UNIVERSITY OF CINCINNATI
______, 20 _____
I,______, hereby submit this as part of the requirements for the degree of:
______in: ______It is entitled: ______
Approved by: ______CRITICAL VALUES: FEMINIST PHILOSOPHY OF SCIENCE AND THE COMPUTING SCIENCES
A dissertation submitted to the
Division of Research and Advanced Studies of the University of Cincinnati
in partial fulfillment of the requirements of the degree of
DOCTORATE OF PHILOSOPHY (Ph.D.)
in the Department of Philosophy of the College of Arts and Sciences
2003
by
Catherine Elizabeth Sherron
B. Phil., Miami University, 1992 M.A., University of Tennessee, 1994
Committee Chair: Chris J. Cuomo, Ph.D. University of Cincinnati
Abstract
Critical Values: Feminist Philosophy of Science and the Computing Sciences
by Catherine Elizabeth Sherron
Committee Chair: Professor Chris J. Cuomo Department of Philosophy
My dissertation is an examination of the intersections between epistemology, philosophy of science, and feminist theory. Feminist philosophy of science creates new and valuable ways of looking at the sciences by using gender as a category of analysis, or a lens through which to critically assess and constructively build projects in science, as well as in the philosophy of science. I employ feminist philosophy of science and a gendered lens in particular to examine the computing sciences. Starting specifically from the underrepresentation of women in computing, the project creates a platform for exploring the dimensions and contributions of feminist philosophy of science. This is not merely a critique of philosophy of science or a feminist review of computing, but a positive project in its own right, examining the epistemological structure of scientific inquiry, including the nature of objectivity, epistemic agency and the composition of an epistemic community, the importance of those epistemic communities, and the role of values in science. A central tenet of the work is that objectivity in science does not require leaving personal and political commitments at the lab door, but that social, ethical, cultural, and other values play a foundational epistemological role in science.
Using gender as a lens uncovers some of those values for critical evaluation. This is not to deny the importance of the natural, empirical world in science. I argue that a philosophical position must at minimum account for our actual relationships—emotional, embodied, social, etc.—in the world and their impact on our theorizing and that dismissing the embodied experience of scientists results in a diminished understanding of the world in addition to diminished epistemological theories.
Acknowledgements
I wish to express my deepest gratitude to my advisor, Chris Cuomo, who could see the merit and completion of this dissertation long before I could. My project and I have flourished in no small part thanks to her.
My other committee members, Bob Richardson and John McEvoy, have been generous, supportive, and tremendously helpful in reviewing drafts of this dissertation, which is much improved for their insightful comments. Errors and confusions that remain are solely my own.
I am also deeply indebted to Thomas More College, its talented faculty and staff, all of whom have been overwhelmingly supportive emotionally, editorially, and financially. I am grateful to be able to continue to working with these energetic people who have taught me so much about the academic life. I wish to thank my students and my colleagues, in particular
Siobhan Barone, Bob Berger, Joe Cronin, Dale Myers, Jim Nelson, Julie Perry, Sherry Cook
Stanforth, Jim Schuttemeyer, and Father Gerald Twaddell.
Many old friends have counseled me through what was at times a rather arduous process.
Thank you Julie, Zac, Colene, Wags, and Jana. Many new friends found at the Women, Work and Computerization 2000 Conference and at the Grace Hopper Celebration of Women in
Computing 2002 encouraged and inspired me. While I did not know the late Anita Borg personally, her passion for developing the presence and talents of women in computing was contagious.
Finally, I wish to acknowledge the loving support of my family: John and Crystal
Sherron, John and Cathy Sherron, Mike and Christel Sherron, Everett Sherron, Christine Young, and especially my new husband, Randy Lee Bailey, for keeping me healthy. I dedicate my dissertation to all of them.
Table of Contents
Introduction 2
Chapter One Stepping Stones: From Feminist Science Scholarship 13 to Artificial Intelligence
Chapter Two Critical Values and the Character of Feminist 34 Philosophy of Science
Chapter Three Computing and Feminist Philosophy of Science 63
Chapter Four Social Epistemology 89
Chapter Five Competition in Science 121
Chapter Six Embodiment and Embeddedness in AI: 161 Drafting a Model of Intelligence
Chapter Seven Computing Women 197
Conclusion Future Paths 221
References 229
1 CRITICAL VALUES:
FEMINIST PHILOSOPHY OF SCIENCE AND THE COMPUTING SCIENCES
This project originated with reading Helen Longino’s Science as Social Knowledge in a philosophy of science course. My undergraduate training had left me somewhat jaded with regard to feminist projects, yet to my surprise, I could not resist Longino’s explicitly feminist analysis in philosophy of science and of science itself. How could there be so much disagreement in science, I wondered? I was convinced by Longino’s argument but perplexed by the conclusion. What was at stake seemed to be the very thing that gave science its power: objectivity. A course in feminist epistemologies and more research uncovered more jewels:
Sandra Harding’s strong objectivity stemming in part from her work with science and technology studies, postcolonial studies, and feminist theories; Lynn Hankinson Nelson’s discussions about epistemic agents; and Nancy Tuana’s work on embodiment in epistemology.
The beginning of my interest in computing is less well-defined. My interest in physics and chemistry was not strong enough to pursue extended study of them, which I thought would be necessary for a grounding of further study in philosophy of science. Biology was interesting to me, but after assisting with a computer ethics course, I decided to pursue a series of courses in programming, data structures, and artificial intelligence (AI). Eventually I discovered that although there had been some exploration of issues surrounding gender in the computing sciences and technology, the approaches were more sociological rather than philosophical, or specifically, epistemological. Thus, there seemed to be a gap in the philosophical literature addressing broader concerns of feminist philosophy of science in the computing sciences, Introduction including AI. Alison Adam’s book, Artificial Knowing: Gender and the Thinking Machine, helped me to start to connect issues in feminist philosophy of science to computing, particularly issues of social, cultural, and ethical values of scientists and their impact on objectivity. It was
Adam’s book that introduced me to the AI project called Cyc.
Attending professional computer and technology conferences such as the Grace Hopper
Celebration of Women in Computing, and the Women, Work and Computerization Conference provided further exposure to computing professionals, mostly women, who were working on problems which I found quite compelling. One set of questions concerned the small numbers of women in computing, why more women were not pursuing degrees and careers in computing, and what, if anything computing professionals, societies, both in and outside of the academy, should or could do about this “shrinking pipeline” as the concern has become labeled. It was these questions which opened up the broader philosophical issues addressed in this dissertation.
What are those issues? In a nutshell, they begin with concerns about the low level of participation by women in computing, and focus on gendered aspects of computing which might explain those low levels, but also serve as the basis of broader and more general critiques of computing. Using gender as a lens for investigation has led me to concerns about how computing is taught, what kinds of projects are appropriately pursued in computing, why one might want to pay attention to the composition of the computing profession, and which epistemological commitments are encoded in those projects and teachings. Christina Björkman’s
(2002) work in gender studies and computing has been invaluable for assisting my thoughts on, especially, the nature of computing and some of its epistemological commitments. A discipline’s guiding metaphor has many values embedded within it and those values guide the teaching, development, and practice of the discipline. For example, Lynn Stein (1999) argues
3 Introduction that computing interpreted through the metaphor of a sequentially ordered series of steps to a specific goal—computation—abstracts too much away from the real processes of computing.
She suggests as a replacement the metaphor of a community of interacting entities.
It is the great power and ubiquity of computing that magnifies the importance of what might otherwise seem like mere academic concerns. Such concerns provide the backdrop and to some degree motivation for more in-depth philosophical discussion of social epistemology, embedded and embodied epistemology, and their relationships to science generally and computing in particular, as well as competition as a social practice in science. The issues are complex and intertwined, and so what I have done is to clarify some of the relationships and to explain why the issues are important. Doing so establishes my robust claims regarding the importance of feminist philosophy of science.
Having an area of focus, e.g., the computing sciences, helps for highlighting the importance and power of feminist philosophy of science. These questions and concerns are raised within the context of computing in order to ground and clarify some rather abstract concerns. This is not merely a critique of philosophy of science or a feminist review of computing, but a positive project in its own right, examining the epistemological structure of scientific inquiry, including the nature of objectivity, epistemic agency and the composition of an epistemic community, the importance of those epistemic communities, and the role of values in science.
Feminist philosophy of science ought to be interested in computing for several reasons.
Feminists are widely concerned with the status of women, which means feminists are interested in how women fare in science, which includes the computing sciences. (I want to note here that although I use “feminist” throughout this dissertation, and do so to connect my work to a specific
4 Introduction history as well as to highlight my explicit political commitments, this work is about gender, which includes both men and women, in all their social roles). Feminist philosophy of science is also concerned with the nature of epistemology and justice. Claims about or arising from an epistemology which threaten or ignore claims of justice is an inferior theory. A third motivation for an inquiry in computing sciences grows out of a general interest in philosophical questions in science. The opportunities to investigate these three concerns about justice, epistemology, and the status of women are all integrated in this project assessing computing science.
OVERVIEW
In chapters one and two I construct and employ three “springboards” for theorizing about science using a feminist lens. The first employs gender as a category of analysis. Feminist scholarship often starts with gender but does not stop at gender. Here I use gender in analyses of the nature of knowledge, particularly scientific knowledge. I champion a position that does not separate values from scientific research. In other words, objectivity in science does not require leaving personal and political commitments at the lab door. While it is legitimate to debate to what degree values permeate scientific research, we cannot progress in science or philosophy by denying that values play a significant role, and examining science using gender as a lens serves to uncover some of those values. While my feminist project does not deny the role of values arising from culture, ideology, personal bias and the like, in the development or practice of science, it also does not deny the importance of anchoring science to the natural world. In doing so, I conclude that feminist philosophy of science provides a superior analysis of scientific knowledge.
5 Introduction
A second springboard is the recognition of the significance of one’s emotional connections with one’s fellow researchers, as well as the subject matter under investigation.
Being passionate can keep one interested in research that can often be tedious, long, and difficult.
It can also make salient subtle details that might otherwise go unnoticed. A passionate commitment to one's work can see a project through the inevitable tough scrutiny and criticism.
Recognizing that women’s lives and women’s bodies—the embodied experiences of all persons—are a source of important information constitutes a third springboard. Dismissing embodied experience results in a diminished understanding of the world and in particular diminished epistemological theories. This account argues that a philosophical position must at minimum account for our actual relationships—emotional, embodied, social, etc.—in the world and their impact on our theorizing.
Chapter three begins with a survey of the degree to which computing is intertwined in our society. Computing increasingly affects more aspects of our lives and affects them more deeply, e.g., in the sciences and biomedical research, in global politics and economics, and in our own self-conceptions and interpersonal relations. Yet women are notably underrepresented in computing. The underrepresentation seems to suggest an unjust situation, one which may have an impact on the computing disciplines overall, including practices constitutive to the disciplines, the particular projects pursued, and the values incorporated into projects and practices. Feminist philosophy of science rightly concerns itself with that underrepresentation in computing because feminist philosophy of science is committed to working for societal justice. These issues are pursued via examination of the relationship between justice and epistemology.
At bottom many of the fundamental questions about science and about technology are epistemological in nature, including questions about confirmation, data and evidence, and the
6 Introduction nature of an epistemological agent in science or technology. Based on arguments developed in earlier chapters I stress the role of values as epistemically foundational to science. Given that computing is neither clearly a science nor a technology, ultimately both philosophy of science and philosophy of technology are useful in different ways for examining computer science. This is appropriate given that a feminist critique can problematize the distinction between practical and conceptual in the first place, a distinction which is often used to differentiate technology from science.
I turn in chapter four to the claim that science, including the computing sciences, develops out of a social epistemology, meaning that the production of knowledge in these disciplines is located in the community, rather than individuals. A (feminist) social epistemology accounts better for the many factors at play in science, from interpersonal idiosyncrasies and ethical commitments, to the use of sophisticated technologies for data collection and the construction of scientific hypotheses. These factors form a social context, including specific histories, to the production of scientific knowledge. Social epistemology is not a sociological study, but a philosophical investigation of the foundations of epistemology, with respect to epistemic agents. The epistemic importance of communities is based on the fact that science is an additive project in which scientists, work over time in networks of groups and subgroups, in their multiple and dynamic social contexts, to elaborate, differentiate, and clarify the knowledge of the past as well as an enormous amount of varied and complex data leading to knowledge of the future. I employ the concept of equipoise to further explain what it means for a community to be an epistemic agent. Briefly, equipoise is a state of uncertainty on the part of individual or groups of researchers with regard to scientific knowledge. None of these researchers can claim to “know” that the treatment (or hypothesis) works until there is consensus
7 Introduction within the broader scientific community. Consensus is most often reached by means of publication, or sometimes presentation, and verification of results. It is not reached by mere political or social maneuvering, but is anchored to scientific evidence and reasoning.
Social epistemology provides major support for my feminist analysis of the computing sciences. Knowledge is particular, not universal or abstract, with social and historical dimensions. If we want to secure objectivity in science, the only way to do it is in and across communities. The community verifies evidence and reasoning, certifying it as knowledge.
Thus, scientific knowledge results from socially negotiated standards of evidence, of reasoning, of experimentation, etc., tested against the empirical world. Standards for good and effective science are discussed and established at the meta-level, in the abstract, beyond any particular project. Using gender as a tool of analysis assists in uncovering assumptions and inaccuracies embedded in those standards, and metaphors in theory construction, as they pertain to specific projects or entire disciplines, which result in harm or neglect to women, men, or children. The example of man-the-hunter vs. women-the-gatherer in anthropology, discussed in the opening chapter, illustrates both the gender symbolism at play and the impact of social context, since the two theories, both of which fit the available data, were constructed by different communities of scientists—a predominantly male group and a predominately female group, respectively.
As individuals embedded within a society, not isolated from it, scientists have a shared stake in the outcomes of science. Building on a argument put forth in chapter four regarding the inherently social nature of epistemology, I argue in chapter five that a model of scientific practice too strongly rooted in hostile or aggressive competition among scientists is disingenuous as well as overly restrictive of both science and of scientists. An approach to science which understands scientists to be lone geniuses working in relative isolation from each other (or at
8 Introduction best, groups work together) sets up a system that not only unnecessarily restricts many persons from becoming scientists, but fails to supply an accurate picture of science in the first place.
That system fails to recognize that lone geniuses do not exist, and that everyone needs others to help refine their ideas. The question here is whether aggressive competition is necessary for maintaining objectivity, which in turn, promotes and protects the integrity of science.
Recharacterizing competitive scientific practice would benefit both male and female scientists by allowing them to work under less rigid definitions and expectations, providing a more accurate picture of how science is actually practiced.
Competition is one factor, among a set of factors, which can inhibit the production of novel ideas; science must generate numerous novel ideas to solve problems, since no straightforward method for deriving solutions to scientific questions exists. Competition can also function to disproportionately discourage certain potential members, such as women. While it is not clear exactly how gender and competition are connected, it appears to be partly related to gendered difference in self-confidence and in communication styles. Both consequences— discouraging new ideas and new members—can compromise the practice of science, and are therefore, potentially detrimental to society. There are two “competing” models at work here
(Longino 1987). The structure of the first model of aggressive competition necessarily excludes multiple winners. Ties are impossible in this model. The game/contest ends only when a winner has emerged. Think of an NCAA championship basketball game: there are not two national champions! It is a winner-take-all model. In the alternative cooperative model there can be more than one winner. Ties are possible in principle, even if they rarely actually happen.
Differences between the contestants usually result in distinct rankings: that runner trained better, this skater was not feeling confident.
9 Introduction
Many feminist scholars have shown how concepts and especially metaphors can structure thinking. The view is that we are particularly susceptible to patterns of thinking in which we organize the world into pairs of opposites, called conceptual dualisms. Examples of dualisms include: Human/Nature; Civilized/Primitive; Male/Female; and Reason/Emotion.
Competition/Collaboration is yet another. The dualistic mindset is not just theoretical: categories can literally impact how we see ourselves and can affect behavior. These conceptual pairs identify what is more (and also less) valuable in society and they appear in models about science. They suggest that persons who are competitive cannot be cooperative, and vice versa. I argue that competition could be serving as a structural barrier, and is not necessary to science; models like that of David Hull’s overemphasizes competition. It is a feminist critique of dualisms that helps us understand the emphasis of competition in models of science, and feminist philosophical critique which can suggest appropriate corrective action, which would work to benefit women, or at least not actively harm them.
More pragmatically, investigating questions about competition in science could help to address concerns about women’s decreasing participation in the computing sciences. In fact, computing culture is complex and is not exclusively competitive, although it is certainly male- dominated. Although competitiveness does seem to have had an impact on women going into computing, it cannot be the sole reason for the “shrinking pipeline.” As Christina Björkman argues, something more fundamental regarding the nature of computing, and especially the epistemological bases, or at least perceptions of such, must be at work.
Reflecting on how gender plays a role in how competitive a scientist seems helps to identify practices which might be discouraged as well as those which should be encouraged. For example, “winning” a verbal exchange by being quick or aggressive—skills typically developed
10 Introduction more in men—might be important on one version of competition, whereas the ability to listen to others, particularly those who are critical of one’s work, might be important in a different model of competition. Similarly, developing a mentoring relationship might be easier for many women, but whether the relationship is encouraged might depend on the model of competition. Another example is proficiency in mathematics or “tinkering” with computers beginning at an early age on the part of young men but not many young women with the result of male college computer science students being more familiar with some aspects of computing which often seems to intimidate intellectually competent yet technically unfamiliar young women. I argue that a reduction in hostile competition opens up discussion for criticism to be taken and offered seriously and sincerely, rather than as a mechanism for, say, advancing one’s one career. Since the valued commitments informing scientific reasoning are subtle and usually invisible in the background, open discussion might reveal commitments that otherwise might remain hidden.
Building on these first five chapters, in chapter six the argument narrows focus to computer science and artificial intelligence. Here, using a variety of sources to help build my case, I argue that good work in AI must view cognition as both socially embedded and physically embodied. The final chapter address a specific AI project, called Cyc, in order to show that pursuing work in AI without thinking of intelligence as embedded and embodied is inadequate.
Chapter six looks at the physical embodiment and social embeddedness of epistemic agents. Drawing from a variety of disciplines, researchers are learning that cognition is rarely something that occurs within the confines of a single person’s brain. Rather, people learn, develop, and produce their work using a wide range of tools within and surrounding their bodies.
They—we—also rely very much on each other for doing our work. Our social embeddedness, as it is called, speaks to feminist concerns about values in science, about the social context of
11 Introduction science, and about the concepts informing science which are inaccurate and can bring about harm to not just women, but men, cultures, society, and the environment.
To give an illustration of how the concepts and issues raised by feminist philosophers of science could be addressed, I turn in the last chapter to review the AI project, Cyc. The argument throughout this dissertation has relied on the premise that diversity of perspective is crucial. At the same time, I am not arguing that no women are competitive, that all men are, or any other broad generalizations. What I do reject are essentialists claims, which divide men and women into binary and opposing groups based only on biological sex, and abstracted away from other social characteristics such as race, culture, class, etc, and then making claims based only on gender, separated from other relationships. A characteristic such as gender, race, or class obviously does not exclusively define any one person. To claim that all members of a class act or believe in one way is to make a claim about essentialism, that all members are essentially and necessarily the same in some respect.
Researchers in computing and AI can learn from the arguments that feminists make, namely, do not assume that your views are shared by others and do not try to represent your beliefs as universally true or universally applicable. Alternatively, try to consider what power relations are at work, who benefits, and who might be disadvantaged by these views. The project under review in chapter seven, I argue, runs a significant risk of presenting its product as something universally applicable when it truly is not. Perhaps it could be, but more careful research, perhaps informed by some of the work in this dissertation, is required to do so.
12
CHAPTER ONE
STEPPING STONES: FROM FEMINIST SCIENCE SCHOLARSHIP
TO ARTIFICIAL INTELLIGENCE
INTO THE RIVER
Feminist philosopher of science, Helen Longino, raises intriguing questions regarding, among many other issues, the influence of money and politics on scientific research and whether that disrupts objectivity in science. The impact of “subjective” factors like politics is still hotly contested in philosophy of science, sociology of science, and scientific venues.1 Medical research for example is not a value-free enterprise, even though we rely on it being free from private interests. Yet, big pharmaceutical companies are increasingly conducting drug research on humans in developing countries, in part to push through the testing phase quicker, which reduces the time until profit (Shah 2002). Similarly, World Press Review reports that the most respected medical journals are becoming concerned with the integrity of research submitted for publication:
The editors of 12 leading medical journals (The Lancet and The New England Journal of Medicine, among others) . . declared that unless certain measures were taken, they could not guarantee in the future the objectivity and independence of published studies [because] research studies are frequently manipulated behind the scenes by the pharmaceutical firms that finance them (Uhlir 2002, 47).
1 Just a small sampling of very recent reports includes: Brown, J. 2001; Longino 2002; Urlir 2002; and Weinberg 2001. Chapter One: Stepping Stones
Given the power of science in the modern world, its status as objective knowledge is not trivial.
Objectivity is supposed to free science from manipulations like those concerning the editors of some leading medical journals, manipulations having to do with political agendas, individual gain, or personal prejudices. Scientific credibility seems to rest squarely on an understanding of objectivity that allegedly protects the public from such abuses.
But is it objectivity alone that protects us? What does “objectivity” mean? Perhaps defining objectivity is not as important as how we account for the biases of those involved in science. Accounting for the biases that do exist will help protect science from charges that it is dissolving into a branch of politics. For Longino a deliberately open system of criticism safeguards the goals of science, one of which includes the production of objective knowledge, allegedly free from prejudices, biases, and ideologies. But Longino’s contention is that background assumptions—including the personal, social, political, and moral values of a scientist—help him or her form connections between empirical evidence and scientific conclusions. This is in contrast to a view of science in which the data “speaks for itself” to any
(interchangeable) scientist, regardless of his or her background, or a view in which evidence is thought to matter much less than, if at all, the interpersonal relations of scientists and their intermediaries. Such background assumptions are usually unarticulated and unacknowledged, which makes them difficult to analyze. Part of Longino’s work, and a major concern of other feminist science scholars, is to uncover those assumptions and articulate them in order to make it easier to discuss how and to what degree they guide science (Wylie 2000, 177).
Longino’s synopsis of other feminist critiques is interwoven into a philosophically dense critique of epistemology in science and proves helpful for understanding how personal, social,
14 Chapter One: Stepping Stones political, and ethical views influence science.2 Many feminist science scholars analyze how values factor into science, which is important for gaining a foothold on the climb to more fundamental questions about the valued dimensions of epistemology. In anthropology for instance opposing positions often account for the same, sparse amount of physical evidence.
This state of affairs demands a way to adjudicate among vying theories if there is to be anything like progress in science. The quotation from Longino below distills the following important point about competing theories. Each theory relies on the same set of evidence, yet draws differing conclusions. In this particular case, the professional disagreement hinges on the identification of who is responsible for the cultural or social evolution of humans. Should the credit go to the hunters, who on many accounts were believed to be the males? Or, are the females, who typically worked as gatherers, the ones who brought about the changes which led to the development of what we now call culture? There is more involved than just lining up the evidence. Those valued background assumptions play supporting if not leading roles, rather than minor or non-speaking parts, in our epistemologies. Longino writes:
The androcentric ‘man-the-hunter’ perspective assigns a major role to the changing behavior of males . . . The gynocentric ‘woman-the-gatherer’ perspective assigns a major role to the changing behavior of females . . . Each perspective assumes the centrality of one sex’s changing behavior (or ‘adaptive strategies’) to the evolution of the entire species. Neither assumption is apparent from the fossil record or dictated by principles of evolutionary theory. Each is an example of a contextually driven background assumption facilitating inferences from data to hypotheses (Longino 1990, 106-7).
Objectivity which simply reads the data is a myth. Yet some of the conceptual components of objectivity are rightly safeguarded, lest political or special interests overrun science. There is a balance to be had in science, but it involves some conceptual rebuilding of objectivity.
2 The illumination comes in no small part from her use of concrete examples to illustrate some rather abstract arguments, something I strive to duplicate, particularly in my final chapter. 15 Chapter One: Stepping Stones
My inquiry into epistemology and how values permeate how we know spans several different areas, including the role of the body in creating knowledge, the importance of the community in which a knower is embedded, the function of values in the production of knowledge and their impact on objectivity, and how those concerns interact with feminist theory.
Throughout this project I try to show how expanding the scope of epistemology to include the embodied and socially embedded context provides a more accurate and useful theory of knowledge. In turn, applying a revised epistemology to science provides a better account of scientific knowledge.
This dissertation builds on the work of Longino and other feminist science scholars, and other philosophers, in drawing connections between several different projects in a range of disciplines. I show that feminist science studies creates bridges across what otherwise might appear to be discontinuous gaps between different projects. Building from feminist theory means beginning with the lives of real women whenever possible and being committed to improving the lives of women. After all, theory should be relevant to the real lives of women.
Many feminist critiques of theory have been based on claims that the theory is not relevant to women. An abstract theory without connection to women’s lives is an inaccurate theory.
Beginning with real women is essential in feminist theory in order to avoid misrepresentation or overgeneralization about women themselves, a mistake made by theorists and activists early on in the feminist movement of the 1960s and 70s. In the past significant differences stemming from race, class, and sexual orientation have been overlooked in order to advance the goal of ending women’s oppression. Unfortunately that resulted in years of marginalization of women of color, of lower socioeconomic status, and of non-heterosexual orientation. Feminists strive for theories that foreground rather than gloss over differences between women, differences that
16 Chapter One: Stepping Stones are often of primary not secondary importance in terms of identity. I attend to the nuanced voices of women working as computer scientists and artificial intelligence researchers and their personal or reflective writings about their work; the voices of women in computing, science, and engineering; and the work of feminist theorists of any gender, particularly feminist science scholars. The heteroglosia of their voices is what I believe to be an important starting point for theorizing.
Questions about knowledge are intimately related to issues of power, and since in
Western culture scientific knowledge is virtually synonymous with power, expanding the scope of feminist science scholarship is a very important project. It is especially important to use feminist criticism in assessing computer science due to the power computer science wields in our culture. The project applies feminist science scholarship specifically to computer science and the development of artificial intelligence (AI), revealing some of the background assumptions that are driving computer science. A great deal of what Westerners know and how we come to know it is a function of science and technology. Increasingly science and technology are based on computing capabilities. Improved computational ability is becoming synonymous with power and success in science, whether one’s research is in astronomy, microbiology, or physical chemistry. We are at the edge of an interesting future with respect to genetics and our ability to modify genotypes to cure disease and to enhance our physical and mental abilities. Such research is only possible with very sophisticated technology, not the least important of which is very powerful computers. Questions about our futures are necessarily involved in determining if and to what extent that scientific knowledge should be used. Such questions are not value-free.
As those values—background assumptions like those mentioned above—are being incorporated
17 Chapter One: Stepping Stones into our computer programs, it seems prudent to examine them before they become so integrated in our lives that they become invisible.
ABSTRACTION AS A VALUE
Researchers in science are often persuaded to some degree by subjective or personal factors such as personal financial gain or success, nationalism, or educational affiliation.
National interest can influence choice of and financial support for a research program or a tool like a computer programming language. In the early 1980s worries about falling behind the
Japanese channeled interest and money into American artificial intelligence research. While the
Japanese were researching their “Fifth Generation” project based on the Prolog language, the
Americans and the British scrambled to get their own projects “optimized for the development of
Lisp programs” (Russell & Norvig 1995, 24).3 It was national interest (in combination with money from industry and researchers’ graduate school affiliations) that helped pick the standard language for research and industry, because the Japanese did not want to be dependent upon U.S. technology, nor the U.S. and British on the Japanese.
Note that the choice of computer language is not unimportant or trivial. One language can subtly encourage certain kinds of research while discouraging others, depending on what is more or less difficult to do in that particular language, as illustrated by this statement from an artificial intelligence textbook: “the format of a language can have a significant effect on its clarity for a human reader. Some things are easier to understand in a graphical notation; some are better shown as strings of characters” (Russell & Norvig 1995, 317). An object-oriented
3 Prolog was developed by the French. Lisp was created by Americans. (Russell & Norvig 1995, 18 & 28). 18 Chapter One: Stepping Stones programming (OOP) language4 such as C++ offers different advantages to the programmer. It allows the programmer to reuse entire sections of code, defined as classes. Part of the motivation behind switching to OOP from C, FORTRAN, or COBAL, is that the programmer can work abstractly, without necessarily knowing much about the specific data involved:
Abstraction, as applied to problem solving, allowed you to concentrate on the problem solution, without worrying about the implementation details. The same is true with ADTs [abstract data types]. Abstract data types allow you to work with data objects without concern for how the data objects are stored inside the computer or how the data operations are performed inside the computer (Staugaard 1997, 37-8).
Abstraction in both computing and in philosophy raises potential concerns and the value assigned to abstraction has itself become invisible in many respects. There are hazards involved with trying to problem-solve without regard to a specific situation. Here is one relatively simple example. When I prepare to teach a course, I can do it rather abstractly without knowing anything about my students. But the class might fail miserably if the material is too difficult or too easy, if the pace is too slow or too fast for a specific group of students. Part of what it means to teach well is that one’s students come to understand the material, and this is something that cannot in the end be achieved if I have only an abstract understanding of who my students are.5
4 To clarify, this is not usually used for artificial intelligence research. 5 Abstraction in medicine and genetic sciences also risks harm. Prescinding from particular persons, one could argue that because fixing genetic disorder X can be done, it should be done in order to prevent a child as well as a family from suffering the effects of condition X. In other words, if we have the technology to “cure” a child, we should. But the idea that we should alter the genetic code to improve the condition relies to a great extent on genetic determinism. Genetic determinism is the belief that having gene Y will necessarily lead to condition X, regardless of the child’s environment. It ignores the fact that changing the code might inadvertently change something else, perhaps resulting in a worse state. Thus, what may seem perfectly acceptable or even laudable in the abstract might have disastrous real consequences. In reality, numerous problems surrounding the pragmatics of using of this technology must be addressed, since doing so assumes certain things about who is using them and on whom they will be used. But do parents or doctors really have the right to alter the genetic code of a child? If so, under what circumstances and limits? The availability and use of genetic technologies has not been subjected to sufficient public scrutiny, but they are not merely products to be distributed on the market. The point is that scientific research and resulting technologies cannot be stripped away from the values contributing to their origin, development, and implementation. See Lori Andrews’ Future Perfect (2001) for more in-depth analysis of the many layers of values, including those of potential parents, doctors, researchers, governments, etc., embedded in the use of genetic technologies. 19 Chapter One: Stepping Stones
Importantly, not only is implementing the abstract difficult, in many respects it is not even a good goal. The difficulty of implementing the abstract occurs also in philosophy, in part because it occurs in our language. In the abstract, this syllogism in modern English seems fine:
All men are mortal. Socrates is a man. Therefore, Socrates is mortal.
Yet as feminist critics have repeatedly pointed out, the problem is that this syllogism is valid only if the instantiation is a man; it does not work if the test case is a woman. The proof seemingly conflates “men” with “humans” as Janice Moulton (1977) shows in her “Myth of the
Neutral ‘Man’”.6 If Sophia is used as the individual’s name instead of Socrates, observes
Moulton, the argument does not work since “men” really is not a neutral generic term. Feminists have consistently argued that abstractions can hide important details, yet abstraction seems to be a hidden value. Lorraine Code (1993) makes the argument about the invisible work of abstraction and the important details concealed by abstraction in terms of highlighting the subjectivity (the gender, race, or age, for example) of an epistemological agent. Making the gender of the epistemological agent explicit is a move that seems odd precisely because we come to expect the abstraction; we often cease to recognize it as an abstraction. We use abstraction even though, as Code notes, the details can be very important to who knows what. My gendered or cultural location or physical abilities (my embodiment) can sometimes make certain things visible to me while hiding others. A person without sight might be able to perceive better with other senses. She might be good at “seeing” when others are making assumptions about their claims based on a shared ability to see, as opposed to a clear train of thought. As in Longino’s
6 Note that in the original Latin or Greek this would not have been a problem because the term here translated as “male” meant “human being” but see also Korsmeyer 1977. 20 Chapter One: Stepping Stones discussion of background assumptions, if the abstracted details are not visible, it is hard to challenge them. Moreover, they need to be challenged because they propagate faulty science.
In “Challenging the Computational Metaphor: Implications for How We Think,” Lynn
Stein, a computer scientist at MIT, argues that conceiving of computation as a series of steps leading to a goal, relies on an outdated metaphor of computation (Stein 1999). Instead, she suggests using the metaphor of computation as a community of interacting entities in order to reconceptualize the work of computing, how computing is taught, and how the computing is employed in other disciplines, namely, cognitive science. It is the abstraction inherent in the conventional metaphor that encourages a limited view of computing. She reinforces my claim above regarding some concerns with OOP, in saying that “The current practice of object- oriented programming has largely been co-opted by the traditional metaphor” (Stein 1999,
485n6). Each agent has abilities, including communicative abilities with other agents. The properties of the system emerge from agents sometimes located in different spaces, and there is no “central command” to give orders. Thus, in the new model, the goal for the programmer is how to “design the community.” That involves identifying the members of the community, specifying how they interact, and which tasks each one can perform (Stein 1999, 484). Stein looks to “cognitive robotics” as a research program successfully utilizing this new model:
Cognitive robotics is an attempt to scale these community-based approaches so popular in robotics into the traditional domains of artificial intelligence: reasoning and problem- solving. The new computational metaphor—in which behavior emerges from interactions, rather than a composition of independent constituents—has a crucial role to play here (Stein 1999, 498).
The old metaphor, while necessary for the birth and early development of computing, abstracts from the real processes of computing, including the myriad ways many computers are embedded physically in the world and socially in other systems (Stein 1999, 475), a phenomenon similar to
21 Chapter One: Stepping Stones one described above by Code, in which important aspects of the subject at hand are oversimplified, resulting in some cases, in a fairly inaccurate model. The new metaphor of interactive computing and the work resulting from it discussed by Stein provide an example of the negative aspect of abstraction in computing at the process level of programming. But abstraction can also be a concern at the level of content.
Inaccurate generalizations lead to inaccurate abstractions that can lead to problems with representations of women in a variety of areas from computer programs to medical research.
General knowledge about women could be abstracted from many particular women, but such generalizations rely on assumptions regarding the “essential” nature of women. When all women are categorized together on the basis of similar characteristics, biology is most often the common feature. This is done without regard to the characteristics that can make any two women differ radically from each other and that therefore can make abstract generalizations false. Race and class are the most obvious of the separators. To clarify the point, a woman of color might have so much less in common with a white woman that the similarities of being female become virtually irrelevant.
Yet if characteristics like race, class, and gender are fundamentally important to how we construct our selves and others and how we understand each other, as feminist philosophers have argued, computer programs representing people need to recognize these as important characteristics. Writing about graphical representation of people in chatspaces, Lisa Nakamura questions the apparent acceptability of making abstractions of people. Remarking on one web space where you can choose how you appear to other members, she writes:
Here, race is constructed as a matter of aesthetics, or finding the color that you ‘like,’ rather than as a matter of ethnic identity or shared cultural referents. This fantasy of ‘color’ divorced from politics, oppression, or racism seems also to celebrate color as
22 Chapter One: Stepping Stones
infinitely changeable and customizable, as entirely elective as well as non-political (Nakamura 2001, 10).
Race cannot just reduce down to a fashion accessory. Race is more than color but by allowing members to choose their “color,” the site designers implicitly condone a view of race as something superficial that is not necessary to include at the most fundamental level of identity.
This is technology seemingly divorced from values, but the decision to not include race or gender is based on the belief that they are not significant enough to be included in the program.
Yet that decision is itself based on values—here, that race or other aspects of subjectivity are not considered important in a way that requires accuracy in representation. When personal characteristics cannot be made visible, the programmers’ values have infiltrated the program.
Some group of persons is designing these venues for interaction, and values are passing unfiltered from the design process through the program to the users.
Problems of abstraction arise not just from the way a program is designed, but from language itself and across different human languages. Uniquenesses in human languages can affect the manner in which scientists are able to present themselves and their data. Just as some computer languages make it easier to represent certain ideas or relations, as mentioned above with C++, human languages have their own hidden assumptions built in. Although this might seem trivial, it can have a substantive impact on the subject of investigation. The language one speaks in the scientific arena opens some possibilities while closing off others. Sharon Traweek, an anthropologist who studied American and Japanese high-energy physicists, highlights certain properties of Japanese language, which leads her to the conclusion that choices made regarding languages have different effects, which are not all equal.
In Japanese one cannot speak or behave properly without signaling the gender and relative age and status of the participants in the conversation . . . Sociolinguists and
23 Chapter One: Stepping Stones
linguistic anthropologists tell us that when all the speakers are fluent in the same languages, the strategic decisions they make together about which language to use in which situation are often determined by issues of power and status. . . . I am also suggesting that to speak about science in Japanese is to choose a certain demeanor and attitude about international science and its American scientific slang (Traweek 1992, 454-5, references omitted).
What is considered the most prestigious work in physics can differ significantly based in part on national culture (Traweek, 1988). Traweek claims that differences in worldviews between Americans and Japanese can account in part for the way research in those countries is conducted, including the kinds of questions deemed important to investigate, the kind of equipment purchased, and the time tables. According to Traweek, Americans are more “high- tech,” purchasing the best of cutting edge technology to use in their research. In contrast, the
Japanese tend to make less frequent purchases and even make different decisions about who can use the equipment and for what purposes. Scientists cannot be abstracted from their cultural contexts. One primary argument in this dissertation reveals some of the hazards involved with valuing abstraction in computer science and artificial intelligence research. This will involve asking questions such as who is doing the abstraction, what qualities or characteristics are being stripped away from the subjects, and what is the end result? Once again, pursuing science in the abstract without regard to its epistemological, social, political, and moral commitments, is simply failing to do good science.
EPISTEMOLOGY AND DIVERSITY
Values—political and ethical—surround the notion of scientific objectivity. Does opening the door to political, social, and ethical values corrupt or strengthen (or both) science?
Chapter Two analyzes questions about negotiating the "proper" place for values in science. For example, supporting diversity is a political and ethical value, but does it have a place in science?
24 Chapter One: Stepping Stones
What reasons are there for supporting diversity of criticism and of practitioners in science?7 If there are not good answer(s) to this question, it won’t matter if some groups and their criticism are kept out while others are welcomed in. Fortunately, there are at least two kinds of response: epistemological and political/ethical. Epistemologically, novel ideas, methods, tests, etc., can arise out of diversity of views strengthening science through rigorous review. Encouraging diversity creates a situation conducive to more explicit and thereby improved communication. If scientists do not share underlying values and assumptions, each party can assist in identifying and articulating the points of divergence. According to Evelyn Fox Keller (2000) language is a powerful motivator, shaping “landscapes of possibilities.” Languages (literally and figuratively) resulting from and informing a variety of worldviews can suggest alternative possibilities regarding which hypotheses to test, what assumptions to examine, and where and when to apply one’s results. Diversity of ideas serves as kindling for improving scientific thinking in myriad ways.
Diversity of group members serves to check the impact of science on a variety of interests, which is important epistemologically but also politically and ethically. One researcher might not realize that the assumption, theory, method, etc. is offensive or harmful to others, while another scientists from a diverse group of scientists, might recognize a potential problem.
For example, having more women involved in overseeing medical research helped to include more women as subjects.8 Programs aiming to increase diversity can assist in bolstering raw numbers of scientists. This is important because the number of scientifically trained and technically skilled persons needed to maintain our technologically and scientifically based
7 On a related note see University of Michigan (2002) and (2003) news releases regarding the importance of diversity on college campuses and the debate over affirmative action programs in college admissions practices. 8 See for example, www.womens-health.org , the web pages for the Society for Women’s Health Research. Visited 24 June 2002. 25 Chapter One: Stepping Stones societies are becoming scarce. If underrepresented groups are recruited, we could have more scientists (Gender Working Group, 2001). Greater diversity does not logically require greater numbers of practicing scientists, but for underrepresented groups it is reasonable to believe that a push for diversity could result in increasing overall participation. On a related point, encouraging diversity among scientists also serves the political and ethical goal of providing equal employment opportunities for women and underrepresented groups (Gender Working
Group, 2001; National Science Foundation, 2000). Actively encouraging and recruiting women and members of minority groups to jobs and careers in science and technology helps empower them financially, socially, emotionally, etc.
Diversity exists among numerous persons, but an individual usually maintains simultaneous membership in different groups frequently identifying now with one group, later with a different group. Memberships are dynamic in that an individual can move between groups without relinquishing membership in any. Those memberships are in flux, changing over time. An individual is a member of one or more racial groups, ethnic traditions, socioeconomic class, gender, sexual orientation, etc. That one has multiple identities, some of which are transitory while others more permanent, results in mapping out diversity in many different and complex ways. One’s standpoint relative to a particular group is relevant to her assessment of knowledge. One of her memberships might invoke a reaction that another membership would not. One’s background and current surroundings—one’s standpoint—can sometimes help one see something that is virtually invisible to another, and that kind of “sight” in combination with the views of many others is valuable to science. Standpoint epistemology is the theoretical approach to knowledge production based on the idea that one’s position in society (especially
26 Chapter One: Stepping Stones those in positions that lack power) provides access to knowledge that others (e.g. those in the power position) are less likely to have. Alison Wylie describes standpoint epistemology:
In concrete terms, what kinds of empirical evidence an epistemic agent has access to, what sense they make of this evidence, what capacity they have to discern the limitations of dominant views about the social and natural world, and what new possibilities for inquiry they envision, may be both enhanced and limited by features of their social location (e.g., the experiences, resources, values, and interests that comprise their standpoint) (Wylie 2000, 178).
Before we had empirical evidence from the sequenced human genome, many thought that we would have a map of human history as well as an answer key to the genetic mistakes causing disease; now we know that the path leading to those keys will be much longer and more difficult, involving not just unraveling the genome, but comparing genomes from multiple persons to get a more complete map, as well as deciphering how proteins are turned on and off and what role they have in growth and disease. The way a scientist goes about pursuing some of these questions can depend in large part upon what equipment and other resources she has at her disposal. For example in less developed countries, scientists cannot always rely on a consistent supply of electricity, an up-to-date library, the funds or the maintenance of the infrastructure making it possible to travel to international conferences, and they alter their approaches in light of these constraints. Similarly, research in computing networks might move in divergent directions due to the local conditions. If the local population cannot afford to have electricity to every home, computers certainly cannot be in every home. Even if the computers were donated, they couldn’t run. Yet this population might have tremendous need to be networked, just not in the way it is done in the modern West. A local researcher’s standpoint might lead her to a
27 Chapter One: Stepping Stones technological or scientific solution to the local problem, a problem that might not have been easily conceivable to someone who takes reliable energy production for granted.9
Although standpoint epistemologists have been sharply criticized for relying on essentialism (Harding 1986, 1998; Wylie 2000), standpoint epistemology does not necessarily do so. Individuals are simultaneously members of multiple groups, some of which might be in conflict. The advantage is in being able to shift from one group to another because it can help one see more clearly the gaps in models or explanations. One sees from one standpoint that those explanations or models do not completely or easily apply to oneself vis-à-vis another standpoint. Multiple actual and possible identities show how persons are embedded socially, culturally, and politically, without committing any one individual to a static or single identity.
Taking account of standpoints can be a way of enhancing diversity in a group, such as a scientific community, leading to improved epistemic vigor in that community. Of course not all criticism is equally valuable and none of the feminist scholars I cite in this work want to give equal time to the KKK, religious radicals, or terrorists and to women faculty at MIT or persons of color working on computer science projects. So, an important question to answer is: who gets to have a say? The answer is other values regulating community interaction outside of science relieve us of having to repeatedly give a group like the KKK equal time and consideration when critiquing science. I argue below in my chapter on social epistemology that science is properly understood as a group effort, with the consequence that the practices forming the group will weed out members who are destructive to the group, points championed by
Longino (1990). More specifically, Longino gives four criteria that together safeguard the social
9 This example comes from a short paper called “Computer Ubiquity - A Mere Fallacy to Most Women in Kenya” delivered on 11 October 2002 at the Grace Hopper Celebration of Women in Computing Conference in Vancouver, Canada by Rachael Wanjiku Mbugua, Senior Lecturer at Highridge Teachers College, Kenya. See www.gracehopper.org for more information on the conference. 28 Chapter One: Stepping Stones processes in science and keep objectivity in science from collapsing on itself (1990, 76). It is the fourth criterion, equality of intellectual authority, which addresses this issue about who gets to count as a legitimate critical voice. Although any group can raise criticism, Longino stipulates that a reasonable time limit must be imposed before a critic must yield the floor, so to speak. If the criticism does not move forward, it will not be considered viable for very long. Specifically, she gives two markers which signal end of time for the criticism. We approach the first “when critical discussion becomes repetitive and fixed at a metalevel” (Longino 1990, 79). Unless there is a prospering, on-going empirical research program, cessation of dialogue at the metalevel will be grounds for dismissing the criticism, thus the second signal is the absence of such a current and productive research program. When the second-order discussion is no longer interesting, the program is dying and it is no longer worth the effort to point out its weaknesses.
SOCIALLY EMBEDDED EPISTEMOLOGY
Chapter Four posits a normative account of an epistemology for science, one that rests on a social epistemology. According to this account a community, in addition to individuals, is a source and repository of knowledge. I introduce the concept of equipoise to illustrate and refine the idea of how communities come to have knowledge. One of the main goals of that chapter is to explain and defend the position that a social epistemology provides a better account of science. “Better” means more fruitful in order to meet a wider range of needs of a wider range of persons and which is based on a more empirically accurate picture of how science really works.
Part of the social aspect requires open dialogue between researchers and critics, including subgroups of the populations that might be impacted by the research. Social epistemology incorporates an explanation of how bias functions in scientific communities and how that serves to strengthen scientific knowledge. As feminists assess how science impacts humans and the
29 Chapter One: Stepping Stones environment, feminist science scholarship has played an increasingly important role in transforming science and scientific practices.
If a community can serve as an epistemological agent, it is important to know how the community is formed (even individual knowers develop in groups—so group and community dynamics are important to consider there too). Another way I examine the values embedded in epistemology is to investigate the accuracy of the assumption that competition is necessary for good science. Using tenets of feminist science scholarship, I argue in Chapter Five that Western scientific methodology and practice champion certain dispositions disproportionately encouraged in men compared to women. If knowledge resides only in the mind, then encouraging competition makes more sense than if knowledge is understood as dispersed across a community, as it is in a social epistemology. But if a community can serve as an epistemological agent, then aggressive competition might harm science, insofar as good science is a practice that must be open to a plurality of views, approaches, methodologies, and questions. Aggressive competition can make it difficult for many women and men to become scientists, which, among other problems, can discourage innovative approaches to scientific research. This is in contrast to
“soft” competition that encourages excellence without stifling others.10
EMBODIED EPISTEMOLOGY
Relying on a naturalized11 approach to epistemology I look to contemporary science for evidence regarding the extent to which the body is involved in knowledge making. Technique and practical knowledge blend with our powers of abstraction to constitute knowledge, and
10Conflating these two types of competition avoids the necessity of a close examination of the reasons contributing to the weak representation of women in science. This is so in part because the void can be written off as an inability on the part of individual women, not to mention many men, to compete. This is a political rather than epistemological concern, however. 11 I use the following definition of Naturalism, taken from Ted Honderich (ed), The Oxford Companion to Philosophy (1995): “the view that everything is natural, i.e. that everything there is belongs to the world of nature, and so can be studied by the methods appropriate for studying that world” (p. 604). 30 Chapter One: Stepping Stones knowers—even experts—often have skills and knowledge that they cannot explicitly codify.
Knowledge is stored in our bodies: the surgeon’s technique, the pianist’s ability, the expert’s intuition. Although philosophical tradition has considered epistemological agents as individual purely rational, thinking things, which might as well be divorced from their bodies, much contemporary scientific evidence does not support a picture of pure rationality.12 Purely rational, disembodied cognition does not exist. A more accurate scientific methodology acknowledges embodied aspects of knowledge and considers them rather than denying or ignoring them. A naturalistic approach allows for integration of current scientific evidence into epistemological theory.
The task of presenting that evidence and building on current work in epistemology and cognitive science in order to structure an epistemology rich enough to accommodate both embodied and socially embedded knowing is the work of Chapter Six. It relies extensively on the work of philosopher Andy Clark to construct bridges between epistemology and feminist theory. I will argue that a socially embedded and embodied epistemology better explains the success of experts than a more traditional account, since the socialized view can locate assumptions that are hidden but necessary to the experts’ functioning and because it fits better with current scientific findings.
Given that rationality is the subject of a set of feminist critiques in philosophy as well as philosophy of science, it should be no surprise that the meaning of rationality is important in feminist discussions of computer science. Rationality is a tremendously important concept in artificial intelligence, as well: agents are created to achieve certain goals, and in so far as they
12 Greene and Haidt (2002), for example, report that “Moral psychology has long focused on reasoning, but recent evidence suggests that moral judgment is more a matter of emotion and affective intuition than deliberate reasoning . . . . These findings indicate the importance of affect, although they allow that reasoning can play a restricted but significant role in moral judgment” (p. 517). 31 Chapter One: Stepping Stones can meet those goals, they are said to be rational agents. My defense of the importance of feminist science scholarship continues in Chapter Seven, by narrowing to the single real life example of the artificial intelligence project called Cyc®. The analysis of this example binds together all the issues introduced above.
Motivation for the Cyc project arises out of problems of expert systems. An expert system is a particular type of artificial intelligence program. It is rational, but only in a formal sense, meaning it can follow rules. Thus its “expertise” is very limited. Querying the system about something outside the scope of its knowledge will fail. It is ironic, really, to claim that expert systems are rational in any broader sense given that they are built upon expertise that itself is rarely rule-based, much less rational. Moreover, these systems are constructed upon an ideal rationality that is disembodied—an ideal itself undermined by scientific research.
The aim of the Cyc project is to create a program that “knows” a huge amount of everyday facts. With that “common sense” the program supposedly could assist the much more narrowly focused expert systems in working together. The long-term goal for a project like Cyc is to have the program learn on its own; but what will it be learning? More propositions perhaps but it cannot learn anything embodied. Chapter Seven is rooted in epistemological questions about what counts as knowledge. An unacknowledged and often unrealized split between embodied and propositional (which is supposed to represent “rational”) knowledge is often lurking beneath the surface in artificial intelligence research. Even when it is acknowledged, these two kinds of knowledge are not equally valued. More perniciously, because the knowledge of women is historically associated with the body, it is devalued, whereas propositional knowledge—historically associated with men—is held out as “real” knowledge, and as such is pursued in AI. The result of this false dichotomy is that the Cyc project and other AI projects by
32 Chapter One: Stepping Stones analogy are ill prepared to understand and incorporate non-propositional knowledge. Fortunately this outlook is changing as an increasing number of researchers are moving into work on robotics and projects involving multiple users and multiple agents, all of which seem to fit very well with embodied and embedded considerations.
33
CHAPTER TWO
CRITICAL VALUES AND
THE CHARACTER OF FEMINIST PHILOSOPHY OF SCIENCE
INTRODUCTION
Feminist philosophy of science includes a complex family of positions at the intersections of natural and social sciences, feminist theory, epistemology, and philosophy of science.
Feminist science scholars share common theoretical assessments of relations of power, along gender lines for example, and analyses of how such power influences and shapes science.
Raising philosophical questions about science is important for many reasons. Feminist concerns arise in part because historically in science and in philosophy gender has remained invisible and so certain questions and practices were never considered suitable for scientific investigation.
The unique role of feminist science scholarship is to develop and encourage increased scrutiny of science by using gender, and other sites of social difference and power, as categories of analysis.
To be clear, feminism does not entail mere reduction to gender issues. It is much broader and more diverse than just gender. In the first place feminist theory could not be about gender exclusively because gender is never isolated from other characteristics such as race, class, or culture (Alcoff & Potter 1993b). Recognizing multiple possible combinations of attributes, many feminists strongly concur with the call to avoid reducing the complexities of human lives and endeavors to any single aspect, one of which might be gender (Cuomo 1998; Jaggar 1983;
Okruhlik 2000; Pearsall 1999b). A one-dimensional analysis along gender lines alone is insufficiently complex, because those categories like gender, race, and class are interwoven. Chapter Two: Critical Values
Difficulties arise from generalizing about any group based on a single characteristic.13 Being a poor woman is a different experience from being wealthier and female, such that two women of differing racial, social, or cultural backgrounds could be more different than a man and a woman of similar social backgrounds. Although issues of identity are important, it will be sufficient for purposes here to limit the discussion to talk about multiple and dynamic roles. Examining the lives of real women is an important starting point, but the analysis must move beyond the personal roles of individuals, to examine the role of women in science without simply reducing them to their gender.
Science commands a position of power in modern Western society. It permeates our institutions, penetrating governmental, educational, social, even trade and international relations policies. The widespread impact of computing in modern Western life also makes computer science a worthwhile subject of study. A computer often seems like merely a powerful or convenient tool, but like the automobile, its power can reach much deeper and further than is noticeable except under direct, purposeful examination (Moor 1995, 1).
Science can also influence cultural ideals (Longino 1990, 164). This is not trivial power.
If it is “scientifically proven” that males are naturally, i.e., biologically, more aggressive than females, some might see that as justification or, less extremely, an excusing condition for behavior like rape (Jaggar 1983). Sherry Turkle has argued that we have come to see ourselves as computing machines (1984, 1997). Turkle and Seymour Papert have said that “both the popular and technical culture have constructed computation as the ultimate embodiment of the abstract and formal” (1990, 128). If computation is abstract and humans are computing
13 Sandra Harding uses “fractured identities” to raise problems for those who generalize about all women (1986, 163-4). A German with Jewish and Christian ancestors who found herself labeled as a “half-breed” by the Nazis perhaps without any knowledge of or affiliation with her Jewish relatives serves as an example of someone with a fractured identity, belonging to neither group. See Cynthia Crane (2000) Divided Lives: The untold stories of Jewish-Christian women in Nazi Germany. New York: Palgrave Macmillian. 35 Chapter Two: Critical Values machines, then in many ways we are coming to see ourselves as ideally disembodied, and merely machines that think. This has many serious implications for human society.
Part of the work of science criticism includes a serious examination of various power relations, and the background beliefs that support them. Helen Longino calls the framework of broad ideological commitments, “contextual values.” These are the personal, social, cultural, ethical values which are present but less noticeable than other aspects of science because they are so often left in the background and given little or no attention. Contextual values are differentiated from “constitutive values”, those that judge accuracy, reliability, and other aspects of scientific methodology and practice (Longino 1990, 4-7). One goal of this dissertation is to make visible some of the contextual values in computer science and in the development of AI.
Although some would claim that the personal characteristics and political, social, ethical values of programmers are irrelevant to the programs they write, I think there are good reasons to conclude otherwise. Scrutinizing how values are present in science traverses many dimensions of science and it leads to numerous investigations. In a later chapter on competition in science, for instance, I examine several epistemological assumptions and methods regarding the appropriate development and inculcation of epistemic agents. I inquire about the influence of gender on epistemology throughout by asking questions about the assumptions regarding epistemic agents, such as are they embodied and are they isolated individuals? Questions about the degree to which science is objectively and/or socially constructed are addressed from various vantage points. Here, following and incorporating the work of many feminist philosophy of science scholars, I argue that science is shaped by values and complicate a more simplistic objective reporting of "truth" by stressing the importance of disclosing those values in order to carefully and openly evaluate them.
36 Chapter Two: Critical Values
DEMARCATING FEMINIST SCIENCE SCHOLARSHIP
Although they overlap and inform each other, feminist criticism is not synonymous with postmodern criticism, social studies of knowledge (SSK), Science and Technology Studies, or cultural studies of science. The feminist criticism that I espouse does not subscribe to the following characterization. In A House Built On Sand (1998), editor Noretta Koertge, describes the social constructivist view which she takes to be broadly representative of feminist science criticism. Speaking for social constructivists, Koertge claims that
In particular, the products of scientific inquiry, the so-called laws of nature, must always be viewed as social constructions. Their validity depends on the consensus of ‘experts’ in just the same way as the legitimacy of a pope depends on a council of cardinals (Koertge 1998, 3).
Koertge correctly asserts in her introductory essay, "Scrutinizing Science Studies," that what ties together the interdisciplinary group of science critics is, among other things, their adherence to questioning scientific objectivity. That claim on its face is true; feminist science scholars do question what has in the past been called objectivity. It does not follow, however, that knowledge claims cannot be assessed for their merit, i.e., as claims related to evidence and methods and not just relative to politics alone. Evidence plays a central role in the type of feminist science criticism I am relying on, in stark contrast to the critics she decries.
Koertge is mistaken in concluding that all science critics believe that "it is futile to exhort scientists and policymakers to try harder to remove ideological bias from the practice of science"
(Koertge 1998, 4). In fact all of the feminist science critics introduced and discussed in this chapter argue that there are many ways to reduce the effect of personal bias in scientific reasoning. Furthermore, many feminist science scholars would probably disagree that the presence of personal bias is always bad. Personal bias refined against empirical data, can be
37 Chapter Two: Critical Values quite productive because it can lead to innovative questions and inquiries of interest to particular populations. Nonetheless, the feminist position I am developing here clearly states that pure scientific objectivity is impossible.
However, it is not the case that because absolute or pure scientific objectivity is impossible that science reduces to base power struggles or political consensus. Londa
Schiebinger responds to the charge that science is either objectively and absolutely true or is merely a tool to be employed by the socially powerful. This response is steeped in the view that scientific knowledge arises from real persons in their material conditions. Scientific knowledge is then refined by methods developed over time to help us draw conclusions that apply to the natural world in ways beyond those that individuals alone can perceive. Schiebinger writes:
Nature, after all, is infinitely rich; there is much in nature we do not know. What we do know is influenced by our history and our values, our national and global priorities; sources of funding and patterns of patronage; the structure of academic institutions, markets, and information networks; personal and professional experiences; technologies and relations with foreign cultures; and much else besides. Culture does not construct reality, but works, as Evelyn Fox Keller has put it, 'to focus our attention in particular ways, conceptually magnifying one set of similarities and differences while dwarfing or blurring others, guiding the construction of instruments that bring certain kinds of objects into view, and eclipsing others.' The goal of revealing gendered structure and polity in science extends the process of continuous critique that is part of the ordinary and remarkable workings of science (1999, 18, emphasis added).
Feminist scholarship often starts with gender but does not stop with gender. The concerns under investigation in these analyses concern the nature of knowledge, particularly scientific knowledge. As such, my feminist project, based on the work of other feminist philosophers of science, does not deny the role of culture, ideology, personal bias and the like, in the development or practice of science, but it also does not deny the importance of tying science to the natural world. In doing so, feminist philosophy of science provides a superior analysis of
38 Chapter Two: Critical Values scientific knowledge. In part it does so by bringing gender and other relations of power to the forefront for analysis regarding their role in the production of knowledge. Work in anthropology, for example, illustrates the role of non-cognitive factors in the production of science. The phenomenon of underdetermination in science regularly occurs but is particularly dicey when the empirical evidence is so slight that most of what is known is theoretical (Curd &
Cover 1998, 255-408). As shown in chapter one, underdetermination in anthropology works to make the "man-the-hunter" theory appear just as plausible as the "women-the-gatherer" theory
(Longino 1990, 106-11). Feminist critics were the first to take the stance that the man-the-hunter theory and its assumptions about women might not be correct.
Another example of a feminist analysis of science that revealed empirical inadequacies involves the disparity in medical research on heart disease in men compared to research on women; there are now laws that require females to be included in research on human subjects
(Laurence & Weinhouse 1994, 60-64). Women have been excluded from clinical trials for various explicit reasons including risk of harm to potential fetuses and the exclusion of elderly patients from trials, even though women with the potential to get pregnant, who have hormone cycles, and who age are treated with kinds of medicines and doses which have been studied primarily in men (Laurence & Weinhouse 1994, 90-1). Does attention to feminist concerns make for better medicine? Yes, when feminist practitioners, researchers, or policy makers ask why there are no female research subjects or whether the results of the male-only studies would be applicable to female patients. We have better medicine now that the concerns and needs of women are more likely to be addressed and researched seriously as “legitimate” problems. As argued by Nancy Tuana, for example, it is the fact of being anchored in one's own subcommunities, which drives one's investigations. It is the collectivity of these various agents
39 Chapter Two: Critical Values and their respective communities that form the valuable system of checks and balances which keeps science healthy.
FEMINIST PHILOSOPHY OF SCIENCE
Does a commitment to feminism alter one’s scientific research or one’s perspective on science? (Honderich 1995, 805). Part of what feminist philosophers of science do is to challenge the boundaries of what science is and who can do it. Although our conception of “scientificity” does need adjustment, I will argue that it is not as radical an overhaul as it might first appear.
There are many feminists working in science and many scientists guided by feminist values, even if they do not identify themselves as such. I want to expand the definition of science to identify explicitly and include more feminist aims. This involves reassessing and possibly redefining what science is, who practices it, under what conditions, and by virtue of what credentials (Harding 1998, 8-12; Cajete 2001). For example, the study of agricultural methods and the environmental sciences are just as much a part of science as is research on human genetics, organic chemistry, or nuclear physics, even though the former are not as “sexy” or well-funded as several of the latter. What we legitimize by calling it science can change. Who can be a scientist can and has changed as well.
According to the politics and social norms of the time and context, women have had varying degrees of access to science, but mostly only very limited access. The way has not been paved for women. Historically many women worked as scientists—often under the auspices of male relatives—but they were not formally recognized as scientists (Wajcman 2003, 138).14
14 Many interesting examples of women's involvement in early science were described at the Metaphysics Into Science: Gender & Knowledge in the Early Modern Period conference jointly held at the University of Cincinnati and Miami University, April 6-8, 2000. 40 Chapter Two: Critical Values
Because there are many factors involved in the separation of women and science, it could take some time before the trend changes course to become easier for women to participate.
How do women scientists participate in science? One response is that women might practice science differently (Adam 1998; Alper 1993; Barinaga 1993; Eisenhart et al. 1998;
Goldberg, J. 1989; Keller 1983; Mahowald 2000). Women scientists from the recent past include
Grace Hopper in computing, Barbara McClintoch in genetics, Candace Pert in biochemistry, and
Jane Goodall in primatology, to name just a few who have been held up as examples of women doing science differently. Many studies report on women in science from philosophical, sociological, historical, or educational perspectives (Alper 1993; Davies & Camp 2000; Harding
1998, 1993, 1986; Herring 2000; McIlwee & Robinson 1992; Rosser 1995; Schiebinger 1999;
Symonds 2000; Travis 1993). But just because a scientist is a woman certainly does not mean that she will act like other female scientists or that she will act differently from male scientists.
Some theorists have concluded that women engage in different science—the means they use and questions they pursue are different than the means used and questions pursued by men.
Of course, this might very well depend on what is meant by science or the practice of science.
Some theorists like philosopher Sandra Harding have been quite direct in asserting the claim that there are separate feminist sciences (1991, 296-312). Other discussions stop short of concluding a separate science exists, but in reporting complex findings clearly indicate that gender and other characteristics do factor into the production of scientific knowledge in a multitude of ways.
Women scientists are not synonymous with feminist scientists, of course, so we need to go beyond simply identifying and counting numbers of women scientists to ask if there are differences between feminist scientists and non-feminist scientists. Unfortunately, since that distinction is rarely if ever noted in a scientific write-up, it is very hard to compare any alleged
41 Chapter Two: Critical Values differences. Biology professor Lisa Weasel, writing for the feminist philosophy journal,
Hypatia, tries to motivate scholars and scientists to give more concrete suggestions for incorporating feminist insights into scientific practice. She claims that “Most practicing scientists remain unfamiliar with feminist writings concerning their fields, and in the rare instances when feminist work is acknowledged within the sciences, it is usually perceived as threatening or irrelevant” (Weasel 2001, 28). One goal of this dissertation is to give some specific suggestions for making AI and computer science more congruent with feminist aims.
Is there a relationship between numbers of women in science and feminist science?
Particularly in computer science the numbers of women seem to be declining beginning with loss of interest in high school: “only 17% of the high school students who took the Advanced
Placement Computer Science test in 1999 were females - the lowest percentage of all tests given.
AB Calculus is up to 47%, Chemistry is 42%, Biology is 56%, and Physics, although still dismal, is over 20%” (Lanius, 2002). From a high of 37% in 1982, female enrollment in university level computer science programs has dropped to 19% in 2000, an especially disturbing trend when women’s participation elsewhere in science and engineering is increasing (Gürer & Cuny
2002).15 It is not clear why more women do not participate in computer science, especially with the promise of interesting, stable, and well-paying jobs and careers.
Simply increasing the numbers of women is rarely, if ever, an adequate solution to the low numbers of women generally in science, because it will likely be only a temporary one.
Special recruiting strategies might be implemented to increase the number of women in the
“pipeline” of those getting degrees in computer science, but it is unlikely that those strategies will result in more women in the field over the long run if fundamental components of computing work are unsatisfying for women. Instead, a careful and close examination of
15 See also Davies & Camp 2000 and Grossman 1998 for similar statistics about women in computer science. 42 Chapter Two: Critical Values practices that are harmful and restrictive to women is needed along with an assessment of how restrictions on any scientist with certain characteristics might inhibit all of science. This marks the difference between merely increasing the numbers of women in science as opposed to informing science with feminism. Thus, even though there are quite a few areas in science where women are well represented (biology is the best example), that does not mean that the science is feminist-friendly, so to speak. Londa Schiebinger, professor of History of Science, observes: "Because modern science is a product of hundreds of years of active shunning of women, the process of bringing women into science has required, and will continue to require, deep structural changes in the culture, methods, and content of science" (1999, 11). If social roles are strongly linked to gender and if ideas about gender govern science, then some social roles might be incompatible with science. For example, if scientists are expected to work long hours, they are not left with much time for family responsibilities. If women are expected, though, to undertake family responsibilities, the social roles for scientists and women are somewhat incompatible. This also means that the role of scientist impedes family responsibilities for men. How should we proceed given that "Women are supposed to assimilate to science rather than vice versa; it is assumed that nothing in either the culture or the content of the sciences need change to accommodate them" (Schiebinger 1999, 4)? These are not merely policy questions. The point is to find out what about the characterization and practice of science discourages both women and feminist values.
Concerns about simple numerical equity can serve as springboards for looking at science under new light. It is due to disproportionate numbers of women in certain fields that feminists thought to look for deeper, structural barriers to women and to feminist values. So, although I will not argue that feminist philosophy of science is fundamentally incompatible with
43 Chapter Two: Critical Values mainstream science,16 it is crucial to note that feminist science scholars have political commitments that motivate them to continue with analysis to improve the status and treatment of women (and other marginalized groups by extension) and to improve science. This is not always the case with scientists or the scientific establishment. Feminist criticism has uncovered perspectives that make visible things that some influential traditions have hidden, particularly the costs associated with being a woman in science and that is in part what makes feminist scholarship valuable to science studies. One key example is that science is not as neutral as it claims to be. Schiebinger describes the contribution of one particular school of feminism as valuable because it helped
to refute the claim that science is gender neutral, revealing that values generally attributed to women have been excluded from science and that gender inequalities have been built into the production and structure of knowledge (Schiebinger 1999, 5).
Illuminating the difficulties is a good first step, but there is more to analyze, which is not a simple task. The bigger project involves identifying problematic deep structures of science and then tracing the impact of those structures:
One would also have to look at other trends in the international structure of science, changes that have furthered goals often associated with feminism, but that may have nothing to do with women or feminism per se, such as the trend away from competition between single investigators to competition between internally cooperative groups (Schiebinger 1999, 12).
I examine in more detail below competition in science as one of those deep structural components of science. Tenure practices would be another component to investigate
16 Louise Antony (1993) makes a similar point about feminists theorizing about epistemology. She argues that feminist philosophy does not have to break radically from mainstream epistemology since mainstream epistemology is a source (à la Quine), which can ground even very radical feminist criticism. 44 Chapter Two: Critical Values
(Massachusetts Institute of Technology [MIT] 1999).17 I look also at the social aspects of creating knowledge in the next chapter.
VALUES
One very important part of the deep conceptual foundation of much feminist science scholarship concerns the role of values in science. Feminist science critics tend to agree that as in all other human endeavors, values are intimately connected to science; there exists no such thing as a value-free science. Science is not neutral, although there certainly is disagreement found regarding the extent to which those values contribute to the development of science.
Many postmodernists see values—personal, subjective factors—as dictating science. Other critics see a balance between the influence of values and empiricism. I align myself with this latter group of critics. Let me stress, though, that values—however conceived—are more detrimental to science when they are hidden than when they are openly displayed. They can more easily escape challenge by remaining under the table.
One reason for the existence of an "old boys' network" is that, in the absence of sufficiently powerful opposition, it “works. ” However, “works” might only be with respect to some, not all, persons. When medical research fails to be applicable to women or children because no representatives from those groups were included in the clinical trials, it cannot be said to work for everyone. Without confrontation things might continue along the way they always have. It is important to stress that the success of this kind of network comes in part from the very real, interpersonal connections forged among the parties. Values—in this case, the personal relationships forged with those sharing basic worldviews—form an important part of the structure upon which science is constructed. That is unless one believes that individuals
17 See also Superson 2002 for specific feminist review of academic tenure processes. 45 Chapter Two: Critical Values working exclusively on their own create science. Even David Hull (1988), who argues that competition is a fundamental component of science, agrees that there are units of scientists who work together to further their ideas, as Schiebinger above noted.
Beyond the role of values in forming and maintaining research groups, political and especially democratic values should be an openly recognized part of guiding science’s efforts.
Openness and responsiveness to criticism—about values, methodology, data, etc.—keep science strong. The history of science shows that scientists make mistakes, so we need safeguards to filter out the mistakes. When the mistakes arise due to value commitments, the most effective or efficient procedure for correcting the mistakes might involve identifying the source of the values at play, rather than, say, arguing about why this set of data does not fit the hypothesis under review. Philosophers have made this point in different ways. For Longino (1990) objectivity is secured by public review: science is social knowledge. Harding’s (1991, 1998) strong objectivity was developed in part to help negotiate knowledge-making between a wide range of standpoints of knowers, contexts, and empirical data. Harding uses standpoint epistemology as the source for a method in science, one that takes differences in culture and other contexts into account without dissolving into relativism (1998, 129). For philosopher Paul Feyerabend objectivity is fundamentally based in theoretical pluralism: “Theoretical pluralism is assumed to be an essential feature of all knowledge that claims to be objective” (1963, 923, emphasis in original). Alternative theories provide criticism that differs in kind from comparison of the received theory with “the facts” and some facts are only available via alternative theories. These are parameters for healthy science. A static notion of objectivity is incongruent with this definition of healthy science. Many feminist science critics recognize that we can never be fully value-neutral in science—nor do they think that it would be desirable to do so even if it were
46 Chapter Two: Critical Values possible—and so spend a significant amount of energy discussing ways to best facilitate critical review of both contextual or background values as well as constitutive value which are used to judge theories by a diverse set of critics. These ideas about reforming science have themselves evolved through critical dialogue among feminist science scholars.
Let me give an illustration to preview how some of these views connect. That the community is crucial to creating scientific knowledge—the socialized epistemology view—is a key to my argument for the seriousness of investigating why there are not more girls and women in science, math, engineering, as well as computing. Redirecting the focus from individuals to individuals-in-community requires examining in detail the composition of and recruitment strategies involved with training members of the scientific community. From a feminist point of view, excluding a significant portion of the population (i.e., women)—intentionally or not— likely results in less accurate and less complete science.
POLITICAL COMMITMENT
These shifts from individuals to communities as epistemic agents, and from science as neutral to science in a context of values, are epistemological shifts, but they have practical and political impact, also. Feminism explicitly engages in political action, such that pursuing a scientific research project can rightly be called into question if its practical effects in the world are politically unacceptable. One facet of the character of feminist philosophy is that feminist arguments “matter not only for the reasons that most philosophers think our claims and arguments matter, but also because of their potential impact on social change that aims to improve the condition of humankind, especially its female members” (Gary 2001, 88).
By paying attention to who does science, anyone can see that, especially in certain fields, there is often a large proportion of persons with homogenous backgrounds doing science,
47 Chapter Two: Critical Values particularly at the upper echelons. Even if, in the final analysis, this situation were to make no difference in epistemology, feminists would still value diversity on a political level, but the point is that the composition of the community of knowers and the science they produce are logically and practically dependent upon each other. Following both Donna Haraway and Sandra
Harding, I claim that position and location of an agent does make a difference in epistemology.
It is not the case that I will necessarily fail to understand the point of view or motivation or explanation of someone with whom I have nothing in common, but we might have to work harder at our communication than two people whose backgrounds are more similar. Having to work hard to be attentive to others is could be very beneficial because it might prove to be more fruitful than communication between people who think that they understand each other—those who assume they share values or beliefs that in fact are not really shared. Communication at that deeper, more attentive level might reveal subtleties that remain hidden in the traditional system and thus, unavailable for inspection. So it is both the process and the result of the feminist challenge that is desirable.
A second line of inquiry often found in feminist science criticism investigates the background intellectual and personal commitments of scientists, in addition to the explicit identification of values and their role in the pursuit of science. Philosopher Lynn Hankinson
Nelson envisions a refreshingly different understanding of science:
The science we should work to bring about will be different from the conception many scientists and philosophers have about science. . . . Unlike much of the present practice of science in which values are incorporated and we simply deny that they are, the incorporation and infusion of values will be self-conscious and subject to critical scrutiny (1990, 5, emphasis added).
Discussion of personal values, social values, scientific values, etc. is necessary as part of a thorough critical review of science. Such a call for disclosure quite obviously presupposes the
48 Chapter Two: Critical Values existence of values in the personal, social, and scientific arenas. Disclosure contributes to good science in that it can make it easier to determine points of disagreement and because it can draw in a wider range of evidence and hypotheses.
While the claim that values do play a role in science might seem to undermine objectivity--a lynchpin in scientific success—this is only if objectivity is understood narrowly.
An objectivity based “only on the evidence” and which denies the contribution of values is narrow. It is brittle in a sense because one can only identify the evidence and fit it to a theory in so many ways without reaching for explanations arising from commitments of value. Including value dimensions is much more complicated but provides many more candidates for explanation.
When scientists are seen as independent from society they are more easily understood to be immune from the influence of values. In a subsequent chapter on socialized epistemology, I strongly challenge the notion of knowers as individuals. Understanding epistemic agents to be completely autonomous with regard to creating knowledge is empirically inaccurate and epistemically undesirable. Knowledge generally and scientific knowledge in particular is more successfully explained as a community enterprise. Mistakes can be corrected, new models or directions for experimentation can be suggested by others working in concert, sometimes collaborating, sometimes criticizing, but always working towards a better understanding of the empirical world. This understanding includes how science itself works.
Both Nelson and Longino are perfectly clear about the presence of values in science as well as in the relationship of science with the wider community. Nelson writes:
Our larger community, including its values and politics, is, in fact, a science community. This result has clear moral implications. If a science community encompasses all of the activities and practices with which special scientific communities are inextricably related, and if, as feminist criticism indicates, there is no separating science from values, or science from politics, then good science must incorporate the taking of responsibility for the directions and use of knowledge
49 Chapter Two: Critical Values
developed and certified in scientific communities, and self-conscious and critical attention to the values incorporated in scientific theorizing (1990, 15, emphasis added).
Science impacts values and politics, and in turn they influence science and each other. Feminists concerned about the community are necessarily concerned with science’s impact on the stability, integrity, and beauty of the community.18 Ignoring the role of science within the larger communities is foolish for anyone committed to the ideals of improving the material conditions of persons, given the strength and permeation of science into the core of society. Not surprisingly, this raises a difficulty: how to balance the demands of values and politics within and against the special methods of scientific inquiry? One solution to this problem is to use the inherent social nature of science to fashion an anchor of objectivity.
Objectivity almost always harkens back to empirical data, how it is used and what qualifies as data. According to Lynn Hankinson Nelson (1990) feminist science criticism poses three general kinds of questions for empiricists to answer, each of which “involves a connection between the social identity of scientists and the content of science” (p. 32). The first includes questions about sex and gender, another set highlights politics and its role in science, and the third inquires about social arrangements and their possible impact on science, positively and negatively. These three sets provide categories for thoroughly examining the many interactions of science, values, and social power. Donna Haraway (1997) and Sandra Harding (1998) focus on questions from the third set. Strong objectivity strengthens the practice of science by making way for alternative, innovative methods and approaches (Harding, 1998). Haraway refers to expanding the scope of science as "a re-tooling of vision.” She claims that unless vision—here a metaphor for knowledge—is acknowledged as situated it will not ever really be able to “see” the
18 I am adapting a description applied to the ecological community by the environmentalist Aldo Leopold (1949). 50 Chapter Two: Critical Values world. Like in a real body, where I cannot see all of myself directly, my embodied knowledge/vision is limited. My knowledge is situated in my body. There is a need to recognize our limited epistemological views. In doing so we can join our knowledges in the social realm, thereby creating an improved objectivity, as suggested by Longino.
Why would science be resistant to something that could improve it? There are political values at work clearly in resistance to the feminist claims I discuss. For one, the "reforms" challenge long-standing assumptions about what science is and what objectivity requires.
Implicitly those opposed to feminist criticism assume that only scientists can practice science and only scientists have a right to critique it, where critiquing is a privilege requiring credentials, which can only be attained in specific ways. But rethinking the definition of science to explicitly incorporate values and context opens up the discussion to all kinds of contributors. That does substantially complicate the situation, yet simultaneously it brings a wealth of resources to science.
Science’s success results from evaluating hypotheses until finding something which works and withstands the scrutiny. Evaluation can come in many forms, one of which is brought about via critical review by others in the field, those who are affected by the particular research, and others who have a vested interest (positive or negative) in the project. Feminist criticism in science plays one of these roles. If the criticism does not harmonize with the empirical results, then perhaps it can be ignored. But given that many feminist critiques have successfully challenged even apparent empirical successes, it must be encouraged. Nelson clearly explains why:
bearing in mind that the rationality of science is partly a function of its being open to criticism, should empiricists consider the issues that feminists are raising about the sciences? Prima facie, the outcome of considering these criticisms would seem to be that, whatever their origins, these criticisms will enhance our sciences and our
51 Chapter Two: Critical Values
understanding of science, or we will find them to be off the mark or vacuous. If they are on the mark, then, of course, the questions of their origins and the inability of current views to accommodate such origins will need to be addressed. We will need to find a way to account for the connections between sex/gender and science, and politics and science, that, unlike current empiricist accounts of science, does not assume that such factors are either irrelevant or inherently distorting (Nelson 1990, 35, emphasis added).
As mentioned above in anthropology, at one time the leading theory argued that males engaged in certain activities and that those activities were ones that led to the advancement of culture.
The theory seemed to fit the data at hand. It was only after the conceptual criticism of feminists that it became apparent that there could be other plausible theories to account for the data.
Critical review of that hypothesis and many others has resulted in a change in the way science is done. We are now somewhat more sensitive to how gender, class, and race, for example, drive assumptions and practices in science. Some practices have changed in light of this and similar criticism. For instance, the United Nations Commission on Science and Technology for
Development established a Gender Working Group in 1995 to advise it on such things (Gender
Working Group 1995). Some of the recommendations include: tracking the statistics regarding the many aspects of women in science and technology across national boarders; encouraging the recruitment of women into science and technology; incorporating gender as a category of analysis in policy, programs, and projects; and assessing the impact of science and technology according to gender (Gender Working Group 1995). The call to assess the impact of programs specifically recognizes, for example, that international development policies have not always taken gender into account, and that they should, because the impact of the policies often falls more heavily on women than men or falls equally to women and men, but the women are not approached to participate in the decision-making process for implementing the policies (Oldham
2000; Ufomata 2003).
52 Chapter Two: Critical Values
In her 1998 Is Science Multicultural? Sandra Harding makes a persuasive case about the motivating factors driving science, arguing that it has usually been local needs at a fairly practical level that have triggered scientific research and development: traveling across this geography; using these native food sources; or searching for cures for those diseases. In our global capitalist economy, the marketplace and the need to sell, rather than searching for truth or pursuing a love of knowledge, is what often seems to motivate and regulate science. Especially in computer science and biotechnology what sells is what is pursued in research. Of course, there is nothing revelatory here: this has happened repeatedly in the past and will continue to do so in the future in capitalist and other settings. My point is to reinforce other feminist science critics in their claims that to see law, politics, and economics as separate spheres is to misinterpret what science really is and to desire a purity about science that does not and cannot exist. My feminist project strives to improve science and possibly though not necessarily rectify the underrepresentation of women in science along the way.
Perhaps some other element within science could eventually facilitate a reform, but feminist criticism makes the issues very explicit and, in a way, urgent. The political, ethical, and social dimensions are highlighted and put out on the table for discussion. Feminism makes a space and a regular time for these important discussions, whereas science usually brings them up only at special times and places that are outside the routine process. A case in point is the
Human Genome Project (HGP), which set aside some money for ELSI (ethical, legal, and social issues) having to do with the Project. It is unclear to me whether the findings of the ELSI group have had real impact on the research. Stem cell research and gene patenting, for example, are still moving along with apparently little integration (at least until very recently) of the concerns of those working on policy. The expansion of technology into more areas of our lives and
53 Chapter Two: Critical Values particularly the increasing dependence upon computerization are other examples. In the case of something like stem cell research, which is controversial for many reasons, the explicit discussion of ethical practice has come to the forefront and in a very public way. But such discussions are still lagging behind the actual research, and in other cases like computerization the lag is severe, when it really should be integrated at many levels. The eight recommendations of the UN Gender Advisory Board attempt to provide a comprehensive structure for assessing such integration.19
Values permeate science. This is the clear conclusion from much recent feminist science scholarship, philosophy of science, postmodernism, and social studies of knowledge (SSK). All of my subsequent chapters expand on this basic conclusion in the service of building a broader argument about how a feminist philosophy of science analysis of computing and AI yield very important theoretical as well as practical results. Having taken some time to discuss the general theoretical framework for feminist philosophy of science, I now move to examine some major themes as well as theorists in the field.
THEORETICAL SCAFFOLDING
Theme: Objectivity
Helen Longino, Lynn Hankinson Nelson, Sandra Harding, and Nancy Tuana, four leading feminist science scholars, interrogate both the concepts and methodologies of science and philosophy of science. Often feminist philosophy of science begins with an analysis of gender. Again, this should not be confused with a baseline examination of numbers of men and women or with other theoretical tools used by feminists. While discussion of values in science is
19 It turns out that Sandra Harding, surprisingly or not, worked with the Gender Working Group of the U.N. Committee on Science and Technology for Development (Harding 1998, ix). 54 Chapter Two: Critical Values today more common in the philosophy of science than in the past, feminists are still more straightforward about explicitly acknowledging the influence of social, political, and ethical values in science and engaging in discussion about how to retain important structures like objectivity and rationality. Although these scholars disagree with each other on many issues, questions about gender and values are treated by each as germinal in the process of revaluating the often-exalted position of science in society.
Philosopher Helen Longino is widely cited in discussions about objectivity in science.
She argues that intersubjective interaction safeguards objectivity enough to keep science productive without allowing political values to dictate the specifications of scientific research.
She claims in "Cognitive and Non-Cognitive Values in Science: Rethinking the Dichotomy"
(Longino 1996) that traditional cognitive values in science, such as simplicity, explanatory power, and empirical adequacy are not neutral values, but can take on socio-political valences in different contexts. For example, what does “simplicity” mean? In one context it might be concise—as in a theory that is easily stated. But it could also mean how intelligible the theory is to a non-specialist. But even what appears as simple could be due to the way the evidence is presented, and other equally valid ways of presentation can quickly complicate the issue. For example, deciding between taking a medical treatment or not can be a function of whether the related statistics are presented as a survival percentage or a non-survival percentage.20 Another meaning is ease of fit of the theory to the data. Acknowledging that cognitive values are not fixed in meaning opens the door for recognizing the influence of non-cognitive values from the realm of ethics or politics. Members of the broader community rely on those other values to know how to use and judge something like “empirical adequacy.” The community negotiates
20 Thanks for Bob Richardson for his assistance in reinforcing this point concerning the alleged distinction between cognitive and non-cognitive values. 55 Chapter Two: Critical Values how these values are used, within a variety of constraints given by the physical world.
Recognizing the need for flexibility helps create a rich, dynamic, and open environment, one that is best for the critical review required for good science.
For some feminist science scholars, strengthening science requires interrogating the universality of scientific knowledge, a characteristic often claimed for objectivity. Sandra
Harding (1996) develops alternatives to understanding knowledge as universally applicable and universally available. She argues that recognizing the existence and power of multiple local knowledges allows us to get a better picture of the world and can help us interact more effectively in the world. Local knowledge addresses particular local problems using the resources of the local epistemic and physical community(s). A community near a body of water has certain types of physical resources at its disposal and presumably has developed a certain body of knowledge about that local area. Solving transportation, housing, medical, economic, or agricultural problems is done within that community using those resources. Those local solutions to such problems could very well differ from the solutions derived in and by a different community. Harding draws on tools found in multicultural and global feminisms to assess philosophy of science and to build the argument that a multiperspective science is more effective than a uniform, comprehensive account.
Lynn Hankinson Nelson who co-edited Feminism, Science, and the Philosophy of
Science (1996) writes about metatheoretic issues involved in the intersections of the disciplines involved with science, philosophy, and feminist theory, focusing on the epistemic agent in science: as individuals, groups, or some combination. Additionally she is concerned about what constitutes evidence. Nelson advocates a holistic position, both with respect to the group as knower and to a broad empirical grounding of knowledge. She calls explicitly for the use of
56 Chapter Two: Critical Values values when judging the merits of a theory. Accounts of science that relegate values to somewhere outside the empirical realm are empirically inadequate on this view, because of underdetermination of theories, because of meaning holism, and due to her naturalistic view.
Nelson’s naturalistic approach does not sharply demarcate between science and epistemology generally, nor between science and the role of values. As discussed above, non-cognitive values shape cognitive values.
I champion a position that does not separate values from scientific research. This is one of the most important contributions of feminist science scholarship and cannot be overstressed.
Objectivity in science does not require leaving personal and political commitments at the lab door. While it is legitimate to debate to what degree values permeate scientific research, we cannot progress in science or philosophy by denying that values play a significant role.
Theme: Springboards
Feminists have called for a focus on the lives and work of women as starting points—as
“springboards”—to theorizing. In applying this to science, feminist philosophers have examined the lives of scientists, medical practitioners, and of women, even if they lacked the formal credentials, working with or among scientists (Berman 1975; Ginzberg 1987; Keller 1983;
Longino 1987; Tuana 1996). Nancy Tuana (1996) discusses three epistemic differences found in the process of examining the work and lives of women scientists, in her "Revaluing Science:
Starting from the Practices of Women." In addition to broadening our understanding of the practices of scientists, Tuana, like Harding and Longino, is committed to widening the circle of scientists to include persons from a diverse set of backgrounds. This is presented as a way to make science better, in part by becoming applicable to a wider set of concerns.
57 Chapter Two: Critical Values
The first epistemic practice concerns the dynamic situation of knowers within communities. Given that an epistemological agent is usually "simultaneously a member of different epistemic communities and subcommunities," Tuana’s model depicts knowers as travelers between various communities (Tuana 1996, 22). Knowers are each simultaneously members of different communities, and communities can be defined by a variety of commonalities, including academic discipline, research institution, affiliations from schools.
Subgroups can then be formed within those communities based on other characteristics such as cultural background, nationality, gender, etc. The role of a knower could differ depending on the context—which subcommunity is most relevant at the time—because context can enhance or inhibit epistemological access. Communities teach knowers to value some beliefs, practices, and things more highly than other beliefs, practices, and things. Those values recommend (or inhibit) certain lines of questioning that protect or enhance (or threaten) what is valued. By investigating the community’s values, critics might be better positioned to understand why researchers do what they do, but more importantly, assess the assumptions they start with in building their cases for why such and such result should become a part of scientific knowledge.
The epistemic impact on knowledge by the community is inevitable, but it is usually also invisible. For example, when computer programmers make assumptions about characters in a game, about how to assess pain, about what is to be included in “common sense,” regarding who can use a program or how users would do so, they are perhaps unreflectively creating programs based on the values held in their respective communities, but not necessarily broadly held across Re the above comment. You might add examples from Merchant multiple communities (Adam 1998, 2000; Rommes 2000). Many feminist science critics have on Bacon's sexist methodology and Easlea on the sexist Manhattan Projectg. See Merchant's The Death of Nature and provided examples of how background assumptions fit into the reasoning processes of various Easlea's Fathering the Unthinkable. I don’t know how far back in the history of science you want to go but there are scientists (Harding 1998; Keller 2000; Longino 1990; Weasel 2001). Assumptions of one plenty of examples in it of sexist theories.
58 Chapter Two: Critical Values community can be revealed when they clash with different assumptions made by different communities.
Sandra Harding echoes this point that we can make more sense out of the process of science by referring to the cultural values that inform researchers, their communities, and their projects. Harding does this on an even broader and more global scale than Tuana. In Is Science
Multicultural? Harding introduces detailed cases in which local needs have encouraged certain directions in science, while dismissing other directions as not worthy of the community’s attention. In her words:
Sciences’ problematics are shaped by their supporters, sponsors, and funders and, more generally, by what is interesting to those groups willing and able to have their concerns conceptualized as ones for systematic empirical research. Bordering the Atlantic Ocean, one group will want to fish it, another to use it as a coastal highway for local trading, a third to use it for trans-Atlantic emigration or trading slaves for sugar, a fourth to desalinize it for drinking water, a fifth to use it as a refuse dump, a sixth to use it as a military highway, and a seventh to mine the minerals and oil beneath its floor. These differing interests have created culturally distinctive patterns of knowledge (and ignorance) about this part of nature’s regularities and their underlying causal tendencies (Harding 1998, 65).
In these ways theorists like Tuana and Harding provide reasons for promoting diversity among epistemic communities.
Importantly, this is not an essentialist claim. The claim is not that non-feminist scientists will necessarily all reason from the same assumptions, or from the same set of assumptions to identical conclusions. But we might be able to identify some assumptions if we look to the groups from which the scientist emerges. If we can then identify the values in the assumptions, it lets us hold them in check while we evaluate the other evidence supporting the claim. The values themselves are open to scrutiny as are other cognitive assumptions and other data that is gathered in support of a claim or theory. The availability of a robust scrutinizing process is a key
59 Chapter Two: Critical Values point in Longino’s Science as Social Process. When the data, the reasoning, etc., including the social and political values embedded in the methods, have been examined, only then the community is in a position to determine if the results are worthy of becoming part of the scientific knowledge base.
A second springboard is to consider the epistemic value found in affective processes— emotions—something historically dismissed as counterproductive in science. The absence of affect from science is really a myth. There are numerous examples of terms heavily laden with emotion and subjectivity. Consider the following examples just from computer science: we
"kill" a program; in object-oriented programming, "families" consist of classes that "inherit"
"behavior" from their "parents"; and, finally, if you anger someone you might get "flamed."
Even "subject" and "object" of research have emotive connotations. Evelyn Fox Keller presented an example of emotionally connected research when she wrote about the work of
Barbara McClintock. The “wars” among scientists also attest to the passion involved in science.
David Hull’s 1988 Science as a Process certainly attests to heated disagreements sustained over years, involving large groups of scientists. More examples of very impassioned involvement with scientific research include the race to identify the structure of DNA (Watson & Bragg
1991), early research to identify HIV (Shilts 1988), and most recently, the contest between publicly and privately funded research efforts to map the human genome (NOVA 2001). No one - even the most fervent positivist would reject the idea Furthermore, Nancy Tuana correctly points out that emotional involvement with one's that emotional involvement is an important motivational factor in science. But this is irrelevant to the feminist claim research is not necessarily nor even usually a detriment. Being passionate can keep one that they shape the content and form of science. This paragraph seems redundant to me, unless you modify it to make interested in what can often be a long, difficult process. It can also make salient subtle details something like the point I'm making in this comment. that might otherwise go unnoticed. A passionate commitment to one's work can see a project through the inevitable tough scrutiny and criticism. Tuana attributes the patience to develop the
60 Chapter Two: Critical Values ability to comprehend complexity to a passionate commitment to one's subject. Too much emotion at one time can override rational processes, but emotion is also a powerful motivator.
Eliminating emotion does science a disservice.21
A third springboard from women’s lives into critical assessment of science arises from the role that the body plays in epistemology, and so also in scientific theorizing. An emotional connection might be one example of embodied knowing. Tuana provides a terrific illustration of knowledge arising from the body:
the example of the engaged, embodied scientist we find illustrated so well in the primary care health provider offers a more adequate model of scientific inquiry than the image of the detached, disinterested, autonomous scientist (1996, 30).
You might want to point out that this notion of 'embodied Epistemological theories that cannot account for—or worse, just ignore—our connections to knowledge' is implied by the idea of "implicit knowledge" contained in Kuhn's conception of a paradigm. This other people and entities are empirically inadequate. Since all scientists are embodied, we must dimension of Kuhn's philosophy of science was also used by 'constructivists' in SSK, who identified science as a 'craft account for how embodiment factors into the creation and development of scientific knowledge. activity.' You might check out Jan Golinski, Making Natural Knowledge, Chapter Six in particular presents scientific evidence regarding all of the faculties used when pp.16-18. people think, make decisions, respond to a stimulus, etc. Some of these faculties are very connected to our bodies. Not only is epistemology embodied, it is also socially embedded.
Chapter Four presents research about learning and human cognition to support the claim that humans do not learn or function well, if at all, in isolation. An understanding of science that fails to account for these aspects of embodied and embedded epistemology is empirically impoverished. Ignoring these kinds of relationships of knowers to the world is insufficient for a theory of human epistemology.
21 See also Marjorie Grene’s (1969) Introduction to Michael Polanyi’s Knowing and Being, for an overview of Polanyi’s tacit knowledge, meaning perception or skills, which emphasizes personal participation of the scientist with the subject of study. 61 Chapter Two: Critical Values
Integrating science with the philosophy of science and integrating feminist theory with philosophy have similar results: breaking down disciplinary boundaries and using the strengths of one area to reexamine ideas in another. I take Nelson’s argument to point to where we should start in the philosophy of science. That is, a philosophical position must at minimum account for our actual relationships in the world and their impact on our theorizing. Empirical inadequacies are sometimes the result of a practice of science which has failed to concern itself with those actual relationships. Thus a feminist critique might call for more empirical data or a review of the data offered as the basis for some conclusion, but the call might just as well be for conceptual clarification or a refocusing. New empirical data might bring about a questioning of the theory from a new point of view, but it seems more likely that because of some point of view—a standpoint, as it is called in feminist and other epistemologies—the evidence is seen to be related in some way. This discussion about the critical role feminist philosophy of science plays in theorizing and evaluating science, serves as the foundation for feminist philosophy of computer science and AI, to which I turn next.
62
CHAPTER THREE
COMPUTING AND FEMINIST PHILOSOPHY OF SCIENCE
Of course feminist philosophers of science should be interested in computer science and artificial intelligence. Computing is ripe for investigation because it has become entwined in our culture, swiftly and directly impacting much of society, transforming how we live and even saturating our views of ourselves (Turkle 1984; 1997). Computing sciences22 are tremendously influential in our society (Björkman 2002, 10-11). Computer scientists Christina Björkman and
Lena Trojer in an article on gender research in computing describe the impact:
Computer science (CS), as one of the core disciplines within the broad area of information technology, has become one of today’s most important disciplines by virtue of its influence on the shaping of technology and thus also society. There is little technical research, development and production done today that does not, is one way or another, involve results from (mostly in the form of applications of) CS. Computer science thus strongly influences the direction and content of technical research and development. It is reasonable to assume that this influence of CS on the current and future developments of technology will continue to grow, and that the discipline will continue to be located at the centre of information technology. This centrality means that what happens within CS will have effects that reach far beyond the discipline as such, having consequences for the whole of society (Björkman & Trojer 2001, 77).
Those current and future consequences include substantial impact on global political and economic structures (recall Y2K concerns) and worldwide development. Additionally, computing touches on individual human lives by way of changes in communication, political
22 I will use “computing science” or “computing” to jointly designate computer science and artificial intelligence research. Chapter Three: Computing organization, and scientific research (especially in fields like genetics), to name just a few areas.23
Yet women are notably underrepresented in computing (Björkman 2002; Camp 2002;
Davies & Camp 2000).24 The underrepresentation seems to suggest an unjust situation, one which may have an impact on the computing disciplines overall, including practices constitutive to the disciplines, the particular projects pursued, and the values incorporated into projects and practices. Feminist philosophy of science rightly concerns itself with that underrepresentation in computing since feminist philosophy generally is committed to working for justice in society.
Computing affects all our lives, but feminist scholars will be particularly attentive when computerization disproportionately and negatively impacts women. Doing so requires using gender as a lens of analysis. This analysis involves investigating the nature of computing itself, its structures and guiding concepts (Björkman 2002), and the pedagogy of computing education as well as determining which projects in the computing sciences are pursued, how they grow and develop, and whether they negatively affect women. Since women do not participate equally with men in computing, and since gender is rarely if ever considered relevant in computing, it is women who are more likely to be at risk. What is harmful to women is likely to be harmful to society, which is the concern of everyone. As I have argued in chapter two, when contextual values are dismissed as real components of science the risk increases since without recognition of something worthy, they escape regular, thorough review.
Issues surrounding gender often function as contextual values. By failing to consider the various ways gender (and race, class, etc) positions one in the world, programmers might
23 If technology and science extensively shape our society, then those who design technology and who research in science contribute greatly to the shape of our society, and they are also subject to critical review. In chapter seven I offer a critical analysis of a particular project, Cyc®, and its creators. 24 See further discussion below in the section, Computing Women. 64 Chapter Three: Computing inadvertently bring about harm to someone by assuming that everyone acts or reacts in the same way. Racial stereotyping can contribute to real harm of persons. Sexist portrayals can also contribute to a hostile environment, as when advertisers use computer generated models or pictures of women constructed from pieces of photographs of real women and assembled in a photo editor program (Kilbourne 2000). Abstraction, which was discussed in chapter one, helps to make the omission of important characteristics less noticeable. Any time a programmer incorporates information about people, she has a responsibility to make sure that her own biases are not masquerading as generalities. It is important and, philosophically speaking, perfectly legitimate, to seriously consider such factors, namely, the values permeating the planning, undertaking, testing, and review of the computing sciences.
The relative lack of women in computing seems indicative of several things. It suggests a continued presence of negative gender stereotypes and gender inequalities in our society. It reveals something about the nature of computing science: what it is and who should practice it.
The lack of women in computing also portends a potentially worrisome future. If the trends continue, those who can influence society via computing will likely be far too homogenous and the economic and political power they command far too great. A commitment to activity in service of truly open opportunities in the computing sciences (and elsewhere) as well as political equality for all, and the theorizing that makes all of that possible, defines this endeavor as feminist.
Although there has been work in computing and philosophy (e.g., Boden 1977, 1990,
1995, 1996; Clark 1997, 2001; Dreyfus 1972, 1979, 1992; Glymour 1992; Haugeland 1998,
2000) as well as feminist critiques of who uses computers (e.g., Adam, Emms, Green, & Owen
1994; Balka & Smith 2000; Rothschild 1983; Webster 1995), there has not been much combined
65 Chapter Three: Computing work of feminist theory and philosophical issues in computing. Björkman (2002) comes close with her assessment of epistemologies in computing, but philosophical analysis is not her orientation. That absence has in part motivated this feminist philosophical critique of computing. Analyses of gender and other power dynamics have proven fruitful in areas such as law (e.g. Dworkin 1997; Estrich 1987; MacKinnon 1989), social and political theory (e.g. Frye
1983; Jaggar 1983; Superson & Cudd 2002), and psychology (e.g. Brown 1994; Enns 1997;
Gilligan 1982; Noddings 1984; Worell & Johnson 1997; Worell & Remer 1992); these citations denote just a few areas and a handful of authors.
A feminist computing science analysis promises much, pursuing the issues via an examination of the relationship between justice and epistemology, which provides a venue for investigating the status of women in computing. Building on earlier scholarship my project explores a rich variety of questions using a range of approaches, including new epistemological stances. Feminist criticism provides new tools and frameworks for investigating the dearth of women in computing, whether their absence impinges on the substantive work of the discipline, and whether computing projects are compatible with feminist goals like reducing oppression.
Questions about who shapes technology and how they do so are fundamentally questions about power and knowledge, which makes them epistemological. Each of the subsequent chapters in this dissertation explores some aspect of epistemology in science, from questions about who or what constitutes an epistemic agent in science and what social conditions are most conducive to growth in scientific knowledge, to the embodied and socially embedded aspects of epistemic agents that must be incorporated into complete models of cognition. These questions are asked and answered in order to establish the claim that feminist philosophy of science can uncover
66 Chapter Three: Computing heretofore invisible or obscured problems and offer unique solutions to concerns in science, philosophy, and the greater society.
I am delineating an important role for feminist philosophy of science by constructing feminist critiques of computer science. My eye is toward the wider project, while being vigilant for issues that are unique to computer science. Avoiding the challenge of philosophically analyzing computing risks missing an opportunity to offer constructive criticism to a youthful, but extremely influential industry and the theoretical work upon which it is based. Given that computing is neither clearly a science nor a technology, ultimately both philosophy of science and philosophy of technology are useful in different ways for examining philosophical questions in the computing sciences. This is appropriate given the feminist critique problematizes the distinction between practical and conceptual in the first place, a distinction which is often used to differentiate technology from science.
DISCUSSIONS OUTSIDE PHILOSOPHY
A commitment to feminist philosophy of computing is certainly not meant to dismiss the important and insightful work of non-philosophers with respect to philosophical concerns.
Clearly not all those interested in the computing sciences and the function of gender in computing are philosophers. Outsiders can be useful to a philosophical investigation, however, for their ability to see straight to the heart of sometimes-obscure questions.
For example, Elaine Bernard has wondered whether it is worthwhile to investigate why women are dissuaded from entering academic computing.25 It might not be that women do not like or cannot succeed in computing per se, but rather that computing as taught and practiced in academia is hostile or intolerable for many women. Perhaps the culture of academic computing
25 Professor Bernard of the Harvard Trade Union Program, Harvard University, United States presented the opening plenary remarks at the Women, Work and Computerization 2000 conference in Vancouver. 67 Chapter Three: Computing should be assessed for its impact on women. Bernard suggests that we might be asking the wrong questions or looking in the wrong places for answers. A different strategy would assess the success and/or impact of women who are already in computing without having completed a degree program. Couldn’t it be, she suggests, that women are in computing, getting along well and making contributions without being present in undergraduate programs or elsewhere in academia? Her point is not that the low percentages of women participating in computer science is acceptable, but rather that those numbers alone do not support the assertion that women do not impact computing. Have academics, worried about the “shrinking pipeline” of women into computer science degree programs of study at the university level, neglected areas of relative strength for women? If so, how and why did that happen? Perhaps there are plenty of women in computing who got there by bypassing computer science degree programs. If that were the case, we ought to investigate how did they did so and whether those paths could be expanded for others. I am not convinced that they can, due to deeper conflicts embedded in computing with regard to gender, to which I now turn assisted by a historian of science and a computer scientist.
Londa Schiebinger, a history of science scholar, has written extensively about women in science. Schiebinger (1999) draws conclusions about gender in science based on evidence gathered across many disciplines (history, philosophy, sociology, etc.) and about many kinds of science (physics, biology, anthropology, etc.). She reminds those working in gender and science studies that "women" is a complex category and that neglecting to carefully analyze the various connections will lead to cheap answers: “Women’s historically wrought differences from men, then, cannot serve as an epistemological base for new theories and practices in the sciences.
There is no ‘feminist’ or ‘female’ style ready to be plugged in at the laboratory bench or the clinical bedside” (Schiebinger 1999, 8). We need to start with answers to questions about
68 Chapter Three: Computing epistemology in science and about the relationships of gender and epistemology. Given the fact that gender is such a complex concept, Schiebinger is thus certainly correct to claim that “[w]hat is needed is a critical understanding of gender, of how it works in science and society” (1999, 13, emphasis added). Getting a grip on this critical understanding is one of her major aims. To this end, she develops a set of conceptual tools for analyzing gender in its various forms in science.
By looking at both the culture and the content of science, she shows how the concept of gender functions in science, affecting outcomes, subjects chosen for study, institutional arrangements
(e.g., gendered patterns of institutional authority regarding tenure and promotion practices), language, theoretical frameworks, and even the very definitions of science (Schiebinger 1999,
186-191).
In addition to Schiebinger’s research, I find the work of Alison Adam and Frances
Grundy, and Christina Björkman to be particularly helpful.26 As practicing computer scientists engaged in feminist science criticism, they have much technical experience upon which to draw.
Using the work of trained computer scientists provides a measure of reassurance that we philosophers have avoided gross misunderstandings of the issues. The potential for misunderstanding the technical side of a discipline is one very good reason to make sure that those well versed in the discipline, (regardless of their commitment or not to feminism) have open access to those very critiques. Adam, Grundy, and Björkman live and work outside the
United States, and thus their perspectives further broaden the base of this project. In combination with philosopher Sandra Harding’s work on sciences in the developing world, I am
26 Similar research, presented at the Women, Work and Computerization (WWC2000) Conference at Simon Fraser University, Vancouver, British Columbia, Canada, June 8-11, 2000, can be located in the conference Proceedings, edited by Ellen Balka and Richard Smith. Proceedings from an earlier WWC conference are listed under Grudy in my bibliography. 69 Chapter Three: Computing developing a comprehensive position about women and computing sciences, but not necessarily one that pertains exclusively to American society.
I have been arguing that asking expansive questions about fundamental components of science is a hallmark of much of the work done by feminist philosophers of science.
Importantly, they are not just concerned about gender, nor are they developing what used to be called “feminine science.” The critiques delve deeper into the roots of how we see ourselves and how we function in the world, as well as the nature of computing.27 Since science does play an important role in human society, it is not surprising that feminist scholars are drawn to investigate it. These feminist computer scientists have worked hard to uncover gender in the methods, institutions, outcomes, and language of computer science, showing that gender does not disappear in computer programs, it merely becomes encoded within a program.
Adam especially raises questions about the epistemological assumptions behind artificial intelligence programs and expert systems, arguing that such projects often fail to consider how such assumptions can exclude and sometimes even harm certain persons, usually those whose backgrounds are quite different from the backgrounds of the programmers themselves. She has claimed in her Artificial Knowing: Gender and the Thinking Machine (1998) that gender, specifically masculinity, is encoded—“inscribed"—in artificial intelligence. The view of human intelligence, upon which these AI programs are based, is overly simplified and abstracted away from or stripped of embodiment. This abstraction ostensibly erases gender, presumably becoming gender neutral, yet Adam argues that the “fingerprints” of masculinity are actually
27 Christina Björkman in particular stresses moving beyond a simplistic solution which merely adds more women into computing, in no small part because this solution has been tried and has failed. Another motivation for deepening the search is the belief that a viable solution to what she terms “The Gender Question in Computing Science,” following the work of Sandra Harding, can only come via an examination of the disciplines of computing themselves. See particularly papers IV and V in Björkman (2002). Delving into a deep discussion about the nature of computing, while very tempting, is beyond the scope of this project. I do, however, intend to pursue the questions at another time. 70 Chapter Three: Computing retained. A neutral gender is a hidden masculine gender, and even though the subject represented in an AI system is rarely described in any detail, it is often assumed to be capable of substituting for any person. Adam argues that one consequence of building systems that utilize these ubiquitous subjects is the loss of responsibility for the use of one's knowledge, and also to erase aspects of epistemological agents that might be associated with femininity.
More recently, Adam has shifted to the broader area of information technology (IT) to argue that the role of feminist criticism in IT is important for ensuring that technology becomes and remains empowering for a broad range of users (Adam 2000b).28 This reinforces a call to keep science accessible to a variety of persons with a variety of needs. All of Adam's work in some way goes to support my claim regarding the importance of acknowledging the role of contextual values in science. I turn in the next chapter to the argument that science, including the computing sciences, develops out of a social epistemology, meaning that the production of knowledge in these disciplines is located in the community, rather than individuals. The social account is better equipped to negotiate a path between the valued dimensions of science on the one hand, and the empirical constraints of the natural world on the other. I then explore the implications for an epistemology that takes physical embodiment and social-situatedness seriously, and in doing so raise the problems of ignoring gender, which tends to result in masculinity becoming the default gender.
Frances Grundy, another feminist computer scientist, poses several interesting challenges to two ideas often believed to be key in computer science.29 One questions the wisdom of basing
28 Another future project will involve a more precise definition of computing. For now, it is important to make some general claim about a variety of disciplines, including computer engineering, information technology, software design, human-computer interaction, AI, etc. I realize, however, that more needs to be said about whether all of these can really be categorized together. Specifically, future research would need to examine the underlying structure and conceptual foundations of these disciplines to see if the generalized critiques really hold. 29 Both of Grundy’s papers, "Mathematics in Computing: A Help or Hindrance for Women?" and "Where is the Science in Computer Science?," were presented at the WWC2000 conference. 71 Chapter Three: Computing computing science curricula on mathematics. Another questions the claim that computer science ought to be considered a science. That leads to the question of whether computer science is properly analyzed by philosophy of science. Let me address the second issue next.
SCIENCE OR TECHNOLOGY: WHAT IS COMPUTING?
Given that computing is neither clearly a science nor a technology, ultimately the resources of both philosophy of science and philosophy of technology are useful in different ways for examining the computing sciences. Employing both is appropriate since many feminist critiques problematize the distinction between practical and conceptual in the first place, a distinction which is often used to differentiate technology from science. Another similarly false dichotomy I reject is that science is objective, while technology is subjective. Just as science and technology have both objective and subjective dimensions, so does computing. Still, investigating computing as a science allows me to compare computing sciences with other sciences in order to ask questions like “Why aren’t the patterns of participation by women in computing more like those found in biology or medicine rather than physics?” Although it is tempting to see the answer(s) as either related to the difficulty of a particular science or to a social setting of a science, context I not so easily or properly separated from the “real” work of science. The question is not an either-or one. Again, the issues are of a more fundamental nature. The focus is on developing and employing the tools of feminist philosophy of science, especially Londa Schiebinger’s conceptual tools of gender analysis, highlighted above, to examine computing sciences. Yet, to reiterate from above, while I am constructing a feminist critique of the computing sciences, my eye is toward the more universal project of positing the critical value of feminist philosophy of science.
72 Chapter Three: Computing
Computer science is not exactly a “hard” science, but its position within the science- technology continuum is otherwise not well-defined.30 Yet even if computing were "merely" a tool, i.e., a technology rather than a science per se, it would still be worth investigating for the same reasons that investigating mathematics or logic is important, namely because they are systems of knowledge. I will follow philosopher Joseph Pitt’s argument in Thinking about
Technology: Foundations for the Philosophy of Technology (2000) that science and technology are intimately intertwined, and suggest that the computing sciences are the clearest example of disciplines which cannot separate their science and technology components. In most sciences, technology in the form of powerful computers is almost always at work, often driving the science
(Björkman 2002).
Research in technology has been financially lucrative, and computing has certainly been profitable for many. Success in the market appeared limitless, at least until the winter of 2002, when the U.S. financial markets began their slide. Computers, as a case in point, clearly have been very successful with consumers and it seems safe to say that no one really cares if they were created by scientists, technologists, or both. Genetic therapies and other biotechnologies, in contrast, stand to benefit tremendously from legitimization in the form of scientific approval and conversely suffer without it; witness recent debates over cloning. People are scared of potentially harmful scientific processes and new technologies, but with the endorsement of the scientific community that fear is allayed somewhat. When the Raelians claimed that a cloned human baby had been born, the world looked for scientific proof of their claim.31 Not surprisingly, none has been offered. Note that there is apparently something important
30 Salmon et al (1992) introductory philosophy of science textbook classifies philosophical investigations into computing as part of the behavioral and social sciences. Historically, as a branch of mathematics, computing is also quite naturally at home within philosophy, given the place of logic in the study of philosophy. See also note 22 above regarding what is meant by “computing.” 31 See news stories by e.g. Suzuki 2003, Edwards 2002. 73 Chapter Three: Computing technology needs from science—endorsement of safety or legitimacy perhaps—whereas the reverse dependency is not as strong—neither the public, nor scientists themselves look for an endorsement from the technology sector about scientific advancements.
An examination of medicine might help clarify some of these claims. The goal here is to establish an analogy between computing and medicine, neither of which is clearly pure science or mere technology. Then by showing that medicine can justifiably be considered a science, I conclude that so too can computer science. At least some of the factors impacting the status of women in computing disciplines are similar to those generally affecting women in science.
More importantly, however, is how those factors impinge upon the way science and technology are conceived and practiced.
In the computing sciences, as in medicine, basic sciences provide frameworks for more applied technologies. Sometimes it is hard to discern what is basic and what is applied in medicine, but following Arthur Zucker in his “Can Medicine Be a Science?” we can separate out at least three areas. They are: the fundamental sciences (biology, chemistry, and physics), medical research, and clinical medicine (Zucker 1996, 240). Like medicine, the computing sciences certainly do encompass technological developments, yet are not comprised merely of technology. Computing sciences also incorporate basic science components. As medicine relies on basic research in chemistry and biology, so too does computing rely on basic research in physics and chemistry, since computing depends upon hardware development that relies on those basic sciences. Research on semiconductors for example might be closer to basic research than work in hardware layout and design. Mathematics and logic of course play a crucial role in
74 Chapter Three: Computing computing sciences as well. Research in AI might involve basic research in cognition whereas work in human-computer interfacing (HCI) is more applied.32
Should medicine be studied under the philosophy of science? It is sometimes since questions about causation, explanation, hypothesis construction, validation and confirmation, the nature of evidence, and so on, found in philosophy of science are clearly applicable in medicine.
Given that medicine deals so much with persons and that persons are complex, sometimes appearing chaotic like weather, it might be helpful if medicine were considered its own subcategory. Medicine might seem to differ from other sciences in that "totally precise and accurate prediction is impossible in clinical medicine" due to having persons as subjects of study
(Gorovitz & MacIntyre 1996). Complexity turns out to be another similarity between AI and medicine since AI also has much complexity, in part because it is also modeled on human
(intelligence).
Some might object to classifying medicine as a science because it is too deeply emeshed with values to be "objective." The proper blend of subjective and objective components in science certainly constitutes an important topic in the philosophy of science. However, as I argued in earlier chapters, objectivity does not mean value-free. In fact, value-free is not a real possibility for any human endeavor. Exemplary reasoning—using cases or paradigmatic examples—works in science but the models one uses can shape the form of the science. Even though medicine involves human patients, who are complex, this does not make medicine any less of a science than physics. Perhaps prediction is not as easy here as in classical physics—
32 In chapter four I make connections between research in cognition, as reported by Andy Clark and John Hagueland, among others, and research paths in AI, showing how basic research becomes blended with technology. Overlooking such research could result in a skewed understanding of some aspects of epistemology. If such misunderstandings serve as the basis for work in AI, the work itself will be flawed. Similarly misconceptions about gender might become folded into the fabric of computing, particularly when issues about gender itself are not considered relevant. 75 Chapter Three: Computing even though that might also be a myth itself—but by using exemplars, predictions can be made
(Caplan 1996). Arthur Caplan gives these concluding remarks in a paper about theoretical reasoning in science:
Philosophers of science use physics as the clearest example of science. They focus on Newtonian mechanics, astronomy, and relativity, not on civil engineering, hydraulics, or acoustics. If they did, they would get a different picture of physics, a different sense for 'particular' --one that was close to medicine. . . when theory is applied to pressing practical problems, the resulting area of study (medicine and certain areas of engineering) will not look like the purely theoretical areas of physics (Caplan 1996, 266).
Medicine is a science, even though it may not look like physics and even if researchers do not always use the same patterns of reasoning: science encompasses many different types of activities.
Computer science can thus also be classified as a science. It is analogous to medicine by virtue of, for example, work in pattern recognition and certain methods of problem solving.
Especially in artificial intelligence research, much of the hard work revolves around getting the agent (the robot or program) to filter out important information from “noise,” using the proper patterns. The agent must be “taught” or programmed to recognize patterns that are significant in some context, e.g., identifying tanks in satellite photos of military camps or discerning pre- cancerous masses from normal tissue. For the computer scientist, devising appropriate ways to construct and program the agent can be substantial research projects in themselves. One question a feminist analysis must address is what about the computing scientists themselves?
How might gender influence their work?
76 Chapter Three: Computing
COMPUTING WOMEN
Women in computer sciences are relatively few, at least with regard to numbers attaining formal education. Accommodation does not appear to be happening satisfactorily in computing, engineering, or physics where women’s participation hovers around 20 to 30%. Yet there are areas of science, such as biology, anthropology, psychology, which are more successful at recruiting, developing, and retaining women as scientists. Classifying computing as a science provides a means to assess similarities and differences with these “successful” sciences.
Although success cannot simply mean greater numbers participating, that might be a good start.
Quite a few assessments have been made resulting in several areas of research regarding women in computing. There are groups studying the uses of computers. For example, work in labor studies and civil liberties assesses how computing impacts jobs and alters the status of work after the introduction of computing (Harris 2000). Of course the impact is not always negative. Wagner (1993) provides a case where computerization of nursing can help to define the tasks of the job, help plan work to better meet the goals of the job, and otherwise help to
“make visible” the scope of the job not only to the practitioners themselves but also to others in surrounding areas. One might also consider the use of supercomputing power to “crack” the human genome or for data mining, in the context of genetic research. Computing can make available information otherwise unknowable. Linking health and medical records to genetic maps can only be done effectively by means of supercomputers. The powerful programs used in the databases create serious concerns regarding safeguarding privacy (Jonietz 2001; Taubes
2001). Should computers be used for these kinds of projects? How might that use impact women? Such concerns are not value-neutral.
77 Chapter Three: Computing
Other researchers inquire about who uses computers and the users’ status. Women who use technology tend to do so in low-level supporting roles, whereas men tend to be involved in the development and use of “sexy,” cutting-edge technology, so when women do work in computing it is often with accompanying lower status and/or pay. It is not clear, however, that the introduction of computing necessarily makes existing gendered divisions of labor any worse
(Adam 2000; Spertus & Pine 2000; Webster 1995). Studies like these do suggest ways to mitigate negative outcomes, e.g., through institutional support or supplemental training in important skills thought to help women succeed and become more visible in the field, move up in the ranks, and recruit other women into the field (Craig et al 2000; Grossman 1998; Sanders Reference data on percent 1995). These studies’ recommendations are closely aligned with the results and suggestions for female higher degree in CS, etc. increasing women’s participation in engineering, sciences, and mathematics (Massachusetts
Institute of Technology [MIT] 1999; McIlwee & Robinson 1992; Rosser 1995).
Looking beyond the empirical data, what do the employment patterns reveal about who can (or not) do what kind of science and what is the science supposed to be? Many theorists have recognized, for example, that simply putting more women into the mix does not guarantee a long-term solution, unless and until the sources of the original exclusion are addressed, which involves looking for underlying patterns and assumptions (Björkman 2002; Harding 1986, 55;
Schiebinger 1999; MIT 1999). Yet unless one were motivated by political or ethical concerns regarding the status of women in the job market, it might not seem worthwhile to look at who works in computer science. If gender is considered irrelevant to employment trends in computing, then recruiting measures won’t likely be reviewed for their effectiveness in reaching girls. In other words, if recruiting targets are conceived of involving merely as genderless individuals, but other factors create expectations of male workers, those expectations will likely
78 Chapter Three: Computing be carried through in the methods of recruitment and then on into the computing culture (Adam
2000).
Political and social equality is often predicated upon financial independence, so upon discovery of the relatively few women in engineering and computer science programs—both of which have excellent job opportunities—an inquiring scholar might wonder why there are not more women present. Some, in the name of equality, would argue that men and women are interchangeable in these jobs, and that gender is not a relevant factor. But, given that we are socialized to be fairly different and that the work environment for a computer science or engineering job—the culture of computing—is heavily masculinized, it is disingenuous to view computing as a neutral environment. I am not claiming that only men can do computer science or engineering, but that the way students are taught and trained in these professions might very well favor what are considered as masculine characteristics. Londa Schiebinger makes this point by saying that the culture, education, and practice of science appears neutral, even when it is not.
Moreover, women are expected to fit into it, rather than ask that it to change to accommodate potentially different needs.
If women are excluded—voluntarily or not—from computing, they are cutting themselves off from important financial, political, business, and leadership positions in the future. They simply will not have the requisite experience needed to qualify for those positions.
A similar, and even more serious problem, is occurring at present with regard to economic status, whereby prosperous nations and groups within nations are getting computer training and setting up sophisticated computing systems and leaving the rest in technology “ghettos.” Internationally women are being left out of and/or are choosing not to be involved in computing (Kuosa 2000;
79 Chapter Three: Computing
Peiris, Gregor, & V 2000; Symonds 2000). Should this be cause for alarm? If we are only concerned about sheer numbers, or lack thereof, of technically literate persons (of whatever gender) prepared for jobs in the computing industry, the answer is a resounding yes. But of course I am arguing that the problem is even more serious:
CS [Computer Science] continually loses women at all stages of the pipeline including elementary, middle, and high schools, college, graduate school and beyond. Thus, the computing industry has lost access to a large pool of potential computer professionals. . . . Our preliminary results suggest that we will see a minimal increase in the number of women graduating with CS degrees in the next few years with a drastic decline between 2001 and 2002 (Davies and Camp 2000, emphasis added).33
In the UK and elsewhere, women's representation in higher education computing courses continues to run at around 10%, a significant decrease from the figures of the late 1970s and early 1980s and which shows little likelihood of improving (Adam 2000).
In Australia and the USA there are now more female undergraduates--right across the board, except in computer science and engineering, which is an issue we need to address--than there are males (Spender 2000, emphasis added).
It is not entirely clear why this is happening. One suggested factor, as mentioned above, is that the academic environment of computer science is unappealing to women, because it is (or is viewed as) a “masculine” area. What might that mean? Computer work might appear to be
“overly rational.” The usual translation for “overly rational” is devoid of emotion—work which many people, not just women, find undesirable. It is sometimes alleged that men are more suited for computer work because they, more so than women, are disinterested and lack emotion, traits that serve them well in a sterile, antisocial world of computing. However, Frances Grundy
33 The details of their findings are as follows. "Specifically, our results show that women earned approximately 16.70% of all bachelor 's degrees in CS last year (1998-1999), a slight increase from 1997-1998. Similarly, we predict women to earn 16.86% in 1999-2000, and 17.23% in 2000-2001. Again, we see slight increases in the percentages from the previous year. However, in 2001-2002, our results indicate that only 16.27% of all bachelor's degrees awarded in CS will be awarded to women, which is a huge decrease from the 17.23% in the previous year." (Davies and Camp 2000). 80 Chapter Three: Computing challenges the anti-emotion assumption by raising the issue of violence and aggression, for example, in gaming, programming, and computing at large. These are examples certainly full of emotional response, even though they are not usually what is meant by "emotional". “Overly rational” is often associated with mathematics, as mentioned above. Specifically, the problem has to do with creating a computing science curriculum based on mathematics. The position of math as a gatekeeper to continued study in computing sciences might block girls from participating and perhaps results in perpetuating a masculine domination of the discipline.
Grundy (2001) wonders if it really is necessary to have mathematics anchor computing science curricula. Girls are not pursuing science, mathematics, technology and computer science in middle and high school and so are cutting themselves off from college work that requires such a background.34 Part of the reason for the lack of interest at the secondary school level appears to be social, rather than educational, in nature. In other words, girls seem to pick up from society that such pursuits are not desirable, and/or the girls see such pursuits as requiring too much sacrifice of what they (the girls) really want, e.g., "normal" work lives with time for families and social lives. Importantly, it is not what their teachers per se are or are not doing that is discouraging girls (Sanders 1995, 150). This explanation provides some evidence for the view that science and technology are, or at least are perceived as, predominately masculine and that as such, not appropriate places for most females. From the August 1998 issue of Scientific
American:
According to a survey of Association for Computing Machinery members, factors that may have contributed to the shrinking supply of female computer scientists include less prior experience playing computer games as children; the long work
34 The Computer Professionals for Social Responsibility (CPSR) website, for example, relates the following: "A recent Washington Post article written by Victoria Bening reports than in recent report, prepared by the Fairfax School Board's Human Relations Advisory Committee, shows that girls only make up 26 percent of the students in computer science classes at Fairfax County high schools. The alarming gender gap in Fairfax mirrors a national trend of young female underepresentation in high-tech classes" (Albright 2000). 81 Chapter Three: Computing
hours common in many programming jobs; gender discrimination; the lack of role models; and the antisocial image of the typical computer hacker (Albright 2000).
In addition to how computing scientists are trained, other researchers suggest that computing outside of academia is inimical to women: "The Internet and the World Wide Web are actively and aggressively hostile to women. Not the technology itself, but the attitudes of the people who are using it" (Eubanks 2000). Eubanks argues that within cyberculture there is a pervasive view of women as subordinate. Disturbingly, the view equates computers with women, such that it becomes permissible to treat them in the same ways, resulting in the dehumanizing and devaluing of women. She writes,
Man uses and dominates computer. Therefore, man uses and dominates woman. This is the pervasive and persistent metaphor working barely beneath the patriarchal and misogynist attitudes that poison so many women's experiences with the Internet. This is how the metaphor of Woman=Body, Man=Mind is perpetuated. And it effects how men and women relate to each other on-line. It makes the Internet just a high-tech place for men to harass women (Eubanks 2000).
If it were just the social factors rather than the nuts and bolts of computing that discourages women, they might be circumvented. For example, we could establish networking groups and use the Internet to do so, showcasing professionals and their roles and lives. But if computing is outright hostile to women, as illustrated above, circumvention involves stepping outside a whole culture, a project likely to fail. Eubanks identifies the "frontier" metaphor particularly as one culprit in a view of computing that contributes to an environment that is hostile to women:
Since the turn of last century, Americans have conceptualized their sexiest new communication or transportation technology as the frontier. Stick with me now, it's not as trivial as it seems; the metaphor comes with a certain set of icons and historical clichés, like homesteading and gunslinging and gold rushing, which help dictate how the technology will be socially integrated. American frontier rhetoric, in both its historical and contemporary incarnations, is both deeply contradictory and shockingly consistent. On one hand, it professes the ideals of self-determination, democracy,
82 Chapter Three: Computing
individual freedom, universal possibility and connectivity. On the other, it authorizes selfishness, profiteering, lack of community responsibility, colonialism, and violent conquest.
Each of the ideals highlighted by Eubanks impacts the broader community, and if women are not involved with the development of computing research and the creation of computing technology they are not equally accessing social and political spheres and the accompanying influence.
Beyond “mere” employment or numbers, the concerns are about who is shaping technology and, thus, influencing the broader culture. Christina Björkman stresses that the concern is deeper than how many women are working or not in computing. She makes this point quite succinctly: “This argument implies that women are regarded as a reserve labour force. In itself, this is not a neutral argument: are women a concern only in their capacity as a ‘reserve’, i.e. when there are not enough (talented) men?” (Björkman 2002, 110). She locates the problem in the discipline itself, rather than with the women who cannot succeed in computing:
What seems to be lacking in many discussions, is deliberation of the ‘nature’ of the discipline itself, i.e., computer science and its knowledge processes. Thus, the issue of female under-representation within the discipline takes us right into the heart and core of CS paradigms and understandings. How these are formed, mediated and mirrored, e.g., in education, is a large, but so far mostly overlooked, part of the complex problem of low female participation in CS (Björkman 2002, 8).
Examining how computing is conceptually situated, how computing and auxiliary technologies are used, who can access them and for what reasons, and which projects pursued are important lines of inquiry. Questions like these provide clues about how science works, both at the level of discovery (how new ideas for research projects are generated) as well as at the level of justification (what evidence is needed, what tests are to be run). In this framework of feminist philosophy of science, questions of values and politics are no longer thought to be outside the process of science.
83 Chapter Three: Computing
SYNTHESIS
What are some of the conceptual foundations of computing? In AI it is the concept of intelligence that must be defined, developed, and analyzed in order to create an artificial intelligence. Critiquing AI can thus build on philosophical work in biology and the cognitive sciences, as well as in feminist studies. Determining how humans undertake a procedure and whether a computational procedure should follow or diverge from that constitutes another important area for philosophical analysis. In other words if the goal is to build rational agents, but humans consistently and regularly violate rational norms of reason as Kahneman and
Tversky (1982) famously suggest, it is not clear why we should stress rationality in AI research.
The decision to model the computational processes on real humans as opposed to creating new agents from scratch is a decision that should be informed by clear and thorough philosophical analysis and current empirical research. Philosophy of science is important to science on the large scale because the pressures on individual scientists and the demands within the disciplines do not usually allow the time or space to think philosophically, as Thomas Kuhn remarked in his
Structure of Scientific Revolutions. Philosophers think about how science as a whole functions, is (or is not) progressing, is meeting its own goals, or even what those goals should be and who has a say in setting them.
Returning to questions of epistemology, is scientific knowledge, for example, properly characterized as social rather than exclusively individual, as suggested by Pitt (2000)? To restate the question: Is knowledge held and transferred by the scientific community or individual scientists? Individuals partake in the evaluation and creation of knowledge through their participation in group(s). While many mechanisms for establishing the truth of a scientific claim are community-based to some degree or another, scientific methods do distinguish science from
84 Chapter Three: Computing many other endeavors. Clearly the empirical component of science is vital to its success. A large part of this empirical nature is built around evidence. However, the empirical world is so complex that it can support multiple interpretations of itself simultaneously, without any one of those interpretations necessarily dominating based on ability to explain and predict natural phenomena. As Harding (1996) states:
While no scientific claims may be trans-historically, uniquely congruent with nature's regularities, only certain ranges of them are consistent with such regularities and thus can generate reliable predictions (p. 263).
Peer-review for funding, presenting in public, publishing in refereed journals, and repeating experiments done in other labs, are all examples of how the community per se maintains epistemological standards, how the community guarantees consistency with the regularities.
Many feminist science scholars would further expand the notion of community by encouraging non-experts to be involved in the critical review process. Others might balk at the suggestion that a non-expert could have anything important to say about the area under review. While I do not see the role of the wider community as being beyond the spirit of real science, I do think that opening the doors to outsiders has been taken as counter to the cherished ideas of objectivity and neutrality and thus insulting to science as arbiter of truth, above and beyond the fray. I am arguing quite the opposite; that these changes I have presented will strengthen science.
These questions about social context can lead to questions about embodied knowledge.
Some feminist science critics who wonder whether there is an important distinction between propositional and embodied knowledge, have done so because empirically it is real women who hold the embodied knowledge and it is that knowledge that is denied status as “real” knowledge.
The emphasis on prepositional knowledge is softening with increased interest in A-life research and robotics. That shift has come about with very little influence from feminist criticism, which
85 Chapter Three: Computing shows that feminists do not always have a corner on the market when it comes to innovative ideas. On the other hand, it demonstrates that the concerns of feminists are not merely dogmatic.
The motivation to think about, for example, how computer programs represent people and their knowledge, comes from a concern about the impact of technology on women, which leads to questions about how knowledge is represented, and what is selected for representation. If embodiment is necessary for epistemology, must not AI then also be embodied? If so, robotics should definitely be an area of interest for feminists, as feminist science criticism is quite at home with an embodied view of self.
CONCLUSIONS
Feminist philosophy of science does need to assess computing in detail. Such a project relies on an examination of science, which I have provided at the beginning of this chapter.
Science is not something “pure” that can be separated from the society and the social norms in which it is practiced. This does not have to entail that science is mere social construction, only that it is an endeavor that is more or less embedded in the surrounding culture and society at large. Helen Longino and Sandra Harding argue that science is necessarily a social activity, and that this, contrary to some interpretations of their criticism, actually safeguards science against being hijacked by and for the interests of those who can exert significant social power; the social character insures, as much as is possible, greater objectivity.
Feminist science criticism reveals important aspects about the connections between science—in this case computer science—and society that are not always apparent within the practice of science itself. Feminist science criticism seeks to strengthen scientific projects against inappropriate and usually unintended narrowness. Feminist science criticism moves well beyond simply recruiting more women into science. Notably, feminist critiques are having a
86 Chapter Three: Computing positive impact. Gender questions have worked their way into policy and funding sources: a recent NSF grant calls for research on gender and science.35
Looking at gender in computing reveals assumptions about social expectations of men and women. This social context is relevant to science, although the extent of the relevance is debatable. Using some recent research in cognitive science I will turn to that social context and build the case for a view of cognition as both embodied—not completely separable from rational processes—and embedded—situated in a social and a physical environment. Assessing the social context in science is important since it is not the cognitive abilities of scientists that differ from other knowledge seekers.
As a feminist committed to equality and open opportunities for women and men, the relatively few numbers of women in computing raises concerns. On the one hand there may be political and social oppression involved, which clearly must be identified and eliminated. This is a political and ethical concern, but it is not isolatable from the practice or theoretical development of science. On the other hand, because science is a social practice, the composition of the scientists might (although it is not guaranteed to do so) impact the deeper level of scientific theorizing.
Observing the wide influence of computing has led me in turn my feminist lens towards computing as a whole, its products, and its creators and in particular to investigate the role of women in computing. I began this chapter by suggesting that one of the crucial tasks for feminist philosophy of science is to show how issues in epistemology connect to concerns about justice (or ethics). I turn in the next chapter to the argument that science, including the computing sciences, develops out of a social epistemology, meaning that the production of knowledge in these disciplines is located in the community, rather than in individuals. The
35 Please see http://www.nsf.gov/pubs/2001/nsf01130/nsf01130.htm#INTRO (Visited 26Aug02) 87 Chapter Three: Computing social account is better equipped to negotiate a path between the valued dimensions of science on the one hand, and the empirical constraints of the natural world, on the other. Later in chapter five, I focus on a small subset of those connections by asking about competition in science, how it affects epistemology and matters of fairness for women in science. Then, in chapter six, I explore the implications for an epistemology that takes physical embodiment and social- situatedness seriously. Doing so raises the problems of ignoring gender because it tends to result in masculinity becoming the default gender.
88
CHAPTER FOUR: SOCIAL EPISTEMOLOGY
INTRODUCTION
If epistemology is the study of how we know what we know, social epistemology is the investigation of that "we." The “social” is not simply a look at how people interact, however. It is not a sociological study, but a philosophical investigation of the foundations of epistemology, with respect to epistemic agents. Historically, epistemology has been driving under the influence of Plato and Descartes, focusing on how individuals come to have knowledge. Some contemporary philosophers, however, have begun to reconsider that fundamental assumption.
Philip Kitcher, Lynn Hankinson Nelson, K. Brad Wray, Helen Longino, Elizabeth Potter, David
Hull, Elizabeth Anderson, Frederick Schmitt, and others, provide accounts regarding how communities generate knowledge.
Science is necessarily a social endeavor. Importantly, it is also grounded in the empirical world. At minimum this means that the scientific theories or data produced by scientists— candidate claims—must be verified by the community before they are entitled to be called scientific knowledge (Longino 2002; Pitt 2000, 129n8). As far as science is concerned, it doesn’t “count” for an individual scientist to know something in isolation. Perhaps individuals can have knowledge of some things,36 but scientific knowledge can only arise out of community interaction. There are several reasons for positing the epistemic importance of communities.
First, science is an additive project, with new knowledge building on previously developed knowledge. Scientists, working over time in networks of groups and subgroups, elaborate,
36 Even if individuals could develop knowledge in complete isolation, which I doubt, the knowledge would pertain to few things or events and the knowledge could not be elaborate or in-depth. Chapter Four: Social Epistemology differentiate, and clarify the knowledge of the past. Second, science covers so many phenomena, which are themselves so complex, to yield conclusively to the efforts of any single individual.
Furthermore, scientists, like any other group of people, do not arise out of thin air.
Minimally, they rely on others to teach them, and to supply other information and tools, such as theories, formulae, data, apparatus, techniques, etc. Scientists develop their ideas with the help of and in reaction to others. From graduate assistants and colleagues to a scientist’s employer or funding agency, many people are involved in shaping the skills and ideas of scientists. Another social aspect of science is the dialogue between researchers and other researchers serving as critics. This constitutes part of the review and justification aspect of science. A community- based epistemic model automatically incorporates these influences, and others, such as financial and political pressures. As vast an endeavor as developing scientific knowledge needs the contributions of many perspectives in order to record the variety of scientific phenomena.
Different perspectives seek out different problems to solve, different justifications and alternative approaches to a single problem, all of which help to assure validity of theories, methods, and practices in science. Using the concept of equipoise, defined and discussed below, will help to further explain what it means for a community to be an epistemic agent.
Social epistemology provides major support for my feminist analysis of the computing sciences. Knowledge is particular, not universal or abstract, with social and historical dimensions. If we want to secure objectivity in science, the only way to do it is in and across communities. Scientific knowledge results from socially negotiated standards of evidence, of reasoning, of experimentation, etc., tested against the empirical world. Using gender as a tool of analysis assists in uncovering assumptions or inaccuracies embedded in those standards, as they
90 Chapter Four: Social Epistemology pertain to specific projects or entire disciplines, which result in harm or neglect to women, men, or children. For example, Londa Schiebinger (1999) suggests that tools for gender analysis should assess a range of factors, including: priorities and outcomes of research; subjects chosen for study; institutional arrangements; how cultures are defined as scientific or not; and the language used in science. Feminist critiques are valuable for highlighting problems in theories, practices, and methods.37 Beyond just raising awareness, though, feminist science criticism has been successful enough38 to encourage deeper investigations of certain disciplines and new investigations of others. Although the computing sciences have generated less critical interest from feminist philosophers than, for example, the biological sciences, there are many good reasons for examining work in computing, as presented above in chapter three.
This project in philosophy of science, focusing primarily on epistemology, is a naturalized project. By naturalized I mean a project which relies on the findings—to various degrees—of science to inform it. Science is both powerful and fruitful, so using its own methods to examine its own practices seems promising. We are able to know the world we live in, which is itself knowable because it contains natural kinds and regularities. Investigating how we know what we know via empirical methods is the hallmark of naturalized epistemology (Kornblith
1994). Empirical investigations give us more information about how those natural processes, including our own abilities, work. With that information we can discover our epistemic strengths and weaknesses and devise ways to enhance our strengths and overcome or mitigate our weaknesses. Viewing philosophy as merely logical or conceptual analysis is selling it short. A
37 The critical function is what I take, for example, Ian Hacking (1999) to refer to when he speaks about raising consciousness as a motivation for debate about the proper role of social practices in science. Even if subjective factors do not affect science as much as some critics would have us believe, those critics are still usually motivated by desires to improve the lives of the disadvantaged. 38 For example in primatology and anthropology; see discussions in above chapters. 91 Chapter Four: Social Epistemology fuller understanding of philosophy, in particular of epistemology, incorporates the empirical findings of science, even though epistemology does not simply reduce to a branch of science.
A commitment to naturalized epistemology reinforces the importance of reviewing the empirical data, for analysis along ethical and political dimensions. Scientific reviews of how medicine is practiced, for example, lead to better ways of maintaining health (e.g. Lewis 2002).
Not only is it rational, it is morally responsible to seek out information relevant to the foundations of knowledge. Philosophical theories like scientific ones must be open to revision, and scientific research can inform and challenge philosophy in productive ways. Scientific findings can help to set the boundaries of philosophical expectations in some cases: investigating “rationality” itself for example (e.g., Gigerenzer 1996) and discovering that it is very difficult for humans to be consistent in their preference orderings or that humans routinely make “irrational” mistakes about judging probabilities (e.g., Kahneman, Slovic, & Tversky
1982). Scientific findings might indicate new or unknown abilities in other cases, creating new venues to be explored, such as the way that humans incorporate “tools” including other persons in their environment to enhance their own abilities in planning (Clark 1997). As Hilary
Kornblith writes, “we also need to know whether our social institutions and practices are helping to inform us or to misinform us” (1994, 97). Science is one of those social institutions.
There are normative dimensions to epistemology, and when scientific research fails to meet the normative epistemic standards, that scientific research is judged deficient. Several normative positions infuse this project. First, neither philosophy, epistemology, nor philosophy of science can ignore the findings of empirical science. To posit a theory which results in harm to or marginalization of persons is to reject crucial information about the theory itself and how it fits with the empirical world. For example, epistemological theories have in the past tended to
92 Chapter Four: Social Epistemology marginalize the experience of many women. The men only heart disease studies also exemplify this point. This does not necessarily mean that a theory should be rejected only if it were to permit harm to a particular group. However, the existence of harm warrants a reexamination of a theory, because it indicates that the theory might be inaccurate.
Another normative charge is that science must not proceed without attending to ethical considerations. If a scientific theory brings about harm to the environment or to people, particularly those who are disadvantaged in some way, we must take great care when considering whether or not to pursue the research or the application. Other strong positive reasons for continuing with the research must be proffered in order to justify further work. Additional empirical research is required when the techniques of an experiment or parameters of a trial fail to represent or consider certain groups. For example, when the subjects of a treatment study are not representative of the afflicted group, the research must be performed again: this has happened in trials that neglected to include women in studies of heart disease (Laurence &
Weinhouse 1994) and in research measuring the impact on women of workplace computerization
(Balka & Smith 2000). Scientific theories that fail to meet ethical requirements are unacceptable theories since values are an indispensable constituent of science. Ethical, political, and social concerns are neither accidental nor irrelevant to science. This applies in both the creation and application of science. Concerns about the ethical treatment of the subjects of research or those affected by science suggest diversity within the epistemic communities as a safeguard. The emphasis on the importance of diversity in epistemic agency, arises from the belief that in diversity lies some reassurance that harms will be identified and prevented, reversed, or at least halted, as different members track their own interests.
93 Chapter Four: Social Epistemology
EPISTEMOLOGY & GENDER
Both naturalistic epistemology and feminist philosophy of science serve as sources for social epistemology, a point noted by Frederick Schmitt in his introduction to Socializing
Epistemology (1994, 3-4). However, the relations between naturalized epistemology, feminist epistemology, and social epistemology vary dramatically according to the theorist. Feminist philosophers have asked many questions about epistemology generally, but epistemic agency, in particular, is a frequent subject of investigation. For my purposes here I want to narrow the scope further to questions about the relationships of gender and epistemology, specifically, how gender can influence scientific theory and the practice of science.
Gender can affect science in at least two distinct ways. One concerns who gets assigned what work. Gender can play a role in the gathering of data and our ways of knowing generally
(Anderson 1995; Harding 1986). Beliefs about what knowledge can be created by men, and what by women, have encouraged entire disciplines to be androcentric, e.g., high-energy physics, and some to be gynocentric, e.g., primatology. Another way gender impacts science involves what kind of work is encouraged, including work in theory construction. Disciplines can be populated by men and women and still be androcentric at the level of theory when sexism and gender symbolism invade theory, as illustrated above with situation in anthropology out of which the man-the-hunter theory developed. Gender can have an impact on the content of theories when, for example, gender relations are used as unarticulated models within the theories.
Examples of such symbolism can be found in all kinds of discussions incorporating models based on ideas about competition, natural selection, and capitalism (Anderson 1995). The important point is not that the particular model is sexist (or racist or whatever), but that a model might be influential enough to subsequently inform a wide range of other projects. For example
94 Chapter Four: Social Epistemology in computing, the computational metaphor discussed in chapter one shows how powerful and pervasive metaphors and models can be (Stein 1999).
Masculine behavior when taken to be normal or a neutral starting point, serves as another illustration of how gender influences theoretical models. This model tends to overlook the value of non-masculine behavior or simply pejoratively mislabels it as abnormal. But clearly not all behavior is exclusively male or female. Much of gendered behavior for example, although historically connected, is not essentially tied to biological sex (e.g, Jaggar 1983). Women, just as easily as men, can adopt behavior associated with masculinity, particularly in long-standing, male-dominated environments. Even if there were no outright barriers to the participation by women, there still could be issues of lowered epistemic authority for women who chose to reject
“masculinist” behavior, which is taken to be gender-neutral by others.
One approach for answering questions about how gender impacts epistemology is to start with questions like “who is the epistemic agent?” Scholars like Longino (1990), Nelson
(1993), and Solomon (1994) try to provide clear answers regarding whether knowledge is created and understood by individuals, by groups, or both. This leads to questions about membership in different groups and what that means for conceptualizing epistemic agency. In particular, epistemic agents are almost always members of multiple epistemic communities. Their memberships shift often and any individual might be more or less connected to any one group at any particular time. The positions of epistemic agents are dynamic. Feminist theory is comfortable with epistemic agents who wear multiple hats (e.g., Björkman 2002; Harding 1986;
Lugones 1994, 1996), and can fairly easily accommodate switching hats (i.e., traveling between those groups) as well as recognizing the epistemological significance of the viewpoints represented by those hats. For instance, Christina Björkman explicitly analyzes professional
95 Chapter Four: Social Epistemology computing papers from various standpoints such as computer scientist, teacher, woman in computing, and researcher in gender studies. Different concerns arise depending on the standpoint, and the combination of standpoints helps her more effectively assess the material
(Björkman 2002).
Feminists must also be mindful of the practical and political goals of taking the real experiences of women seriously. Dismissing the contributions of women is almost certainly not an explicit aim of most theorists, of course, but if women’s contributions are neglected by an epistemological theory, feminist tenets require a reexamination of those theories. The experiences of many women as knowers, and sometimes as scientists, have sometimes fit poorly within traditional accounts of epistemology, which have been based on a picture of individualistic, rationalist minds working in isolation. Björkman (2002) argues that the ruling epistemology in computer sciences is based on positivism, which she defines as “the idea of science as neutral and objective” (p. 14). This epistemology, she claims, does not allow issues of identity or subjectivity to enter into discussions about computer science. On such accounts of how knowledge is produced, it could seem that some women do not have “real” knowledge and/or are not capable of contributing to the creation of “real” knowledge (Code 1993). This conclusion arises after the conceptual separation of ways of knowing and the hierarchical valuing of the separate ways: the primary contrast is usually between rational (the preferred form of) and embodied (a lesser form of) knowing.39
When some disciplines encourage and reinforce the separation of ways of knowing, the result has been an impoverished investigation of some of those ways, typically but not exclusively those associated with women. For example, a model of so-called male
39 For more general discussions about binary logic or dualistic reasoning see e. g. Cuomo (1998); French (1981); Keller (2000); Kershner (1993); Merchant (1980); Plumwood (1991); and chapter five below. 96 Chapter Four: Social Epistemology characteristics as neutral or normal appears in artificial intelligence. This has happened in the very defining of intelligence as rational thinking or behavior, a characteristic viewed as preferable and praiseworthy. The focus up until quite recently in AI was on how to replicate rationality, by either making systems that think rationally or act rationally (Russell & Norvig
1995, 5). Feminist philosophers of science, however, tend to argue for valuing several kinds of knowing, without always privileging rationality as the most distinguished form of knowledge.
From this position, feminist theorists can raise questions for AI regarding the possibilities of altering the definition of intelligence or expanding the approaches from which artificial intelligence is developed.
By starting with questions about how knowledge is gained, and turning to questions about epistemic agency, we have landed in a place which challenges the common conceptual understanding of science, as well as the actual practices themselves. “Successful theorizing,” writes Elizabeth Anderson, “deeply depends on personal knowledge, particularly embodied skills, and often depends on emotional engagement with the subjects of study” (1995, 9). This is true for all theorizers, both men and women. It is not something, though, that has been part of the standard view of scientific practice. Neither is the notion of epistemic agency as social, to which I turn next.
IS KNOWLEDGE SOCIAL?
Here I focus on just two possible interpretations of this question. “Is knowledge social?” could mean that a group, rather than an individual, or perhaps in addition to an individual, produces knowledge. In other words the group is the epistemic agent. It could also mean that the knowledge making occurs in a social context. My argument is that the epistemic agent is the
97 Chapter Four: Social Epistemology group or community, which exist in contexts. My conclusion is based on meaning holism, the nature of scientific knowledge, which occurs over time, builds on past experiments and knowledge, etc., and the nature of human cognition and socialization. There are various social contexts which dynamically encompass epistemic communities. Contrary to the claims of some critics this does not reduce to merely political consensus, but is instead secured by interactionism, a robust, dynamic dialogue within the community. The concept of equipoise illustrates how interactionism works.
KNOWLEDGE PRODUCED BY A GROUP
The discovery, practice, and justification of science are better explained when epistemic subjects are taken to be communities rather than individual scientists. Lynn Hankinson Nelson argues that communities, rather than individuals, must be the "primary generators and repositories of knowledge" (1993, 151). Her argument is anchored on a view of evidence, which she presents as necessarily connected to many communities by virtue of multiple webs of belief.
Individuals alone could not sufficiently understand the significance of evidence without reference to the webs of belief in which the evidence lies. The process of building a theory takes more time and energy than any one individual could expend and is based on too many other theories and pieces of evidence for any one individual to be solely responsible for its articulation. A theory like Nelson's, which posits communities as epistemic agents, is better equipped to handle a system of checks and balances whereby knowledge that works (explains more, incorporates more diverse data, is more easily applicable, etc.) is developed further and knowledge that does not work as well is set aside. Such a theory is ultimately better able to
98 Chapter Four: Social Epistemology explain why science is so productive and how pieces from many different "systems" or communities come together in science and are combined in various ways.
If individuals were the epistemic agents, a single Truth would have to exist; otherwise, we would not be able to make sense out of how any of us could know much in common with any others of us (Nelson 1993). We are different enough in our cognitive and sensory abilities, that it would be a miracle if we would come across enough of the same information and process it similarly enough to develop common knowledge. Alternatively, generating knowledge could be a joint process of each scientist contributing in a small way to the group’s understanding.
Scientific knowledge is the sum of those contributions, and nothing more mysterious. With regard to proving theories, we do not rely on isolated experiences to determine whether or not to accept a theory; we utilize whole webs of beliefs, rich with detail, interconnected to other theories and beliefs. This alternative notion of evidence, knowledge, and epistemic agency seems more plausible, is better able to accommodate the real experiences of people, both men and women, and it does not isolate science from other kinds of knowledge producing activities.
Part of the plausibility of this account arises from a holistic view of language. Some of the attraction of holism is the view that, for example, a single scientific theory is unintelligible out of context of broader scientific knowledge. The webs of belief fundamental to Nelson’s claim that communities are the primary repositories of knowledge are rooted in Quinian holism, in which “sentences of a theory have their meaning only together as a body” (Quine 1969, 80).
Meaning is holistic, and theories have meaning via implications in the world, which is the basis for naturalism since those implications can be empirically tested in the world. It is only in this matrix of theories and evidence that other theories and other evidence can be accepted or rejected. How we test and what we test are determined by community-negotiated standards.
99 Chapter Four: Social Epistemology
Whether or not a particular scientist is sure about a theory or a result, it is the community itself, having the necessary resources, that will eventually determine an answer, as the knowledge about that area develops.
Having a community as the epistemic agent does not mean that every member of the group always follows the same plan. For example, in the medical research arena, we would not want every researcher to be following the same line of inquiry. We benefit when different, even contradictory research projects are pursued because that group work functions to increase and develop scientific knowledge has increased. The final results get determined at the community, not individual, level. Miriam Solomon’s account of social epistemology is built on this point.
Solomon (1994) locates epistemological agency within the scientific community and not solely in individuals, to develop what she calls a “more social epistemology.” Solomon argues that this is a superior theory because it satisfactorily accommodates the social factors that play a role in epistemology, particularly in the realm of scientific knowledge production. For her it is not necessary for individual scientists to be rational, and in some sense it is better if they are not, in order for outcomes of the scientific enterprise to be rational: "A few scientists may, on occasion, reason with full scientific rationality, but it is not necessary or even desirable that they do so" (1994, 219). We want scientists to follow whatever (within certain limits) motivates them so we can have a multitude of competing theories, in part because:
underdog, yet still empirically successful, theories often maintain important connections with the world that later theories build on . . . And the consequences of suppressing less successful theories. . . can be bad for science not only because of the failure to make widely known some empirical successes, but also because the theories suggest directions for development of other theories (1994, 227, emphasis in original).
100 Chapter Four: Social Epistemology
Social epistemology is a better explanation for how science works on the whole. We want and need disagreement and a variety of projects to continue especially when they do not seem to fit the accepted view(s) because that keeps lines of inquiry open that might prove valuable in the future.
The piecemeal, cooperative work found often in computing seems to lend itself easily to the group-agent model. Programs are often designed to be modular, which allows big projects to be subdivided into chunks manageable by individuals. Breaking down the research into smaller parts lets the work progress more quickly than if the research were done in sequential order.
Group division of cognitive labor also facilitates a solution, where a single individual would likely fail, either conceptually—because it is impossible to see the entire problem and solution— or practically—because it would take too long for one person to construct a solution. When those parts are pulled together the community can judge whether the solution is successful. Not only do groups produce, they also verify scientific knowledge. To illustrate, due to possible bias, scientific clinical research must be done on a double-blind, placebo-controlled standard, where members of the broader medical community help to illuminate potential bias on the part of individual researchers or small research groups and mitigate against its effects by comparing the research of differing groups against other findings and against previously proven knowledge.
As in science, the self-reflective nature of naturalized epistemology insists on empirical investigations of the factors that alter how people come to know what they know. Understanding the complexity of the communities, how they function, how knowledge is actually gathered, etc., can be informed by empirical research. This is not to suggest that acquiring such empirical data is itself easy or that it leads effortlessly to an obvious solution, but it is an important first step.
Racial and gender bias in medicine has sometimes resulted in harm to patients, either because
101 Chapter Four: Social Epistemology there are no treatments available for that group or because members of the group are repeatedly mistreated. For example, a 1999 Schulman et al. study in the New England Journal of Medicine reports “that the race and sex of a patient,” rather than symptoms themselves, “independently influence how physicians manage chest pain.” The lack of consistent treatment across racial groups is a concern of theorists, practitioners, and feminists who are concerned with ensuring that social institutions benefit all of us. Preventing the harm requires that research take into account relevant racial and sex differences (Lewis 2002) and relevancy is something to be determined empirically.
The Schulman et al. study was not flawless, but at least now the medical community has been made aware of the potential problems of race and gender bias and can try to be vigilant for them and heed further empirical studies. Funding and regulatory agencies can put safeguarding measures into place, instead of merely relying on individual researchers to know and do the right thing. This is fundamentally a social solution. Clinicians do assess their practices and have in the past determined that changes such as double-blind studies, careful selection of subjects to include a balance of characteristics, placebo controls, and institutional review board (IRB) procedures will improve the results of their inquiries. The use of placebos, for instance, keeps patient-subjects from skewing the study due to their eagerness for a treatment—any treatment— to work. Double-blind studies keep researchers from inadvertently influencing the selection of patients and from overlooking or downplaying negative effects of the drug or treatment. These procedures are put into effect due to the recognition that sometimes different groups can discover different information about the same situation. The remarks about double-blind studies illustrated social controls on the discovery of knowledge based on other empirical work which has shown us how our own biases can derail our discoveries.
102 Chapter Four: Social Epistemology
Medicine is no different in this respect from other scientific endeavors, including those in the computing sciences. There are biases but also self-corrective measures in science. The social manner in which scientific research is conducted, reviewed, and tested generally makes the results reliable and accurate. Using empirical data helps in critiquing knowledge. These examples are also important because they outline the process of group negotiation of knowledge.
KNOWLEDGE IN A SOCIAL CONTEXT
The phrase “scientific knowledge is social” can be interpreted in a variety of ways
(Schmitt 1994). One conception is based on the understanding that the conditions of that knowledge are social. As I understand this argument, local research needs influence which theories become scientific knowledge in so far as they present opportunities for generation, study, and refinement of theories in the first place. Here the research opportunities per se constitute the social conditions of knowledge. The research opportunities are social because they are embedded in social practice and arise from social factors. By “local needs” I include the context of a researcher or research group, who may be inclined to pursue an inquiry in order to cure a loved one, to gain recognition, or to help a local concern. Following Harding (1998) I think that local needs do in fact suggest certain research questions and solutions and generally serve as “toolboxes” for scientists.40 Problem identification then is one aspect shaped by the social context. In this aspect, too, social forces constitute a mechanism that ensures the pursuit of a plurality of topics (Solomon 1994, Hull 1988, Kitcher 1993, Longino 1990). In some sense of plurality social pressures arising from politics, economics, etc., lead to the earliest formation of a scientific theory.
40 See Harding’s chapter four for more extensive discussion of “toolboxes”. 103 Chapter Four: Social Epistemology
“Social” could also reference factors in terms of group dynamics, the definition and criteria for rationality, or those arising from cross-cultural studies of learning. Miriam Solomon
(1994) distinguishes individual motivational factors from the social backdrop or context in which a group functions. Some think that competition, for example, might motivate an individual or group. Both David Hull (1988) and Philip Kitcher (1993) bank on this in their accounts of how scientific knowledge develops. Given that human individuals make predictable and common cognitive mistakes (e.g., Kahneman, Slovic, & Tversky 1982), an account of social epistemology serves to safeguard scientific knowledge because it does not posit knowledge as the product of any single individual. Rationality then becomes a function of how well the group seeks out and responds to criticism, rather than how well any one person reasons. Social factors could also relate to the complex pressures of maintaining group identity. This could be spelled out in terms of national, political, ideological, regional, or other identities. Sharon Traweek's (1988) research on high-energy physics, particularly the distinctions between American and Japanese research groups is just one of many good examples of this. These pressures can dictate which questions are investigated and which methods are used. The composition of and environment surrounding the group then flavors the knowledge produced.
Another line of interest in and justification for a social epistemology is based on the need for shared (public) standards. These public standards function most importantly to identify and employ evidence for hypotheses or theories. For Helen Longino, social, political, economic, and the like interests do interact in epistemological processes, however, there is much more to science than politics. It is the conditions of knowledge, for Longino, as well as motivations behind scientific research, which are social. She explicitly states that scientific knowledge is social because it "is produced by cognitive processes that are fundamentally social" (Longino
104 Chapter Four: Social Epistemology
1994, 139). Observations get their stability through review and correction by others. Reasoning processes must be open, i.e., open to challenge and revision. Knowing what to posit as a hypothesis to the community and knowing how to critique a hypothesis each involve using community assumptions to support or challenge the relevance of data to a hypothesis. Social actions are involved in epistemology, but it is more than just a matter of what criteria are used by the community. Rather it is what criteria ought to be used to govern scientific theory choice. An utter lack of criteria will not work but neither will mere consensus. This is a normative account because it is not the case that knowledge is simply what the group believes or accepts since:
The results of both reasoning and observation are socially processed before incorporation into the body of ideas ratified for circulation and use. What deserves the honorific 'knowledge,' on this view, is an outcome of the critical dialogue about observation, reasoning, and material practices among individuals and groups holding different points of view. It is constructed not by individuals, but by an interactive dialogic community . . . [which] must be characterized by equality of intellectual authority. What consensus exists must be the result not of the exercise of political or economic power . . . but a result of critical dialogue in which all relevant perspectives are represented (Longino 1994, 144-5, emphasis added).
Although Frederick Schmitt (1994) allows that social forces could influence choice of scientific theory, he concludes that a coherency problem exists with the general claim of scientific knowledge as caused by social, economic, and political interests. I think that here the question is whether "caused by" must mean something like "triggered by" or something more like "ontologically derived from." For Schmitt the claim is either a priori or empirical. If it is a priori, it is due to underdetermination of theories such that non-rational factors, such as social considerations, are responsible for theory choice. Rational choice theory plays a crucial role here: "the rational theory choice is the choice that is accepted from each of various perspectives representing opposing interests" (Schmitt 1994, 26). On the other hand, the claim could be empirical: interests cause theory choice. If interests lead to “irrational” choices of theory,
105 Chapter Four: Social Epistemology perhaps the best we could do is to have competing agents who can "cancel" each other out.
Again, if it really were the case—empirically demonstrated—that our methods lead to irrational choices, we might want to try to mitigate that result with other rational constraints. Let me try to explain the problem.
Consensus, according to Schmitt, is necessary for choosing a theory in the absence of rational means. Furthermore, for him, consensus rests on the epistemological claim that reducing the effects of interests on rational theory choice is desirable. However, "a multiperspectival view can be motivated by the claim that interests cause theory choices only if it is taken as derivative from the textbook theory and is proposed, not as a theory of rational choice, but as a procedure for ameliorating the interference of interests with textbook rationality" (Schmitt 1994, 26). He identifies Helen Longino's social epistemology as belonging to a branch of social epistemology, which he calls the social constructivist model. He writes:
If we embrace the conclusion that actual choices are generally nonrational, our degree of pessimism will turn on the details of the sociological findings about interests. If interests do not inevitably cause theory choices, we may ask how to eliminate their effects. If, on the other hand, they do inevitably cause theory choices, then we may ask whether there is any hope of reducing their effects. For example, one might hope that opposing interests would reduce each other’s detrimental effects on textbook rationality. This hope has led some to a multiperspectival or consensus theory of rational choice: the rational theory choice is the choice that is accepted from each of the various perspectives representing opposing interests. Helen Longino (1990) offers a sophisticated version of this view, as she does in her contribution to this volume [Longino 1994] (Schmitt 1994, 26, emphasis in the original).
While I agree that proponents of the Strong Program do make this type of argument,
Schmitt mischaracterizes Longino's theory by categorizing it as a version of a social constructivism. I believe this is a misunderstanding of Longino’s position. Consensus can occur in a context of rationally, as noted above in the quote from Longino. She is keenly aware of the
106 Chapter Four: Social Epistemology importance of sociological factors, and is interested in dampening some of their effects (although not all because there are many social factors which are positive). Still, there are other highly significant factors also at work in her epistemological theory. She clearly does not posit that scientific knowledge is caused by social, economic, or political interests, in the simple useage of the word, "caused." However, because of the flaws with textbook rational choice theory41, it could not be the sole force in her theory. It is the dynamic, rich process of review and discussion of standards of evidence and hypotheses acceptance within the scientific community that is crucial to creating scientific knowledge.
Thus, scientific communities surely are formed by social factors, but it is not those social factors alone from which scientific knowledge arises. While I think that social context greatly influences research, it does not determine the outcomes of science because of the requirement to account for empirical data. Still the social context is quite broad and includes many dimensions, as just detailed, all of which can impact the outcome of science.
EQUIPOISE: SOCIAL EPISTEMOLOGY IN SCIENCE
I employ the concept of equipoise in the medical research setting to illustrate how a community per se can be considered the epistemic agent. The key concept in Benjamin
Freedman's 1987 "Equipoise and the Ethics of Clinical Research," is that of "equipoise."
Equipoise is the central element in a model of how knowledge develops in science. The primary epistemic agent in this model is the medical research community. Individual or even teams of medical researchers hypothesize, develop, and test treatments, e.g., drugs, procedures, regimens,
41 For extensive discussion about systematic errors in human reasoning see Kahneman, Slovic, & Tversky (1982) as well as Gigerenzer 1996, 1999, 2000. 107 Chapter Four: Social Epistemology etc., but everyone remains in a state of equipoise until there is community agreement about the efficacy of the treatment. This means that no one, including the original researchers, can claim to “know” that the treatment works until there is consensus within the broader medical community. Consensus is most often reached by means of publication, or sometimes presentation, and verification of results. Just as Longino stresses, consensus is not reached by mere political or social maneuvering, but is anchored to scientific evidence and reasoning.
The potential conflict between a physician and her duty to a patient versus her duty as a researcher to science serves as backdrop for explaining equipoise. For physicians, determining how much information, at what level of detail and pace, to provide for a patient needing treatment can be very difficult. The limits of the physician's duty to her patient are complex and sometimes hard to discern even in normal medical settings. Doctors must act under the principle of beneficence (doing good for the patient) and must respect the patient’s autonomy (her right to govern her own life and make her own choices), principles which together create a significant obligation in medical ethics called informed consent.
In research situations the doctor-patient relationship is often further strained. Consider situations that call for enrolling a patient in a clinical trial. Putting a patient in a clinical trial usually requires that the patient is very sick and that the normal course of treatment has failed; experimental treatment is the only remaining option. In these cases the doctor might be both researcher and physician. As such, the physician has different sets of obligations to adhere to, one set as a physician and one set as a research. Notably, these sets can appear to conflict.
Consider the following scenario: if the doctor believes that treatment A really is much better than treatment B (the two treatments under scrutiny in Trial X), even though the double-blind,
108 Chapter Four: Social Epistemology placebo-controlled, multi-center testing phase (in other words, a very thorough, sophisticated research protocol) has not been completed, what should she do? Her obligation as a physician to give her patient the best treatment possible is seemingly at odds with her obligation to science to participate in a process that will lead to unbiased, or at least thoroughly tested, knowledge. The obligation of physician qua scientist is further complicated since that knowledge will be used to treat other patients in the future. So even if she thought her duty to this patient is to give the patient the "effective" treatment before the trial ended, the physician might be risking the health and safety of this particular patient if she chooses to employ the untested treatment. On the other hand, she might save the patient’s life or prevent suffering, both of which are clearly morally desirable actions. However, if the treatment is not fully tested she might thereby fail her own future patients in similar circumstances. Importantly, the stakes are very high in situations where the number of patients in the trial is so small that removing even one patient would render the trial ineffective. The physician-researcher thus fails the next patient because she does not have all the information she could have gotten, information which could tell her if the treatment will work over the long term, for a group of similar patients. Note too that this is not just a personal conflict for an individual doctor. Social institutions, such as the Food and Drug Administration
(FDA), also face such conflicts when they determine the standards for “knowing” what works.
Another example of this type of conflict occurs when a therapy seems to be efficacious in treating a serious condition, but for scientific purposes the trial calls for tests using a placebo.
Treating HIV infection in a sub-Saharan African country where the rates of infection are the highest in the world provides a suitably compelling case.42 If the physician gives the treatment
42 For example, the region contains "71 percent of the global total of people living with HIV/AIDS, 79 percent of cumulative AIDS deaths, and 92 percent of the cumulative total of the world's AIDS orphans (UNAIDS defines AIDS orphans as children who have lost their mother or both parents to AIDS before the age of 15)". Specific statistics on rates of infection in the Sub-Saharan region of Africa are taken from University of California at San 109 Chapter Four: Social Epistemology to a patient in the placebo arm of the trial before the trials are completed she could help a desperate patient. On the other hand, more extensive knowledge about the treatment might be compromised by effectively ending the trial.43 This patient, the doctor believes, could get better now if she had the treatment. Yet effectiveness for the target population can only be assessed by completed studies. What should the doctor do?
The physician really cannot act except on scientific knowledge. Her intuitions about this patient might be wrong, plus she is compromising results which could help others. If she wants to help this patient she should find some way to do so without risking the whole study. So is it the case that the physician must choose between this real person in front of her and unfamiliar others? I suggest that this is a false dilemma. Reframing the problem in terms of moving the location of epistemic agency with regard to scientific knowledge from the individual doctors to the medical community at large dissolves the dilemma. It also involves the concept of equipoise.
Equipoise is a state of genuine uncertainty on the behalf of any particular doctor- researcher who weighing the costs and benefits of some treatment as compared to those of some other treatment. Clinical equipoise is uncertainty within the greater medical community, regardless of the beliefs or knowledge of any particular doctor-researcher, regarding efficacious treatment(s). The key here is the difference between what one individual physician "knows, or has good reason to believe" and what the medical community can verify. Neither the individual physician nor individual researcher (or research groups) can know prior to knowing by the medical research community. This is because scientific knowledge must be verified by the
Francisco's web page on HIV/AIDS called HIV InSite, 30 October 2000. For more information, please see: http://hivinsite.ucsf.edu/international/africa/ 43 Protocols are created by various social institutions in order to ensure quality. In this example it is really more of following-through on the established protocol, rather than believing that more knowledge will be produced. Adhering to scientific methods is a matter of principle because that principle is known to be effective and because laws and industry regulations require it. 110 Chapter Four: Social Epistemology group. One way to navigate this epistemic highwire between duty to patient and duty to science is for individual physician-researchers to suspend judgment until the community per se has gathered enough data to make an informed judgment on the efficacy of the treatment under investigation. Like Kant's fear that a few lies might destroy the practice of truth-telling, the worry here is that if every physician-researcher refused to place in or withdrew her patients from clinical trials a "practical threat to clinical research" would exist. Without the suspension of claims to know what is best for the particular patient in a clinical trial "there is an irresolvable conflict.” The apparent conflict is “between the requirement that a patient be offered the best treatment known (the principle underlying the requirement for equipoise) and the conduct of clinical trials" (Freedman 1987, 261).
In the state of clinical equipoise, i.e., before the data from the trial(s) has been analyzed, there is no "consensus" within the medical community regarding which treatment is preferable.
Instead there is "conflict" or a "split" in the community with each side having its own evidence.
When enough evidence is gathered to convince most of the community, clinical equipoise has dissolved. Otherwise, "clinical equipoise persists as long as those results are too weak to influence the judgment of the community of the clinicians . . . " (Freedman 1987, 263). At the point when clinical equipoise is dissolved, the need to acquiesce to the community's judgment becomes a serious, ethical issue; no longer can the physician practice against the findings of the community (Note that this is a somewhat different situation than is usually the case for scientists, whose work does not necessarily impact the health and lives of other people). The knowledge resulting from these clinical trials is imbued with normative powers, based on its status in a social institution, i.e., the scientific medical research community:
The ethics of medical practice grants no ethical or normative meaning to a treatment preference, however powerful, that is based on a hunch or on anything less than
111 Chapter Four: Social Epistemology
evidence publicly presented and convincing to the clinical community. Persons are licensed as physicians after they demonstrate the acquisition of this professionally validated knowledge, not after they reveal a superior capacity for guessing. Normative judgments of their behavior—e.g., malpractice actions—rely on a comparison with what is done by the community of medical practitioners. Failure to follow a 'treatment preference' not shared by this community and not based on information that would convince it could not be the basis for an allegation of legal or ethical malpractice" (Freedman 1987, 263, emphasis added).
I quote extensively to emphasize the epistemological implications: doctors are held accountable for knowing what the community knows and, furthermore, acting on one's own beliefs not only does not count as acting on knowledge, but it can actually result in grave professional sanctions.
Knowledge does not get recognized as such until it is "professionally validated." The same is true across all scientific disciplines.
Based on this discussion of equipoise, the social epistemology model in science serves explains partial verification or split judgments. The group can be split in judgment about what works or which avenues to pursue without jeopardizing knowledge itself, something which is harder to explain if individuals are the producers and repositories of knowledge. While it might be incoherent for an individual to know “X” and “not X,” it is quite reasonable for a scientific community to hold both claims simultaneously, without any overall damage to the processes or outcomes of the cognitive community. For instance, even when one scientist regards some drug as possessing curative properties, it must be tested and reviewed for subsequent acceptance into the community's knowledge base. Over time and across different laboratories claims are tested and verified or rejected. Until then, those beliefs cannot rightly be called scientific knowledge.
Individuals can be certain of their data, but others cannot rely on it until the group sanctions it— usually this is done in a peer-reviewed publication. In cases where consensus is lacking, scientific knowledge status is not warranted.
112 Chapter Four: Social Epistemology
SOCIAL EPISTEMOLOGY IN AI
Regarding the very idea of artificial intelligence, several very vocal proponents of AI, e.g. Allen Newell and Herbert Simon, said early on that computers would easily acquire human level intelligence in short order (Russell & Norvig 1995, 17, 20). AI was “born” in the late
1940s, early 1950s, and still has not realized those original bold predictions. The AI community has been continually conflicted about which methods for AI research are truly viable and which cannot be resurrected. Neural networks, for example, originated in the late 1960s, then were neglected due to a perceived or real lack of promise. They were “rediscovered” in the mid-1980s
(Russell & Norvig 1995, 21 & 24). Some critics say that computers will never, and in principle could never, have real intelligence like humans (Dreyfus 1972, 1979, 1992).
In the meantime the AI community has come to redefine “intelligence” with respect to goals for creating it artificially, and has reevaluated the methods for achieving those goals. They have been slow to show success at pattern recognition or natural language processing, but researchers in those and other areas of AI continue to make improvements. For example, areas such as robotics were not originally thought important to artificial intelligence, but some current researchers are challenging that idea. These researchers are redefining the concept of intelligence away from a rational, proposition-based model to one arising from an embodied position in an environment. This approach is radically different from programming methods that guide a “rational” search through an abstract space of possibilities. Although not everyone assents to the importance of robotics with regard to AI, it might very well be more important than first believed.
I will discuss a specific AI project called Cyc in greater detail in chapter seven. For now,
I want only to highlight the fact that it has only been after the project had time to mature that the
113 Chapter Four: Social Epistemology broader AI community could make good judgments about the soundness of this particular approach. To draw an analogy to the medical setting discussed above with equipoise, when a community is not certain about knowledge it is because there is not enough evidence for or against a hypothesis for the community—not the individual researcher—to judge it knowledge.
It is not that individual researchers should give up all their personal, political, valued commitments, but rather that as a community we balance out those commitments, so that no one set dominates. I might really believe that artificial intelligence must incorporate embodied knowledge, but other researchers might think that foolish. With each group pushing its own projects, more evidence becomes available for the entire community to assess. When empirical evidence is required no particular set of value commitments dominates, unless and until the community together and over time establishes that those standards serve the entire community fairly. There will always be meta-discussion about what constitutes a good and effective community and such discussion is itself a social endeavor.
EQUIPOISE & DEFINING SOCIAL EPISTEMOLOGY
Using the concept of equipoise offers a way to balance out or mitigate the worst effects of essentialized knowledge. People have limited knowledge. That fact can be accepted without penalty to our understanding of epistemology. These limits are a threat to science but only if individual scientists are taken to be the source of (scientific) knowledge. When the community as a whole is the repository and source of the knowledge, ultimately it does not matter if different individuals can study different things. Individual scientists must pass the information on to the community at large for testing, verification, and dissemination. It is this move of taking the community to be the repository of the knowledge that is crucial to positing the concept of
114 Chapter Four: Social Epistemology equipoise as a way of understanding who has what knowledge. Worries about individual idiosyncrasies (or prejudices or worse) and their potential distortions on science take a back seat when all the biases get lumped together. It is less likely, although not impossible, that any particular bias will hijack science when scientific knowledge comes from the community rather than individuals.
Still, equipoise should not be understood as something that justifies inequalities on the pretense that the knowledge will arise sooner or later. Systemic barriers are antithetical to having theories out in the community to discuss and test. Equipoise simply requires that judgment about preferred theories be withheld until the empirical evidence has been gathered and analyzed. If at any point along that path of data collection, analysis, review, etc., the findings are not open to challenge, the findings are not as strong as if they were able to withstand a wide variety of challenges. Avoiding challenges removes a chance to strengthen one’s theory as it is being developed, rather than trying to accommodate conflicting data after a lot of time and energy has been spent fine-tuning it.
A social epistemology for science recognizes that individuals do produce data and information, but producing scientific knowledge requires more. Individuals can vary in their cognitive abilities. They have different political, social, ethical values and commitments. Not all members share those commitments, which is healthy and beneficial for epistemology on the whole. Individuals are frequently members of several different social communities and move dynamically between those communities. Thus there is a social context, including specific histories, to knowledge production. A social epistemology also recognizes that there is more than one reliable view of the world. The standards for determining what theories or evidence should count as knowledge, are socially negotiated. This means that standards for good and
115 Chapter Four: Social Epistemology effective science are discussed and established at the meta-level, in the abstract, beyond any particular project. It is the group that verifies evidence and reasoning, certifying it as knowledge.
A social epistemology is perfectly compatible with the defeasibility of knowledge, too.
Scientific knowledge can always be revised, but until such revisions, we can put stock in what is called knowledge as having practical application in the “real” world.
VIRTUES OF AND CONCERNS WITH SOCIAL EPISTEMOLOGY
There are several virtues of the social epistemological theory I have been sewing together across the backing of equipoise. It provides a plausible normative account that includes social factors, while allowing and encouraging the rigor we demand from science in the form of strong objectivity. Disagreement or suspending judgment is okay; there is not just one answer and science does not have to be a zero sum process. This social account creates room for dissention as an important part of the process, without one party fearing penalty. The presence of many, conflicting views makes the process more robust. There is an analogy here to John Stewart
Mill's argument against censorship on behalf of freedom of speech: the more ideas out there on the table for discussion, the more likely they are to get moved around and connected to something interesting or useful. Emphasizing the importance of individuals encourages spending energy on promoting or discrediting individuals, whereas conceiving of the epistemic agent as a community puts the focus on the knowledge itself rather than the people. One minor example of this is the current movement to have scientific papers published in electronic journals accessible by all for little cost (e.g. Grimwade 2002). Starting from the requirements of equipoise and encompassing all of these important aspects of social epistemology, we now have a model of how community knowledge could develop.
116 Chapter Four: Social Epistemology
Along the lines of who counts as part of the community and whether there must be some kind of credentialing process, can just anyone make suggestions that will be taken seriously?
Although some commentators want to preserve the integrity of science by restricting entry into science (i.e., the positivists and de facto by the educational process), Philip Kitcher (1993) and
Miriam Solomon (1994) offer a different understanding of the appropriate role of researchers.
Kitcher and Solomon both want researchers to wholeheartedly and unabashedly investigate all kinds of questions and follow them through to the end. This is for the good of everyone, since
(1) researchers must be committed to their projects or otherwise they would never get off the ground and (2) we never know for sure where a "winning" idea (or even part of an idea) will come from or where it will lead. Thus, it seems for Kitcher and Solomon, even ridiculous questions should be pursued. Clearly there are practical limits to this, since resources are always limited and it is impossible that all questions can be pursued. For Kitcher, the epistemic agent remains located in the individual but if it can be shown that her knowledge is mistaken, no harm will befall the community. For Solomon, irrationality is not threatening because knowledge is located in the community. She goes so far as to say that irrational researchers ultimately don't make any difference as far as knowledge production: "social groups can work to attain and even recognize epistemic goals without individual rationality or individual cognizance of the overall epistemic situation" (219, emphasis in original).
Regarding the suggestion that any ridiculous question should be researched, there must be restrictions. Certain beliefs can lead to particular practices that the community cannot support. In other words, even in the name of objectivity, certain research projects are ruled out from the start. There are basic research projects, which eventually would be tested on animals and humans, but which simply should not be pursued, because they violate other ethical
117 Chapter Four: Social Epistemology commitments. In fact, we try to restrict research with rules that are devised abstractly, by the wider community, before any particular research proposals are reviewed, accepted, or refused, in order to protect against such abuses.
CONCLUSION: REAL WORLD PHILOSOPHY
Feminist naturalized epistemology situated in the scientific and wider epistemic communities, invokes another challenge, this one specifically for philosophy and its role in the public discussion about science. Philosophy has a responsibility to be involved with public discussions about science because it can uncover androcentrism and other inconsistencies in theory. Philosophers should lead discussions about science, ethics, and politics because we are able to see across disciplines and separate out the important questions, the problematic assumptions, and well reasoned answers. Philosophers are not the only ones capable of doing this, but philosophers are well-suited to it due to their training and their positions in society.
Being located in the academy provides the time and space for such broad analyses, time that is not usually available to scientists who must concentrate on their specific research projects.
Philosophy is rightly charged with needing to be more connected and involved in the world and feminist science scholarship can assume a leadership position in doing so. Although philosophy per se and philosophy of science are both capable of assuming this role, feminist philosophy of science historically has consistently and openly articulated a position that explicitly unites political and social goals as vital to discussions about and research within science. Scientists in the 1940s tried to talk to politicians about safeguarding the world against the range of potential problems arising from the creation and use of the atomic bomb (Rhodes 1986). They were only somewhat successful, and they obviously were not able to keep the bomb from being used. One
118 Chapter Four: Social Epistemology lesson between then and now is that politics, values, and science are intertwined; pretending that they are not results in missed opportunities to make those connections public, to assess them openly, and to take political action if necessary. Just as it is the case that a non-feminist could provide a similar analysis of an issue within philosophy of science, philosophy is perfectly capable to providing a critical review of issues in the mix of the social, political, and scientific.
But such a critical review process of real-world issues has not materialized. Philosophy at large has been reluctant to take up that challenge in the same way that feminist science scholarship has. Feminist science scholarship has the desire, the momentum, and the ability to serve in this capacity. An outsider by some measure, take gender for example, who is also an insider by, say, training, can bring up useful critiques precisely because of her social position. She can explain to other scientists why she does not share some crucial assumption upon which the proposed research is based.
The scientists who are doing science on the “front lines” so to speak, do not always make the time or have the motivation to think about the bigger picture, such as, whether or not research on the human genome should be limited in certain respects. This is where philosophers can make important, substantial contributions to the process and it is an understanding of social epistemology that helps ground and encourages this role for philosophy. It is the community at large that serves as a check and balance system on science; philosophy, especially feminist science criticism is well equipped to guide that system. There is plenty of work to do from this vantage point in the process of educating scientists (not to mention physicians), in new fields in genetics and biotechnology, and in AI and computing science research.
Focusing on the community as epistemic agent raises questions about the composition and practices of the community, and whether those social dimensions and activities affect the
119 Chapter Four: Social Epistemology knowledge that is created by the community. In the next chapter I challenge the conception of competition in science, by considering the effects of institutionalized competition among scientists. Particularly with respect to the implications for women as scientists, I argue that alternative arrangements, i.e., less hostile competition, might serve science better. If women are not expected to be competitive in their daily personal and work lives, and are rewarded for being otherwise, it is unlikely that many of them will in fact be highly competitive. And if being competitive is some sort of precondition for success in science, then women will generally be less successful as scientists or fewer women will be successful. Gerald Holton, a Harvard professor of physics and history of science, reports research connecting “careerist” behavior, which is “more aggressive, combative and self-promoting,” with success as a scientist, as measured by productivity. Productivity, in turn, is related to another measure, namely, securing tenure (Holton 1998). Not surprisingly men tend to do better, to the tune of being twice as likely to gain tenure at the top universities (Holton 1998, 1). The idea of competition itself is extremely complex, though. Examining the complexity leads to a deeper and richer understanding of how science communities work and how to make science more “open” to women.
120
CHAPTER FIVE: COMPETITION IN SCIENCE
INTRODUCTION
Chemist Michael Polyani, writing about issues in the philosophy of science in the late
1940s, stated “While science imposes an immense range of authoritative pronouncements, it not merely tolerates dissent in some particulars but grants its highest encouragement to creative dissent” (1967, 68). This could be read as an endorsement of competitiveness in science.
Alternatively it could be understood as support for collaboration, an interpretation also compatible with Polyani’s views about the interconnectedness of scientists to their wider communities. For him it is clear that all scientists have a shared stake in the outcomes of science because scientists are individuals embedded within a society, not isolated from it. Building on a argument put forth in chapter four regarding the inherently social nature of epistemology, I will argue in this chapter that a model of scientific practice too strongly rooted in competition among scientists is disingenuous as well as overly restrictive of both science and of scientists.
As I showed in my first and second chapters, part of the debate of philosophers of science involves whether—or to what extent—factors like gender, race, culture and worldview, and class influences the content of science (Cajete 2000; Harding 1998; Keller 1985, 2000; Kitcher 1993;
Longino 1990). Presupposing knowers as isolated individuals helps to keep the lines between so-called “personal factors” and characteristics of good science neat and clean. However, as I argued in chapter four, science cannot be so narrowly defined, and individuals do not really undertake their cognitive work as isolated individuals.
Competition is one factor, among a set of factors, which can inhibit the production of novel ideas and can function to disproportionately discourage certain potential members, such as Chapter Five: Competition in Science women: “Science (outside of medicine) is usually perceived as a task that isolates the individual and involves competitive achievement on abstract problems” (Huff 2002, 114). Both consequences, discouraging new ideas and new members, can compromise the practice of science, and are therefore potentially detrimental to society. Science must generate numerous novel ideas to solve problems, since no straightforward method for deriving solutions to scientific questions exists. Based on the availability of more new hypotheses for testing, innovation increases the possibility of finding good solutions to problems in the world and increasing scientific knowledge. This chapter explores feminist critiques of competition in the context of arguments in philosophy of science that posit competition as necessary for science. In doing so, I explore several conceptions of competition. My claim is that something like
Polayni’s “mediated consensus” is what organizes science, rather than competition (Polayni
1967, 73). Science, on this view, involves the “principle of mutual control” where “each scientist is both subject to criticism by all others and encouraged by their appreciation of him”
(1967, 72). I argue that what is needed is to build on Polayni’s understanding by reconceptualizing competition in ways outlined by Helen Longino. More pragmatically, investigating questions about competition in science could help to address concerns about women’s decreasing participation in the computing sciences.
GENDER & COMPETITION
What is competition and does it incorporate a gendered dimension? It is not at all clear how gender and competition are connected. It appears to be partly related to self-confidence, and somewhat related to gendered differences in communication. Women tend to lose self- confidence after they enter college, and it continues to decrease throughout their stay in the
122 Chapter Five: Competition in Science academy (Pearl et al 2002). This diminished lack of self-esteem translates into gender differences in accepting criticism, speaking up, and confidence about ability to perform well
(Pearl et al 2002, 137-8). If women tend to be less confident then hostility might be more disruptive to them.
Gender and competition might be linked also due to apparent gender differences in communication. Communication which is meant and/or perceived to be hostile and competitive might contribute to a “chilly climate” for women, when women who are already uncomfortable with the situation are interrupted or silenced, even unintentionally. Teague reports: “The arrogance and competitiveness of men in the industry and their unwillingness to accept women on their merits, were stated dislikes of several women” (Teague 2002, 154-5). Pearl et al (2002) note that “such experiences are likely to exacerbate an existing lack of self-confidence” (p. 138).
Computer professionals Anita Borg and Telle Whitney provide this explanation regarding gendered differences in communication as one reason for establishing a computing conference specifically targeted on women in computing:
[Deborah] Tannen describes the male purpose for communication as establishing the speaker’s place in a hierarchy. At technical conferences, this is exhibited as competitiveness, posturing, and an abundance of ego. This is not necessarily a natural or comfortable mode of interaction for women even if some of us are experts in using it as a survival skill . . . In large part, the CHC [sic; Grace Hopper Celebration] grew from our feeling that a technical conference in which the vast majority of participants were female would provide a less confrontational and more cooperative communication environment” (Borg & Whitney 2002, 182).
Fortunately, there appears to be a great deal of specific action which can be taken to provide support for women, including conferences like the Grace Hopper Celebration, as well as recruiting, mentoring, and other programs (Cuny & Aspray 2002; Fisher & Margolis 2002; Gürer
& Camp 2002).
123 Chapter Five: Competition in Science
INCREASING DIVERSITY
These gendered aspects of competition are neither solely the result nor the responsibility of men or of women. Yet if we are committed to diversity for political ideals such as democracy and ethical ideals such as fairness, we need to examine competitiveness in computing in order to reduce the harmful effects on women (Johnson & Miller 2002). Competition could be serving as a structural barrier, even though it is not necessary to science. That is the lamentable part. I do not think that it is necessary to science, so models like that of David Hull, discussed below, overemphasizes competition.
Competition is supposed to maintain quality in science, in terms of which students are admitted into educational programs, which papers are accepted for publication, which candidate gets the job, etc. Clearly, there must be “admissions” standards and clearly a superficial diversity is no a guarantee of quality. It is very true that just adding women and stirring is not at all sufficient for improving the lives of women, nor does it improve science. Nonetheless, inquiring about what could be done to diversify a situation, here I will focus on the computing disciplines, triggers a close examination of the practices of the discipline. Thus even if some level of competition must remain—I will argue that some form of competition is helpful—other non-competitive practices could be instituted, to mitigate the negative effects of competition. In the college setting, which is a major area of concern for computing, this could take several forms including: a mentoring program between new and experienced students or new students and professionals (Townsend 2002); a peer-to-peer program or learning community between new students to teach them how to work together (e.g. Systers or ACM-W); providing “bridge” programs for teaching important academic or research skills (Roberts, Kassianidou, Irani 2002).
Many researchers have suggested ways to improve the recruitment, teaching, retention, and
124 Chapter Five: Competition in Science placement of women in computing.44 What is less clear is whether these programs will work over the long-term, especially when the social and cultural disincentives for women working in computing are relatively high, but that is a subject for a separate discussion.
Another desirable aspect of diversity is that diversity in participants might bring diversity of values, presuming that the participants have not all trained in the same program, under the same supervisors, or were raised in very similar environments, such as the same small town.
Diversity of moral, political, social, and other values, strengthens science, as argued in the opening chapters. This is part of what strong objectivity in a social epistemology contributes.
Participants with vested interests propose candidate epistemological claims. The variety of valued commitments leads to variety in candidate claims and with proper interactive dialogue, the community assesses the claims.
Feminist projects are committed to analyzing real and apparent inequities in social systems; this project asks why there are relatively few women in the computing sciences, especially in the higher echelons, what might be the cause, and what effect that has on work in those disciplines. One effect can be that designers in computing incorporate their own stereotypes of users into their software: “Certainly, it is true that software design is a social process and that gender stereotypes can infiltrate the design” (Huff 2002, 112).45 A social environment with a gender imbalance can impact work in different ways. A social environment can make it easy to avoid thinking about certain questions; it makes certain valued commitments invisible. For example, if there are no women on a work team it might be fairly easy to sidestep difficult issues facing working parents, when the men on the team are not the ones primarily responsible for child care, in a society that is structured around women having primary
44 See also www.acm.org/women ; Cohoon 2002; Cuny & Aspray 2002; Gabbert & Meeker 2002; Lazowska 2002; McIlwee & Robinson 1992; Rosser 1993, 1995. 45 See also Adam 1993, 1995, 1998; and chapter seven below. 125 Chapter Five: Competition in Science responsibility for children (Hochschild 1997). In the absence of women in the computing subculture, the permeation of child care issues into the actual lives of the computer scientists is less likely to occur and therefore is less likely to surface within the programs and research projects themselves. Programmers in this environment are less likely to design programs having a working parent in mind as a primary user. Although programs for use by a childless person might not vary widely from those used by persons with children, I mean simply to point out that in an environment where there is no reason for certain issues to ever arise, particular assumptions about users and use might never be challenged. Having parents in a workgroup, though, might make it more likely or easier to bring up concerns of parenting, including issues of software design relevant to parents (adult-material blocks in internet browsers, for instance). In a different case having persons of color in the workplace or classroom increases the likelihood of considering how some minority views are or are not addressed by assumptions made in the design and manufacture of a piece of hardware or software.
QUESTIONS ABOUT ESSENTIALISM
This is not to say that no women are competitive, that all men are, or to make any other sweeping generalization. What I am rejecting when I object to essentialism is the classifications of men and women which divide into binary opposites and are based on biological sex only, abstracting away from other social dimensions such as race, culture, class, etc. and making claims based only on gender, separated from other relationships (Alcoff & Potter 1993, 3-4). A characteristic such as gender, race, or class obviously does not exclusively define any one person. Not all women are the same, nor are all persons of color, nor all members of a privileged economic class. To claim that all members of a class act or believe in one way is to make a
126 Chapter Five: Competition in Science claim about essentialism, that all the members are essentially the same in some respect. Most feminist scholars reject essentialism, as do I.46 Nonetheless, categories like gender are often good starting points for assessing a state of affairs, patterns of practice, and hidden assumptions.
We can look at women as a group because they share a great deal in terms of socialization. Women tend to be less competitive than men because they are socialized to be as such. There might be biological differences between the sexes which result in gendered differences, but I doubt it. Furthermore, even if there were some biological distinctions, the socialization still plays an extremely important role in the traits and abilities of men and women
(Holmstrom 1999; Jaggar 1983). We can use gender as a lens without succumbing to essentializing.
DIVERSITY & NOVELTY
Another motivation for increasing diversity in science, is that it improves when the opportunity for novel and varied criticism produces something useful (Negroponte 2003).
Critics from one area can supply questions about another that can lead to fruitful reconsiderations for one or both parties, as historian Londa Schiebinger relates using an illustration from archaeology:
collaborations in archaeology have sparked its practitioners, as [philosopher of science, Alison] Wylie reports, 'to think differently about their discipline and their subject matter, to identify gaps in analysis, to question taken-for-granted assumptions about women and gender, and to envision a range of alternatives for inquiry and interpretation'—surely the very stuff of scientific creativity (1999, 144).
46 Cross-cultural studies are particularly valuable here for illuminating what, if any, gender-associated traits, abilities, or practices hold up across cultures (e.g. Galpin 2002). 127 Chapter Five: Competition in Science
Diversity among scientists generates novel questions (Negroponte 2003; Wallace 2001, 2001b).
Variety in age, nationality, institutional affiliation, gender, etc., in a community fosters diversity:
“A very heterogeneous culture . . . breeds innovation by virtue of its people, who look at everything from different viewpoints” (Negroponte 2003, 34). Diversity also translates into variety in approach: “the interaction of individuals with different styles and temperaments, as well as different professional expertise, is what provides synergy. That is, it’s what makes the overall ability of the team much greater than just the sum of the abilities of the individual members” (Wallace 2001, 2). Recruiting, training, and promoting a diverse work force is one way to foster diversity. Diversity can be promoted through mentoring, as well as fostering a range of methodologies, which employ different background assumptions.
Interdisciplinary laboratories are good examples of the benefits of diversity. The
Whitehead Institute houses scientists trained in mathematics, molecular genetics and cell biology, and molecular biophysics and biochemistry to research problems in comparative genomics. The collaborative, interdisciplinary research environment is credited with much of the
Institute’s success: it ranked second in the world in genetics and molecular biology research in
2001, according to Science Watch (Calandra 2001). Diversity does not guarantee results, but it makes them more likely over the long term.
Philosopher David Hull, whose model of competition in science I will turn to below, does not believe that it is possible that science could be unduly affected by mere social factors and he rejects “some form of social relativism with regard to empirical truth” (Hull 1988, 4). He disagrees, for instance, with the claim that Western science has a masculine bias, as theorists like
128 Chapter Five: Competition in Science
Sandra Harding (1990), Carolyn Merchant (1980), and many others have argued. For example, he writes:
Extrapolated to science itself, this view of the pervasive influence of society on our conceptual systems has given rise to claims that science as it has been practiced since its beginnings in the West is itself ‘capitalistic’ or ‘male’” (Hull 1988, 2-3).
Although Hull himself refers to studies of bees in which the leader bee was assumed male until evidence showed otherwise, he does not take this to be an example of science as usual, where evidence is the lynchpin converting followers to a theory. He rejects the idea that worldview could alter what is considered to be evidence. In doing so he puts stock into methods that are supposed to reliably find truth.
OBJECTIVITY & TRUTH
In a 1998 Presidential Address delivered to the American Philosophical Association entitled, “The Viewpoint from No One In Particular,” Arthur Fine asked whether opening up the halls of science, as advocated by feminists and followers of philosopher Paul Feyeraband, leads to a less reliable science, because it does not necessarily lead to truth, to the “really real” (1998,
15). The decrease in reliability correlates with an alleged diminishing of objectivity, where objectivity is assumed to be achievable, and is the guarantor of true knowledge, that which is supposed to separate science from less reliable forms of knowledge. In condensed form: the less objective a procedure is, the less true, and thus the less reliable. But what is it about objectivity that guarantees truth and why is truth taken to be so important? Fine details several possible meanings or components of objectivity. It could mean: including and strengthening the presence of multiple voices; being free from bias; or being impartial. Each of these, he claims, is
129 Chapter Five: Competition in Science important in some—Fine uses “locally appropriate”— situations, but none of them guarantees objectivity for every situation. There is no one general method that will work in every situation.
The best that any process for achieving objectivity can do is to assist in identifying error so that future work can be more helpful. When theories and ideas are questioned about flaws uncovered by these procedures, science moves away from error, thus progressing.
Fine’s analysis is meant to reveal that there is no distinctive mode of thought in science, but that a variety of methods are important for devising theories which are more, rather than less, accurate descriptions of the world. Science from this point of view is not “mysterious.” As Fine illustrates by use of analogy to politics, a scientific enterprise structured in this way is better than any other way. Referring to Winston Churchill’s observation that democracy is a horrible form of government, yet the best of all the alternatives, Fine is arguing that there is no one good way to do science, only locally appropriate methods. In part this is because even so-called objective processes do not necessarily yield objective products. Science, he says, is in the business of
“trust-making not real-making.” That is, science is supposed to meet practical needs in a consistent manner. It is not reliability in terms of Truth but in terms of practice that is important.
Our practices need to reliably help to solve practical problems. Solving problems is not necessarily finding Truth, although it is certainly anchored in the empirical world. The realm of medicine offers a good illustration. Whether or not our current understanding of medical science is 100% accurate is less important if we cannot predict and control natural processes. I do need to trust that the medicine that my doctor prescribes for my illness is going to help fight my illness and not make me worse. But, whether or not the underlying medical theories involved are wholly accurate is not as important: acupuncture might work well for me even if it does not seem explainable in (Western) medical terms.
130 Chapter Five: Competition in Science
Fine is echoing the feminist critique which begins with noticing that traditional science is not as objective as it takes itself to be, and that its biases are both a problem and a normal, unavoidable thing in science. As Fine suggests, and I agree, it is trust-making by science, more than truth-finding, that is key in science, given that there are multiple and conflicting ways of interpreting the world. Thus, a feminist project might demarcate ways to promote trust in reliable interpretations. Not surprisingly, as I discussed above, this involves increasing diversity.
In this case I mean the diversity of those who will be affected by the science.
For example, given the history of relations between African-Americans and the American medical establishment, trust-making procedures in medical research must incorporate persons who are responsible for protecting the interests of African-Americans.47 Of course, this is a problematic mission, since there is no one set of African-American interests. There might be some commonalities or sets of commonalities, but how could such a diverse group have a single set of definitive interests? One good place to start is with whatever the group deems important to itself. If someone is designated as on watch for the needs of the group, presumably the members would be somewhat more willing to trust the results of a scientific study, and various interests would have a way of being heard and assessed.
In another instance, more diverse participation in current biotechnology research might help reduce fear about and promote responsible practices in areas like stem-cell research, cloning, genetically-modified food and organisms, gene patenting, and DNA “fingerprinting”
(e.g., Begley 2001, Cajete 2000; Harding 1998; Schulman et al 1999). Even if the science (as in medicine) is justified theoretically, when its application is impossible due to fundamental conflicts with local beliefs, the science is rendered substantially less effective. To give an
47 See Lewis 2002 regarding the legacy of mistreatment of African-Americans by the medical establishment, and Angell 2001 for details on one of the most egregious medical crimes, the Tuskegee Syphilis Study. 131 Chapter Five: Competition in Science illustration, some large part of the global scientific community may believe that they understand how HIV is transmitted and how the rate of infection can be slowed, but for many reasons, it is extremely difficult to disseminate that knowledge in the local communities, plagued by infection, where frank discussions of sexuality are taboo. Greg Cajete, in a different case, claims that the scientific establishment is viewed with suspicion by Native American peoples and from the view of Native science because of histories of the misuse of science against their interests (Cajete
2001).
In fact these are not political concerns in a simple sense. According to Sandra Harding, science should be directed by political needs and moral concerns, rather than some abstract search for “truth.” The perspectives of non-scientists are potentially quite valuable to science in that they allow the non-scientists to formulate questions, observe data, and develop theories that are practically invisible to scientists. Diversity in science, then, can both direct science to different ends as well as provide different means for reaching those ends. Fine makes a similar point regarding his “locally appropriate” methods: “Who can possibly say in general what policy will always be best? What we need here [is] good, local judgment about particulars” (1998, 17).
He states “there is no relativism in working out procedures that are neither absolutely universal nor specifically local” (Fine 1998, 13). Science is a complex, value-laden enterprise, and trying to fit every scientific project under an overly specific, single set of guidelines will lead to more frustration than success.
Just as we believe it is important to have political representation that reflects the diversity of the population, it is reasonable to expect that a representative body also directs our science, especially when public funds are used:
It has been and should be moral and political beliefs that direct the development of both the intellectual and social structures of science. The problematics, concepts, theories,
132 Chapter Five: Competition in Science
methodologies, interpretations of experiments, and uses have been and should be selected with moral and political goals in mind, not merely cognitive ones (Harding 1986, 250).
An alternative political agenda for directing scientific research at the national level could focus on health issues rather than on space exploration. The U.S. could, as a nation, decide that finding cures for diseases that strike a large portion of the population is more important than putting a lot of money into, say, particle accelerators or space exploration. Those choices are political ones.48
Some theorists and commentators believe that women bring differences to scientific observation such that their results are actually different from those of male scientists. If this were the case, and I will return to a more in-depth examination of it below, then that would be a second and distinct reason to want to increase diversity among scientists, specifically with respect to gender. To reiterate from above, this is not a claim of essentialism. The claim is not that all women have special abilities that no men have. It is based in the general claim that outsider status is a rich source of input into the process of developing science. Nicholas
Negroponte in “Creating a Culture of Ideas,” written for Technology Review, offers this support for my claim: “Many engineering deadlocks have been broken by people who are not engineers at all. This is simply because perspective is more important than IQ” (2003, 34). Although I do not think that engineers have higher IQs than other people, I do agree with the general point that perspective can be more important than education or professional training. In this way, the particularities of gender socialization can be useful in science. For example, Barbara
McClintock, relying on a methodology characterized by stereotypically feminine qualities such as patience, care, and an eye for detail49, was able to pursue research that eventually led to a
Nobel Prize (Keller 1983). Similarly, persons of minority status might be able to “see” research
48 Interestingly, it is non-agreement among “experts” that might worry non-experts about the safety of something like a nuclear reactor. 49 See Longino 1987, p. 249. 133 Chapter Five: Competition in Science projects in new ways, ways that are effectively unavailable to most scientists (Harding 1986).
Often-cited examples supporting (to various degrees) the claim that women bring unique perspectives to science include the work of Barbara McClintock but also the primatologists Dian
Fossey, Birute Galdikas, and Jane Goodall.50 Carol Gilligan’s (1982) work in moral psychology relies in part on the claim that different genders are socialized differently. A different style could mean something as innocuous as the way one sets up a research lab. Marcia Barinaga (1993b) reports on one possibility:
Rather than fueling her students’ competitive fires by setting them against one another—something that’s often done in high-pressure labs—Searles [a female molecular geneticist interviewed for the article] adopted a lab management style in which she consciously refrains from encouraging rivalry among her students or holding them up to absolute standards for performance (1993b, 385, emphasis added).
Still, competition has become privileged over collaboration in an ideal picture of science (not to mention other disciplines like philosophy). In many venues it is considered more important for scientists to be competitive rather than cooperative. Why has this happened?
DUALISMS & GENDER
Consider one of many examples of masculine-bias discussed by feminist science critics: males of most species are associated with activity while females are identified with passivity, lacking initiative or ability to direct interactions between themselves and males. This masculine- bias was played out in research on human fertilization, when, in the 1960s, researchers portrayed sperm as very active and aggressive and eggs as passive, docile recipients. Decades later research has revealed that the egg in fact does play an active role and that sperm is not as
50 See Keller (1983) regarding McClintock and see Comfort (2001) for a reevaluation of McClintock’s outsider status. For a brief introduction to the work of the female primatologists see Morell (1993). 134 Chapter Five: Competition in Science
“macho” as originally portrayed (Begely 1997, 54; Longino 1990; Martin 1991). Why wasn’t this evidence available in the 1960s? Because most researchers were not looking for it. The egg was assumed passive, so attention was given to the role of the sperm. That research about the role of the sperm was perfectly good and valuable work. Because observation is theory-laden, the significance of the evidence was understood in light of the working theory and so some data was not seen as evidence due to the constraints of the theory. The theory was colored by social forces; in this case, stereotypes and expectations about the nature of male and of female.
It is hard for any person to step outside her own worldview and associated beliefs and expectations; scientists are not fundamentally different. Their work is informed by their worldviews. If biases like those favoring masculine characteristics are assumed to be separable from “good” science, there is no reason to consider them further. As I have been arguing all along, though, the work of scientists will in some ways reflect their broader views. An understanding of science based on strong objectivity, however, can recognize those real biases and use them to help strengthen science. One way to do that is to increase diversity of the scientific community. Another way is to closely examine what views one holds and how they function in a belief system.
Many feminist scholars have shown how concepts and especially metaphors can structure thinking. Recent examples come from Cuomo (1998), Keller (2000), Plumwood (1991), and
Merchant (1980). Their view is that we are particularly susceptible to patterns of thinking in which we organize the world into pairs of opposites, called conceptual dualisms. Examples of dualisms include: Human/Nature; Civilization/Primitive; Male/Female; and Reason/Emotion.
Competition/Collaboration is yet another. These conceptual pairs identify what is more (and also less) valuable in society; they also appear in scientific thinking. Dualisms are able to mold
135 Chapter Five: Competition in Science thinking in such a way that, for example, competition becomes cognitively separated from collaboration and cooperation. According to this reasoning, persons who are competitive cannot be cooperative, and vice versa. Merchant (1980) argues that as the dualisms are forced into a hierarchical order, one half of the pair is given a higher status. Thus competition has assumed a higher status than collaboration.
Feminists argue that dualistic assumptions create biases and therefore get in the way of objectivity. Viewing the world as sets of polar opposites blinds observers to those entities which exist somewhere in the middle. If, though, there are some researchers looking for one pole, other researchers looking for the opposite pole, and still more researchers looking in the middle, science stands a better chance of solving some important problems. This is the essence of strong objectivity: valued commitments in the sense of worldviews or paradigms are welcomed, and openly reviewed.
Western science is in many ways masculine-biased both in terms of its content and its practices. It tends to exclude women, and in a few places to exclude men, which is bad for science. As I will argue below, the competitive model that Hull sees as grounding science, assists in furthering a masculine-biased view of science and continuing to exclude many women from science. According to Hull, competition is the primary source of fuel for science, a necessary factor of doing science in the modern world, because everyone is motivated to sustain research over long periods of time in order to attain public recognition. The sort of model he employs has clear winners and losers.
An alternative practice in science, called a ‘female style’ by some as mentioned above, might be one means of recharacterizing science by including a wider variety of scientists and research styles. The words “female style” are used in some of the literature, but they are too
136 Chapter Five: Competition in Science essentialist even if they are not meant to be. So instead I will use “alternative” style. One alternative style, mentioned above, organizes laboratories to encourage collaboration and seek out fringe research projects (Baringa 1993b; Calandra 2001; Negroponte 2003). Although it is dualistic even to say that there are collaborative and competitive methods, because it is unlikely that any one person or group would act one way exclusively for a sustained period of time, I do think it is important to discuss them separately at first. Although both stances can be useful, I am suggesting that collaborative models are getting short-thrift in the models of Hull and other theorists. Giving scientists more options for interaction is very likely to benefit science. It is a feminist critique of dualisms that helps us understand the emphasis of competition in models of science, and feminist philosophical critique which can suggest appropriate corrective action.
THE PRACTICE OF SCIENCE: COMPETITION & ALTERNATIVES
What are alternatives to competition, by which I mean to focus on combative, hostile, and aggressive behaviors? Evidence for a different style can turn up in the individual practices and preferences of women scientists. Barinaga states that among women scientists there appears to be “less intrinsic appetite” for competition and, further, that women “shun fierce competition in their scientific careers” (1993b, 386). One strategy for avoiding highly competitive situations has been dubbed the “niche approach” as it appears that some women literally do not compete by choosing to work in areas with few investigators.51 This is what one researcher reported about women in computer science:
51 At the Grace Hopper Women in Computing Conference in September 1997 in San Jose, California, I heard several women computer scientists report that they used a niche strategy when picking a research topic as graduate students. For more information on the conference and the organization see http://www.systers.org/hopper/ . 137 Chapter Five: Competition in Science
Another motive for their theoretical orientation was that they preferred the culture in theoretical working groups: In theoretical computer science competence is less a question of competition, and success is less a question of elbowing one’s way than in other fields of computer science, because in theoretical computer science competence is well defined; it depends on the correct application of logical rules and on an ability to argue logically. For women, who typically are less socialised in competition and elbowing than men, the possibility of proving their competence without competing seems to be important in getting a footing in computer science (Erb 1997, 203, emphasis added).52
Clinging to objectivity as a way of safeguarding against bias is not a defensible practice.
Fine concludes similarly in his address: “The ‘traditional connection’ turns out to be no connection at all, and so there is no connection for the libertarians to break” (1998, 17). In other words, there is no general procedure, which would guarantee reaching the “truth of the matter” in science, and so there is no guarantee of consistently reliable results. At least, though, we can promote diversity that would naturally expose bias. Again, we return to Harding’s “strong objectivity.” By examining the roots of beliefs and practices we can see where the science under review has more or less of a local character. It reveals that there is not just one overarching standard for knowledge, but different standards for different purposes, which is just what Fine called “locally appropriate” standards. Notably this does not relativize all knowledge, so strong objectivity cannot be defeated by the charge of relativism. Harding stresses this: “Different cultures’ knowledge systems have different resources and limitations for producing knowledge; they are not all ‘equal,’ but there is no single possible perfect one, either” (1998, 19).
Additionally, it is better that we come to terms with the political grounding of modern science and straightforwardly address the moral and political issues involved with the pursuit of scientific research projects, as Harding advocates.
52 Perhaps it is really women who are more logical, since they are not socialized, and therefore distracted, by competition. 138 Chapter Five: Competition in Science
COMPETITION: NECESSARY TO SCIENCE?
Returning to Hull, he does suggest in Science as a Process that philosophy of science would do well to at least consider how factors besides logical reasoning contribute to the construction of science. Included as other factors are personal or social factors that affect scientists themselves. He is interested in devising “an evolutionary account of the interrelationships between social and conceptual development in science” (Hull 1988, 12).
Central to Hull's thesis is that, while it appears to be an external feature, competition among scientists is essential to the practice of science. Hull is aware of the appeal of the externalist's position—that personal or social factors can influence science through, for example, shaping what counts as evidence for a claim—yet he remains unconvinced that social factors can constitute a serious, substantive force in the creation of scientific ideas. In contrast to Hull,
Helen Longino (1990) explicitly acknowledges the tremendous success of science while maintaining that “social factors” actually do restrict access to the means of production in science.
For example, women do seek opportunities in the sciences but leave after becoming dissatisfied with the social environment.53
I agree with Hull that competition—a social phenomenon—does arise in science.
However, I disagree with him concerning its importance in a model of the workings of science, insofar as he defines it as an aggressive or hostile competition. It might have some function with respect to encouraging some scientists to pick up a research project just because some other scientist has said that the puzzle could not be solved (or something like that). And it clearly has a great deal to do with the financial considerations involved, both for a scientist personally as
53 It is not at all difficult to find data that supports the assertion that women have not fared well in this system of science. See, for example, the National Research Council's report, Science and Engineering Programs: On Target for Women; Sue Rosser's Teaching the Majority; McIlwee and Robinson's 1992 book on women engineers; Bradley’s (2000) “Too Few at the Top: Women in Science”; and Mandalvilli’s October 2001 report on a gender gap with regard to salaries in biology. 139 Chapter Five: Competition in Science well as for her institution. But hostile competition is neither a central nor necessary component of science.
Hull says that social factors create a force within science, but he stops short of acknowledging their role in the greater society. If social factors are an influence within science, then they must also be influential outside of science, and if that is the case then why would aggressive competition as outlined by Hull be restricted only to science? He accepts that personalities fuel the testing of hypotheses—that aggressive competition exists and is necessary for progress in science—but he also rejects that Western science is male-biased.
Both Hull and Longino agree that the social structures of science, particularly research group connections, are important. The difference is that Hull does not accept that social factors could have as great a degree of or as consistent an influence. In contrast, Longino claims that not only have women qua women historically been denied access to scientific research, but also that few scientific projects adequately address problems that would benefit women in particular.54
Moreover, the vocabulary and metaphors used in scientific practice sometimes reveal unconsciously sexist commitments, as was illustrated earlier in the discussion of human fertilization research. It is the widespread use of such vocabulary, in addition to the plethora of actual scientific projects, which lends credence to the claim that Western science is biased against and sometimes harmful to women. To name just a handful of concerns with respect to women and science (these happen to focus on health and medicine): insufficient research on contraception; insufficient research using female subjects to determine female-specific treatments and dosages, especially for major diseases affecting women like heart-disease and stroke; environmental pollution and its effects on women, especially those of child-bearing age;
54 Sue Rosser (1997) offers, as an example of a women-centered problem, the lack of thoroughly tested, widely available, affordable birth control (presentation at the University of Cincinnati, Fall 1997). 140 Chapter Five: Competition in Science and a lack of research on older and elderly women. Although notably the summer 2002 findings about hormone replacement therapy are a step in the right direction, they still expose the very real problems I have been discussing.
This is not to argue that science is not successful. Clearly science has done tremendous things. To highlight three distinct examples: increasing crop production; devising medical techniques for saving and improving lives; and constructing housing materials that are more energy efficient. The issue here is that the price for success seems to include “systematically perpetuating women's cognitive and political disempowerment" (Longino 1991, 667). Longino concludes that the practice of science—its methodologies, models, and theorizing—must be scrutinized in order to determine if there is some inherent limitation to science as it is currently practiced. I believe that overemphasizing the competitive nature of science, as described by
Hull, creates one such limitation. Less competition could attract to science those who are
“naturally” less competitive, many of whom might be women (see opening discussion). This might be good for changing the course of scientific research so that it works to benefit women, or at least does not actively pursue projects that might harm them.
Hull argues that the sometimes-fierce competition of the kind he has observed between groups of scientists ensures that poor data or erroneous theories or other problems will be rejected or solved in the normal course of events. According to Hull, scientists thrive on a system that pits them against each other in a quest to be publicly recognized as contributors of knowledge. Hull is careful to point out that a scientist’s world is also often insulated from such criticism by supportive research groups, so not every aspect of science is adversarial. Since a scientist's name "lives on" in his work, it is in the best interest of the scientist to have his name appended to some important piece of work which is subsequently often cited by others.
141 Chapter Five: Competition in Science
Although I do not want to deny the motivation of the credit and reward system, it does not seem a wholly sufficient mechanism for consistently motivating all scientists throughout their careers, nor for promoting the best possible science. For one thing, there just are not an adequate number of rewards or recognitions. Most scientists work hard all of their careers without ever receiving a significant reward. Perhaps they have contributed to a handful of publications, most of which are about very routine science. Clearly, something else must be motivating these researchers. So it seems fair to question the legitimization of cutthroat competition when something else less fierce would work. The problem here is not so much with competition per se, but rather the nature of the competition. Aggressiveness serves to promote only a few persons, while it can be quite hazardous to many others. For example, if it is the case that women tend or are trained to behave less aggressively than men (especially in contexts which are male-dominated), then a highly aggressive kind of competition could effectively silence most women (and men) who are not combative.55
Hull explains how science works by showing how some social and conceptual forces constitute part of the process that molds science. He does recognize the potential effects of factors like race, religion, and gender on the practice of science. Nonetheless he is very much committed to a model of science founded on evidence, reason, and argument, isolated from social factors. Yes, he concedes, science does take place in a social setting:
Periodically in the history of science, national origin, sex, race, religious affiliation, and the like have influenced the spread of scientific ideas, but such influences have been extremely variable. The effects of affiliation of scientists with particular
55 Although not about scientists, consider the following written about the practice of philosophy under the same aggressiveness for a glimpse at what is meant by the “chilling effect”. Reporting from his students of their distaste for the “palpable feeling of combat” at presentations, Norman Swartz writes that “certain anecdotal evidence suggests that aggressive challenging of guest speakers’ theses has chilling effects on many of our students” (Swartz, p. 3). To be sure, Swartz is not the only person who has noticed or written about this “trend” in philosophy. Feminists such as Moulton (1983) have called it “antagonistic philosophy” and Gary (2001) discusses, somewhat tongue-in-cheek, “proper manners” for constructive criticism, which notably is not adversarial in nature. 142 Chapter Five: Competition in Science
research groups have been constant and widespread (Hull 1988,15 emphasis added).56
While Hull does not find race or gender to be irrelevant in scientific research, he does not see that they have been consistently influential in the substantive work of science. This contrasts with the feminist argument that there exists a systematic bias in science in favor of the masculine. That research groups and educational settings continue to dissuade women from pursuing careers in science is evidence of systematic bias as well as the strength of a claim that a social epistemology is the primary force in science (e.g., Mandavilli 2001).57
According to Hull, within the group—and the focus is on research groups as the most important unit—individual scientists have a natural curiosity that functions to motivate all of the requisite hard work. Ideas are generated by individuals, are tested with/by others, and are
"disseminated" in groups. Groups have a tremendous impact on the scientific process.
Success—or the promise of success—further motivates researchers to continue working and produce well-supported work.
Success for Hull means success in publication. Publication is both foundational and beneficial to science because it allows public access to data and theories, and authenticates the material at hand (Hull 1988, 323). The mechanism for scientific development is one that relies on a structure of credit and verification. Included in the process of citing others is a system of verification, since it would be risky to quote someone else's work in support of one's own unless one were sure that it could withstand scrutiny. According to Hull, success in publication is measured by the number of times a scientist is cited; the more often a scientist's work is used by another, the more successful the scientist. The system of scientific publication rewards the
56 These effects are not always positive, however. Fraud is often perpetrated under the name of an established, respected senior scientist or prestigious laboratory. 57For recent statistics (2002) about women in science see http://www.chillyclimate.org/statistics.asp linked from the Association of Women in Science webpage. (Visited 25 Aug 2002). 143 Chapter Five: Competition in Science person(s) who gets his work into print first, although it does not necessarily filter out the best work.58
A better measure of good science might instead be influence, rather than temporal priority. This approach is compatible with Hull’s own account. It turns out that women, for various reasons, might produce better work by the measure of influence (Barinaga 1993, 390).
Still both of these concepts—being first and having influence—have their problems. For example, the recent upheaval in medicine over hormone replacement therapy (HRT) shows how someone’s work can be very influential because it is very profitable, when as was recently discovered, the research supporting the widespread use of HRT for women going through menopause is much less conclusive than many people—doctors, patients, and researchers—once believed.
Unfortunately, regardless of how much time a scientist has spent on a research problem, the credit often goes to the one who publishes first. Virginia Morell, in her 1993 article about three women primatologists, highlights a serious problem in this system, namely, some research simply requires years in the field to become familiar with subjects. This commitment is none too compatible with traditional standards for publishing one's research: "Galdikas' slowness in publishing--in fact, her overall indifference to conventional academic standards--is what made her scientific success possible" (Morell 1993, 425, emphasis added). It could be that Hull is mistaken in characterizing competition as equally relevant to all scientific pursuits. A system of credit is a motivator in a less-competitive environment, as exemplified by the success and productivity of individual research groups. This is important: Hull and I could both be correct.
In some fields aggressive competition works and in other areas it is detrimental, but it tends to
58It is not exactly clear what “first” means, however. Sometimes it means first in the most accessible place. (Thanks to Bob Richardson for bringing this to my attention). Hull himself wonders "What influence do various sorts of 'priority' have on science?" (p. 521). 144 Chapter Five: Competition in Science have a gendered effect, with women paying a price for being competitive that men do not pay.
Women are still socialized to be nurturing and cooperative, and when they are not they are taken to be too masculine and “bitchy”. Even within the same field one could be cooperative with one’s own research group, while the group is competitive with other groups (see Gender &
Competition in Computing section below).
Hull’s model of science centered on competition strikes me as limited when applied generically across Science. Competition that fosters the best work of scientists without pitting them against each other could be the force that fuels science. His position includes no way to filter out those social influences that might discourage women or others from participating, which are not desirable on moral or political grounds. Social factors like learning to be aggressive seem to help science—in fact are supposedly essential—but throughout Hull’s exposition, other social/external factors like gender are excluded as irrelevant. Hull cannot have it both ways. A feminist conception of competition, as outlined by Longino, may provide a richer conception of scientific interaction, as well as a more realistic picture of scientists.
What does it even mean to say that scientists are competitive? It is possible that a scientist could be cooperative within his research group and still be hostile with members of opposing groups. It is quite plausible that scientists are in fact highly competitive in certain arenas (between research groups), but in others are perfectly willing to collaborate (within the research groups), without ever being rightly described as exclusively one or the other. A fuller explanation below of dualisms will be helpful for seeing that persons can be both. Hull has perhaps neglected to explain enough here, and his own words acknowledge that the system he describes might in fact be unfair:
if I am right about the central role of competition and aggression in science and if these characteristics are more common among males than females (regardless of
145 Chapter Five: Competition in Science
why), then there may be a sense in which the social organization of science is male- biased (1988, 390n4).
Part of the problem, also, is over reliance on individual scientists working in groups, rather than groups as the basic level of epistemic agents, when science is actually highly likely to include work at both levels.
Hull also notes that "women are less likely than men to go into most branches of science" but that if they do enter into science, their contributions are treated no differently than men's.
While this may be an accurate characterization of the treatment of women scientists, amazingly,
Hull seems to completely miss an important point: that science disproportionately excludes women is a problem, and a very serious one at that. Furthermore, women who go into science under these conditions are likely to exhibit “masculine traits” in the first place. Second, excluding women often means that concerns of women are not adequately addressed.59 Third, and more interesting from an epistemological point of view, it appears that some women can bring a different perspective to science, one that provides them special insights typically unavailable to many men. If women are excluded intentionally or not from science, either because they do not picture themselves, or others such as teachers or peers do not picture them as fitting into the model, one can argue that science and therefore, society, is being short-changed.
Morell lends support to such a claim by saying that "some insiders in the field" believe that "a male approach" could not have resulted in the discoveries of the female primatologists Dian
Fossey, Jane Goodall, and Birute Galdikas (1993, 420). As Morell reports:
59Please see, for example, Laurence, L. & Weinhouse, B. (1994). This book documents how scientific research often overlooks the specific needs and interests of women. Two examples of this neglect include the Aspirin a Day Heart study that included no women among its test subjects, and the failure in some drug studies to include female test animals or women at the later stages because of the "difficulties" of factoring hormone fluctuations due to the female cycle. 146 Chapter Five: Competition in Science
Leakey thought women were better than men when it came to studying wild primates. . . ., Leakey. . . 'trusted women for their patience, persistence, and perception--traits which he thought made them better students of primate behavior (1993, 420-1).
If women’s perspectives might enrich scientific inquiry, we would be well advised to provide a supportive environment for both women, and men, those who are “naturally” aggressive and those who are not, in science. If Hull is right in assigning import to the research groups, then women’s lack of equal access to them is a huge disadvantage for all.60 Claiming that men are essentially one way and women essentially another is dangerous and it could hardly be true.
Essentialism is often used to justify inequities and injustice. An examination of some intellectual history should be illuminating.
BRIDGING DUALISMS
To recap, one claim of some feminist theory is that the rationalist mindset has worked to divide the world into polar opposites and hierarchical dualisms. For example, under a dualistic schema, humans are definitively and hierarchically distinguished from non-humans, or nature.
Similarly, the civilized world is sharply demarcated from the so-called primitive one. Males are easily grouped into one class, all other persons are females. There is no ambiguity, no one cannot be placed in one of the two groups. Those persons or behaviors that are rational, obviously cannot be emotional, on this view. A crucial realization by feminists was uncovering the unequal valuing of the halves of the dualisms (e.g. Merchant 1980). Generally certain halves—the rational, civilized, thinking aspects were assigned to males and those were the characteristics that were valued. The remaining parts—primitive, female, emotional, and
60 See, for example, Harding's (1986) The Science Question in Feminism, Rosser's (1995) Teaching the Majority, Keller & Longino's (1996) Feminism and Science, and Kohlstedt & Longino's (1997) Women, Gender, and Science: New Directions for several accounts of how and why unequal access is occurring. 147 Chapter Five: Competition in Science natural—were all connected by virtue of their devaluation. Val Plumwood (1991), observing that there are many things in the world that do not comfortably fit into this scheme, argues that these dualisms reduce reality into just one or the other of the categories, but never both of them.
The dualistic framework effectively erases any blends or mixtures of the two. Thus, on this scheme, either a person is rational or emotional. Men are expected to be the rational ones, while women are assumed to be emotional and therefore less rational. Any kind of response that is, for instance, both rational and emotional must be flattened or simplified in order to be categorized.
Extending this argument, behavior defined as competitive, under Hull’s definition, may automatically prevent the scientist from collaborating at a later point. The scientist has already been labeled as a competitor, which is incompatible with being a collaborator. I will show below with the help of Longino’s work that redefining competition can however avoid the need for such strict compartmentalization, and can help us move beyond dualisms.
The dualistic mindset is not just theoretical. Categories can literally impact how we see ourselves and can affect behavior. Plumwood shows how conceptions of the self are refracted through these categories, such that the self is likewise flattened as it becomes identified with one or the other, but never both or any combination of, these dualisms. Competition and its
“opposite,” collaboration, do seem to fit easily into the set of dualisms. Notice, though, that the fit is too easy and overly simplistic. Just as with the other dualisms, this one also obscures more complicated pictures of persons (or groups of persons), their relationships with others, and their modes of interaction. It reinforces the false notion that the appropriate modes of interaction are either competition or collaboration. Plumwood’s analysis reveals, in contrast, that there are many other possibilities; the nature of our actual relationships is much more complex and dynamic. Yet competition as generally understood reinforces the isolation, separateness, and
148 Chapter Five: Competition in Science independence championed by the hierarchical system of dualities. Independence and its related traits are not only associated with males, they are traits that are praised in males. An independent man is valued and partially fulfills the qualifications for a good man. An independent woman, however, is not similarly viewed. Moreover, by being independent she is more likely to fail to meet other defining characteristics such as being nurturing, passive, or emotional. This is how the rationalist schema defines men and women, and how it creates mutually exclusive definitions by combining the “appropriate” halves of the dualisms.
Redefining competitive scientific practice, as well as the scientists themselves, in a manner informed by Plumwood's 'diagnosis', would benefit both male and female scientists by allowing them to exist under less rigid definitions and expectations, and may also provide a more accurate picture of how science is actually practiced.
Ultimately this change in view would benefit science, too, since many more people could participate in science without having to overcome stereotypes and rigid, confining expectations, established and enforced by scientific culture. This is especially needed when that culture is internally maintained, because extra-culture solutions will likely fail. Erb (1997) reports a study which “found that the dominance of the freak culture [in computing] is produced and reproduced by all members of the computer science department, that is not only by the computer freaks, other male students and the male staff but also by the female students of this subject” (p. 202, emphasis in original). Unfortunately for philosophers and other theorists, allowing these categories to be more complex, open, and less rigidly defined means that they are much harder to study. But the benefits are more productive science, especially in new areas on the edge of old fields, as artificial intelligence is and continues to be, or in the new fields of biotechnology where the component disciplines are so varied, no one person could really be competent in each area
149 Chapter Five: Competition in Science
(Bahls & Schacter 2002; Calandra 2001). In a social epistemology framework cooperation is much easier as well as expected, since too much competition is detrimental to group cohesion.
Working on an open project is easier in a group because no one person must figure everything out and other members of the group can compensate for any one member’s weakness (Wallace
2001, 2001b). In a context that encourages competition, cooperation would be seen as threatening or less rigorous and thus unacceptable, because it is considered diametrically opposed to competition. Helen Longino’s alternative model of competition suggests ways to productively challenge this dualistic framework.
LONGINO'S MODELS OF COMPETITION
There is more than one way to understand competition. Helen Longino identifies two in her 1987 paper, “The Ideology of Competition.” The first model seems to be what Hull has in mind when he discusses competition. The structure of this model necessarily excludes multiple winners. Ties are impossible in this model. The game/contest ends only when a winner has emerged. Think of an NCAA championship basketball game: there are not two national champions! It is a winner-take-all model. Importantly, though, there is an alternative model of competition. In this model, there can be more than one winner. Ties are possible in principle, even if they rarely actually happen. Differences between the contestants usually result in distinct rankings: that runner trained better, this skater was not feeling confident. Longino's claim is that, "The blanket condemnation of competition as well as its uncritical idealization ignore the distinctions suggested by these examples (Longino, 1987, 250). This quotation underscores the complexity of “competition” and thus supports an important part of Plumwood’s argument.
150 Chapter Five: Competition in Science
The question is which of these models more accurately describes science. Hull's description of science more closely resembles a model in which there is a single winner. Yet there could easily be a scientific structure that employs the alternative. The first model dictates not that all contestants must win, but that it is possible to have more than just a single winner.
This model would not lower the standards for scientific excellence. All scientists would still be responsible for striving to find the best data to support their arguments, for devising well- conceived experiments, and attending to the other practices of scientific research. Each scientist would work just as hard as on the other model, but no longer would it be possible to ruin someone's career by beating her to publication. This is somewhat oversimplified and overstated, but it does demonstrate the plausibility of the alternative.
Longino notes that those who endorse the winner-take-all model tend to assume that there is a scarcity of goods. Is this true in science? According to Hull, the "goods" sought after in science include credit given to scientists for their ideas and research. These goods are scarce in part because of the priority system in publishing. It is fair, though, to question the value of this system. Does it really encourage the best in science? Does it, alternatively, encourage fraud? Is there some alternative that could work just as well, or is there something inherently superior about the priority system? To quote Longino,
Our conceptual linking of competition with domination, hierarchy, and scarcity prevents us from appreciating the value of competitive challenge in developing skills and talents and ultimately undermines our potential to change ourselves and our worlds (Longino, 1987, 256).
151 Chapter Five: Competition in Science
COMPETITION FOR WHAT & WHY?
The value of competition depends on what is scarce. Ideas are not scarce, but money and positions of power are. Can we avoid encouraging competition, after all, given that we live in a globally capitalistic economy? At some point it would be prudent to concede that there is an inescapable level of competition embedded in this system. Of course money has significant influence in our society and it would be naïve to think that science is immune from such pressures. Profit is a powerful motivator in this system; products or procedures that can be patented or otherwise marketed are going to get funded, if only by private money. The case of
LYMErix reflects this state of affairs (Weintraub 2001). Allegedly a big pharmaceutical company, perhaps condoned by the Federal Drug Administration (FDA), conducted less than adequate research before rushing a vaccine for LYME disease to market. The reporter,
Weintraub, observed that
In the end, the problems of LYMErix may be rooted in something far less organized than insidious – the hubris of medical science, which has sold its soul to industry for the funding it needs to survive. To be true to itself, science must acknowledge the gray areas, but to fit the needs of business, it must deal in black and white (Weintraub 2001, 9).
Money does produce “asymmetries,” especially resource asymmetries (Kitcher 1993, 311).
Sometimes resources advantage one country over another, sometimes it is a lab or a company.
Obviously capitalism will not be going away anytime soon; neither will sexism or racism. The entire situation is complicated by very real financial stakes, which are made increasingly important in today’s scientific research. Profit can be a tremendous negative factor in openness and idea sharing. Especially in computing, research and business are becoming increasingly difficult to separate:
152 Chapter Five: Competition in Science
being open about ideas was particularly hard for computer scientists because people saw riches coming from not sharing their ideas. Students would withhold ideas until after graduation. As one person held his or her cards close, another followed, and as a result, many research labs declined in value and effectiveness (Negroponte 2003, 35, underscore added).
A related issue concerns the identification of the innovators in science, clearly a difficult task. Providing the proper conditions for their work depends on who they are. Are they usually at the fringe of a discipline? Are they the big shots or the unknowns? Senior researchers ought to be able to command enough respect, and enough resources in turn to get their ideas off the ground, whereas novices might very well flounder. Again, lab directors and those in similar positions have the power to create an environment conducive to the production of good science.
Laboratory and program directors, as well as professors and mentors, can find ways to lean towards cooperation
Encouraging as many ideas as possible at the lower levels (within a lab as opposed to between two labs), can only further the goals of science. Recent trends in architecture, for instance, echo the value of collaboration. Laboratory design elements include more open spaces and fewer permanent walls so “scientists can compare notes and work in teams” (Fiske 2000).
The story of Barbara McClintock and her research on corn is a case in point (Keller 1985).
Probably a giant corporation would not have funded her research at the beginning. But if it had never had the opportunity to develop, we would be missing out on some important research.
Evelyn Fox Keller, as is well-known, raises issues about the ideology of science (Keller 1983,
2001). Like Hull and Longino, Keller identifies social aspects like competition as fairly important for the interaction of scientists. Her view is more distanced and she stresses diversity as an outcome:
153 Chapter Five: Competition in Science
The contest many scientists feel themselves engaged in, . . . reflects the contest they feel themselves engaged in with human others. . . . Fortunately, however, the practice of science is in fact quite different from its ideological prescriptions. Scientists differ greatly in their approaches to their subjects and in their styles of work. . . . Indeed such differences are essential to the vitality of the scientific enterprise. They are also responsible for a basic thesis of this [Keller’s own] book: actual science is more faithfully described by the multiplicity of styles and approaches that constitute its practice than by dominant rhetoric or ideology (Keller 1985, 124-5, emphasis added).
Perhaps the situation determines the need for hostile-competitive behavior. If it is especially urgent a scientist might use data before completely testing it. Conversely, an opponent might attack the data or theory more swiftly when the stakes are high. In such cases it is possible that the “merits” of a hostile competition might outweigh its harms. There is a substantial danger in this though, as shown in the previous chapter’s discussion of equipoise. It might also depend on the stage of development of the idea/theory/hypothesis. Relaxing the restraints for generating and developing ideas seems the most promising time for open dialogue.
On the other hand, during human drug trials more scrutiny is welcomed in order to protect human health and life, even though hostility may only rarely be appropriate, where appropriate is determined by the community. Thus, while it might be important to train scientists to know how to pick their fights, so to speak, as far as the long-term interests of science are concerned, hostility is not necessary. Hostile competition should not be the norm or default for every level of science, especially at the early stages of education and training where it seems unnecessarily destructive.
Importantly, our modern Western system of science is not the only model; there are others in which domination is not emphasized. A system that ostensibly equates quality with the result of an aggressive competition, is hard to escape. But what is evidence of quality? What constitutes progress in science? One answer would be the success in terms of practical
154 Chapter Five: Competition in Science application of scientific research. Hull claims that the number of citations is a good measure of the success of an individual scientist. Barinaga reports that one study found that women were actually cited more often than men were, even though they did not publish as many articles
(1993, 390). Women were publishing fewer, but more thorough papers. It is not necessarily that women are more influential in science, but that due to social roles, women work differently.
Their social roles include different priorities at work as well as different views about the place of work in their lives. For example, white women are often not a family’s primary income and there can be different things at stake for a woman’s ego with regard to success.61 In some cases in the past, ignoring the standard rules about publishing was the only way to do the research, as indicated above regarding Birute Galdikas' work with orangutans that made it necessary for her to spend years in the field, a practice that is not conducive to frequent publication. On some level it seems that Galdikas was motivated by desire for the truth, or love of her subjects. Was she going out on a limb to do something that she felt needed to be done, but would not get done by the other researchers? If she did not, did her mentor, Louis Leakey? For whatever reason,
Galdikas does not seem to have been motivated by the same kinds of things as her fellow (male) primatologists. Maybe it is the case that there is always a minority group (regardless of race, gender, etc.) that is defined by its outsider approach.
61 See Ellis Cose, “The Black Gender Gap,” Newsweek (3 March 2003), 46-51 and Allison Samuels, “Time to Tell It Like It Is,” Newsweek (3 March 2003), 52-5, for examples of how black American professional women differ from white women and from black men. 155 Chapter Five: Competition in Science
GENDER & COMPETITION IN COMPUTING
In fact, computing culture is complex. The findings of Häpnes & Sørensen concerning
Norwegian hacker culture indicate that there is both competition and collaboration in all male computing groups:
The combination of competition and collaboration, of individualism and caring, may be typical of male middle-class groups, inside or outside of science and technology. What we see here is probably a useful corrective to the picture of male scientists as being dominantly competitive and individualistic. Without some collaboration and caring, a group or a community would find it difficult to exist. Even in settings of competitiveness and individualism, there may be an undercurrent of collaboration and caring which is important to the work performed (Häpnes & Sørensen 1995).
Women might look elsewhere, i.e., not major in computing, rather than engage in a competitive academic environment. From a study about recruiting and retaining women as graduate students in computing, it is important to actively counter negative stereotypes and misperceptions about computer science and engineering, including “You cannot be successful in graduate school unless you are highly competitive” (Cuny & Aspray 2002, 171).
“Competition” thus signals several interrelated practices and behaviors. It can impact, for example, pedagogy and admittance practices, cultures or sub-cultures, and the structure of research groups. Competition can force out potential positive contributions, as when in some sciences and in computing particularly the discipline is viewed as female-unfriendly. Moreover, it does not really serve to cull out negative contributions. It is lamentable that science is taken to be overly competitive—that the model of science which informs our view of science is structured competitively—on the ground that competitiveness will protect the integrity of science by keeping it objective. This is part of the concern of someone like Grundy who objects to computer science curricula based fundamentally in mathematics. It does not need to be that way in the culture, which includes pedagogical practices (Turkle & Papert 1990). If one learning
156 Chapter Five: Competition in Science style is favored, other styles are effectively closed out, such that students with the different styles could be disadvantaged, and for no reason that pertains to being an effective computer scientist.
This is what concerns me:
Beginning programmers have historically been taught to decompose problems logically into sequences of steps. . . Goal-directed, endpoint driven planning is seen as preferable to a more fluidly serendipitous exploration. This approach privileges certain epistemological styles over others and leads successful students to fluency with one particular set of techniques, discouraging others (Stein 1999, 478).
Regarding a change in guiding metaphor in computing from one of computation to one of a community of interacting entities, Stein suggests: “A further benefit is that it reaches out to those whose natural epistemological styles may not accord well with the purely hierarchical, functional, black-box-based approaches common in the traditional paradigm” (Stein 1999, 495).
She continues,
This way of approaching computation [the new interactive metaphor] has also profound implications for the kinds of thinking we do. For students, it means that we harness their native intuition about how to survive in an inherently concurrent and asynchronous world. We never put on the blinders of calculational sequentialism. Students and professionals alike are encouraged to interact with computational artifacts, to experiment, to tinker. And we no longer silence those students whose problem-solving skills derive from experiential rather than mathematical and logical approaches (Stein 1999, 501).
Competitiveness does seem to have had an impact on women going into computing, although it cannot be the sole reason for the “shrinking pipeline.” As Christina Björkman argues, something more fundamental regarding the nature of computing, and especially the epistemological bases, or at least perceptions of such, must be at work: “The approach taken [in her recent work] is to move focus from women/gender to the discipline of computer science itself. This means the question is raised towards a more general level, towards ‘the science question’, discussing the discipline its paradigms and knowledge processes” (Björkman 1999,
157 Chapter Five: Competition in Science back cover). Furthermore, as a social endeavor science is necessarily collaborative to some degree, as noted above by Häpnes & Sørensen (1995).62 From an open letter to a prospective female college students written by two computer science professors trying to encourage more women into computing:
Perhaps you are concerned about the environment being overly competitive and cutthroat. I think you will discover what a truly cooperative discipline computer science actually is. One of the most important skills that we try to impart to our students is the ability to work in a team setting. I predict that the bonding that takes place in the computer lab as you work with your colleagues to finish a project by the deadline will be included among your fondest memories of college (Prey & Treu 2002, 19).
CONCLUSIONS
How can we account for the apparent existence of cooperation even within this so-called competitive system, in which competition has definite negative connotations? Actually, the existence of those groups proves my point. Collaboration is very fruitful, that is why research groups do collaborate. The members know that their own work will benefit from review by others. The previous chapter on social epistemology presented evidence for thinking that groups, not individuals, are the primary epistemic source. Chapter six will supplement that discussion by referring to cognitive science research which demonstrates that we do not learn or think strictly as individuals. Competition is just one aspect of socialization that discourages the participation of women in the core groups. Having to choose between family and work is another, and so is having to live under the social stigma of being a “nerd” or otherwise less than feminine.
Breaking up the competitive model involves identifying both its weaknesses and strengths. Hull highlights some of its strengths. The weaknesses of the aggressive model, however, make
62 Note that the competitive-collaborative dualism itself is objectionable to a feminist account trying to dismantle binary logic (see chapter four, note 4). 158 Chapter Five: Competition in Science science unnecessarily exclusionary and homogenous, can decrease the presence of novel approaches, can promote rushed experimentation lacking in thoroughness, and can encourage fraud.
Hull's position, while interesting and insightful in many aspects, is ultimately unsatisfactory. His analysis could benefit from the insights of feminist philosophers of science.
First, his conception of competition could be broadened, as could his view of scientists and science. His stress on its importance is not necessary. As reported in the journal Science in a special series on women in science, there are women who explicitly shun highly competitive models for setting up laboratories and for researching in groups (Barinaga, 1993, 384-7).
Second, his analysis leaves no room for accounting for the highly skewed numbers of white, middle class, men populating the ranks of science. Interestingly though, if there is something about competitiveness that attracts men and repels women, perhaps Hull has already indirectly explained the skewed numbers. McIlwee and Robinson, authors of Women in Engineering:
Gender, Power, and Workplace Culture, write that “To the extent that the image, activities, and interactional styles of an occupation are male-identified, women will have a harder time conforming to the cultural expectations of their peers and superiors” (1992, 190). Hull maintains
"Things such as social class and sexual preference have, at most, contingent, idiosyncratic effects on the conceptual development of science" (1988, 514). Yet his reliance on a particular kind of competition is itself an external, contingent factor.
In contrast, the positions of Longino, Plumwood, and other feminists provide space for understanding the factors that helped to foster the environment that created the imbalance. One change advocated by many feminists is to increase diversity among scientists, since people with different cultural experiences may bring different perspectives to their work. To quote Barinaga,
159 Chapter Five: Competition in Science
"science can be improved by the recognition that cultural context does influence one's perspective" (1993, 393). Unfortunately, it appears that aggressive, competitive environments block many women from participating. More strongly, McIlwee and Robinson report that
“feminist scientists argue that scientific developments are hampered by the posturing, competition and secrecy that so often pervade the field” (1992, 142, emphasis in original). By relying on a model of science which does not depend on hostile competition, we can foster an environment favoring more cooperative and collaborative interaction, which will benefit science and in turn will benefit society. In one way this is so because it allows more freedom in science.
Freedom in this case allows exploration of more complex ideas and collaboration of problems across disciplines (Calandra 2001; Negroponte 2003).
160
CHAPTER SIX
EMBODIMENT AND EMBEDDEDNESS IN AI:
DRAFTING A MODEL OF INTELLIGENCE
FEMINIST CONCERNS IN ARTIFICIAL INTELLIGENCE
Feminists including philosophers have assessed a range of scientific fields including biology63, genetics64, psychology65, archaeology66, physics67, and medicine68. In many cases they have employed the tools of philosophy of science to answer "new" questions about the role of women as subjects and objects of scientific study69, the appropriateness of publicly funding military and other projects of domination70, or the use of scientific studies as the basis of public policy decisions.71 Following in these footsteps, I have chosen to examine computer science, particularly by investigating artificial intelligence (AI) in some detail. The previous chapters have been building a foundation for a feminist analysis of a specific AI project, found in the next, final chapter.
I begin with a few words on feminist work in computing and then briefly examine the concept of rationality. As this is a feminist project, it is important to look at the women involved in computing, and it is important to situate this project within other feminist analyses of
63 Haraway 1985; Keller 1983; Longino 1990; Mahowald 2000; Martin 1991; Okruhlik 1998. 64 Keller 2000; Nichols 2001. 65 Flax 1996; Gilligan 1982. 66 Wylie 1992, 1996. 67 Traweek 1988, 1992. 68 Conley 1998; Laurence & Weinhouse 1994; Tuana 1996; Weasel 2001. 69 Conley 1998; Schibinger 1999. 70 Cohn 1996, 1996b; Harding 1998, 1993; Longino 1997. 71 See more general treatments in Barinaga (1993); Eisenhart et al (1998); Harding (1998, 1996); Kohlstedt & Longino (eds.) 1997; Matyas & Dix (Eds) 1992; Morell 1993a, 1993b; Nelson & Nelson (1996); Oldham 2000. Chapter Six: Embodied & Embedded Epistemology computing. The bulk of those analyses are more sociological, rather than philosophical in nature. Hence, the importance of reviewing the concept of rationality, a concept which has great importance in artificial intelligence research, for its philosophical implications. From that I move to investigate a view of intelligence and cognition based on embeddedness and embodiment, which I will define and discuss in detail below.
Computer science is not a gender-neutral field. On a pragmatic level feminists have the same concerns with computing as in science generally: not enough women, too much male bias, and no clear answers as to why this state of affairs persists, despite many efforts to change it.
Some of these feminist studies assess usage of and training on computers. Such researchers ask questions about how programming should be taught. Others focus on representations of women and minorities in computing: Why might it be important to think about users? Might programmers under-represent certain groups? How is the dominant cultural group or the most affluent groups represented by programmers? What do those images reveal about the programmers’ views regarding the identities of the other folks? Other critics consider the allocation of resources to and the role of computing in both local and global communities.
Although feminists and non-feminists alike can raise the same concerns, a feminist critique can provide a unique analysis, which can see problems that others fail to notice. For example, for feminists, gender is relevant in many respects to AI. Alison Adam for one has asserted “that gender can be used to explore alternative epistemologies represented within AI systems” (1993, 311). I disagree that there are multiple epistemologies, although I appreciate
Adam’s point. I would say that there is a richer, fuller epistemology, with embodied and embedded dimensions, which is not broadly represented in AI. By using the lens of gender, those other dimensions are noticed.
162 Chapter Six: Embodied & Embedded Epistemology
The conceptual issues are just as interesting, however, and are perhaps more important in furthering feminist values and goals.72 I will only briefly touch on some of these challenges, restricting my remarks to AI. Feminists in philosophy of science and elsewhere assess the concept of rationality, which I discussed in chapter two, so it should be of no surprise that it turns up as an issue for feminist scholars in computer science, too.73 Many question the definition and application of rationality to the particular issue. In this chapter I scrutinize some of the results of privileging rationality in AI. I argue that it risks the displacement of embodied and embedded perspectives, a loss with serious negative consequences.
Rationality is a tremendously important concept in AI because agents are deliberately created to achieve certain goals, and the very essence of rationality is setting and meeting appropriate goals. Insofar as they can follow the rules to meet their goals, agents are said to be rational. Rationality, however, is not a neutral term. To be irrational is to be out of control, but the supposed control of rationality is almost always praised. I have tried to show that being rational is not always praiseworthy, especially if it comes at the price of suppressing one’s emotions. Privileging rationality thus often betrays a wider view about human nature. As I discussed in chapter five, dichotomizing human traits, assigning them to a specific gender, and then ranking the traits is hierarchically is damaging to men and women, restricting them to a narrow range of “acceptable” behaviors (e.g. Kilbourne 2000).
72 The following sources address a range of topics in computing. Note that some of these are collections of essays and not all of the authors listed in this note are explicitly feminist; however, they all are critical of AI: Adam 1995, 1998; Albright 2000; Balka & Smith (Eds) 2000; Bush 1983; Clark 1997; Collins 1990; Craig et al 2000; Dautenhan 1997; Davies & Camp (2000); Donath 1997; Dreyfus 1979, 1992, 2000; Dreyfus, H. & Dreyfus, S. 1986); Eubanks 2000; Grundy et al (Eds) 1997; Haugeland (Ed) 2000; Kantrowitz 1994; Kuosa 2000; Perry & Greber 1990; Stack et al 1998; Tuana & Morgen (Eds) 2001; Turkle 1984, 1997; Turkle & Papert 1990; Wagner 1993; Webster 1995. 73 For example, see Tuana & Morgan (Eds) 2001 and Antony, L. & Witt, C. (Eds) 1993, for a wide range of general philosophical papers on rationality and Adam 1995, 1998 (especially chapter 4) for discussions specifically about AI. 163 Chapter Six: Embodied & Embedded Epistemology
In AI a particular type of program called an expert system (ES) is said to be rational, but only in a strictly formal sense: it usually meets the very limited goals set for it. It seems odd though to praise a machine agent, who is merely following the rules, as intelligent. Following the rules can be done in a very narrow scope without any intelligence. Circuit breakers in this sense follow the rules, and they clearly are not rational in any broader sense of the term, a sense we use in everyday language to refer to something like intelligence.
The formal sense of rationality, then, is only minimally useful for actually developing artificial intelligence, because that term implies a much broader set of skills. The classic definition of intelligence used in AI comes from the Turing Test, whereby an agent is intelligent if it can learn, if it can reason on its own, if it can communicate, and if it can store information that it attains.74 It is incoherent to claim that an ES is intelligent in the broader, richer, Turing
Test sense. This is especially true given that expert systems are modeled upon human expertise that itself is rarely rule-based, much less rational. Rationality then is necessary in some sense, but is not sufficient for, intelligence. For example, a person could be irrational and still very intelligent, in that she might know a great deal. Yet, irrationality in the common sense of being crazy or mentally ill, seems to disqualify persons as fully intelligent insofar as they often fail to communicate effectively.
Furthermore, the quest for artificial intelligence, understood as fundamentally anchored to rationality, diverts attention away from other components of intelligence, namely embodiment.
John Haugeland makes a similar claim from a different angle: “Such an architecture [which is only capable of handling simple instructions] is implicitly assumed by much of philosophy and most of AI” (1998, 227). Such an assumption has no need for considerations arising from an understanding of intelligence which draws on not just embodiment, but, as Haugeland argues, the
74 See Russell & Norvig 1995, 5-6. 164 Chapter Six: Embodied & Embedded Epistemology surrounding context, i.e., the world, about which I elaborate below. Only in the ‘updated’ “total
Turing Test” is the ability to move considered important; in the original test there was no such consideration of movement, much less anything else embodied. As the authors of Artificial
Intelligence: A Modern Approach note, “within AI, there has not been a big effort to try to pass the Turing test” (Russell & Norvig 1995, 6). The focus on rationality seems to be fallout from a history of dualisms with respect to human traits, where masculinity is connected to rationality and to superiority over that which is feminine, emotional, and embodied.
ARTIFICIAL INTELLIGENCE
A basic dictionary defines intelligence as: "The capacity to acquire and apply knowledge". Most of the related words explicitly refer to the mental. In the essay titled "Mind
Embodied and Embedded" from his 1998 Having Thought, John Haugeland discusses intelligence and the cognitive (Haugeland 1998b). How you define the problem, of course, dictates to some extent what you will find for your answer. Stipulating that intelligence is part of the mental automatically precludes it from being embodied--at least as long as you conceive of the mental as a separate entity. Haugeland wants to deconstruct the view of mind as disembodied: "There is little reason to believe that symbol processing has much to do with it-- unless one is already committed to the view that reasoning must underlie all flexible competence" (1998b, 221). He is saying that the focus on symbolic processing was a result of a view that reasoning is the only important component of intelligence, and intelligence is what allows us to successfully navigate in the world. His own view is that in numerous cases we are easily and skillfully able to do things with our bodies in our immediate environments; thus, it
165 Chapter Six: Embodied & Embedded Epistemology seems unlikely that symbol processing could be at the bottom of such fluid and successful manipulations.
Symbol processing is taken to be representative of the purely mental, the abstract, and the disembodied. Haugeland thinks that the Cartesian dualistic worldview is deficient and that an appropriate alternative is one that sees body, mind, and world as continuous and integrated. To reiterate, for him the connection between body, mind, and world is so tight it is intimate. They are each crucial to getting the job of cognition done: "that would mean that the 'content' of any given neural output pattern would depend not only on the particular body that it's connected to, but also on the concrete details of its current worldly situation" (1998b, 226). Intelligence is not just in the mind or just in the body or just in the world. It lies at the intersection of mind, body, and world. This point about including the world in our concept of intelligence will be revisited below.
It is true that representation--the mainstay of traditional philosophical theories--is helpful because it allows us to think about things both absent and distant by standing in for or holding the place of the distant thing. This point was made in chapter one during the discussion of abstraction. The problem is that there are other ways to have meaning and those alternatives have been ignored. For example, things in their social contexts are meaningful. George Lakoff and Mark Johnson (2000) wrote an entire book, Philosophy in the Flesh: The embodied mind and its challenge to Western thought, in service of the claim that meaning of concepts is fundamentally and unavoidably rooted in our embodied experience. Haugeland defines
"meaningful" as: "significant in terms of something beyond itself" which is "subject to normative evaluation according to that significance" (1998b, 232). Intelligence cannot be
166 Chapter Six: Embodied & Embedded Epistemology isolated—it is not located just in the mind apart from the body. Haugeland provides this illustration of his view:
Even in so self-conscious a domain as a scientific laboratory, whether research or development oriented, much of the intelligent ability to investigate, distinguish, and manipulate natural phenomena is embodied in the specialized instrumentation, the manual and perceptual skills required to use and maintain it, and the general laboratory ethos of cleanliness, deliberation, and record keeping. Without these, science would be impossible; they are integral to it (Haugeland 1998b, 236).
Supported by Haugeland, I am claiming that the intelligence under investigation here cannot be purely rational, but is necessarily embodied. But that is not how computing science has generally understood intelligence.
ARTIFICIAL INTELLIGENCE
Computer scientist Aaron Sloman calls artificial intelligence (AI) the science of knowledge or the science of intelligence (2001, 2). In AI rationality is taken to be the "ideal concept of intelligence" (Russell & Norvig 1995, 4, emphasis in original). Some if not much AI tends to view the discipline as one that strives to mimic human reasoning or human intelligence, both of which are primarily understood to be mental phenomena. While research on situated AI is occasionally mentioned, agency still seems located only within the agent, which acts on or in its environment, an environment that is separate from the agent. Researchers spend much energy on knowledge representation: what and how to represent knowledge.
According to philosopher Margaret Boden, a long-time contributor to the philosophy of artificial intelligence, representation figures heavily in the two main branches of AI: connectionist and classical AI (Boden 1995). Classical AI is defined by symbol manipulation, mentioned above by Haugeland. Take information, encode it, and let the computer analyze it.
167 Chapter Six: Embodied & Embedded Epistemology
This approach has realized much success. Its strength lies in the ability of the programmer to clearly define the problem and create a step-by-step plan for solving the problem. One example of this picture of AI research comes from the text, Artificial Intelligence: A Modern Approach.75
The conception of the mind as a central processor seems to call for structuring research in certain ways. Philosopher Andy Clark claims that the central processor mindset has led to an overemphasis on computation in AI. He has a specific diagnosis of the problem: "Once mind is cast as a controller of bodily action. . . . The distinction between perception and cognition, the idea of executive control centers in the brain, and a widespread vision of rationality itself are all called into question" (Clark 1997, 8, emphasis added). For Haugeland, the picture of agents with multiple, rich channels of feedback between mind, body, and world, contrasts with the standard picture of simple, narrow-bandwidth instructions, running from the central processor to all of the components. The instructions can be easily copied and supposedly can be used to instruct any set of interchangeable pieces of equipment, such as robot fingers. To repeat from above "Such an architecture is implicitly assumed by much of philosophy and most of AI" (Haugeland 1998b,
227). But, as I, together with Clark and Hagueland, have been arguing, this standard view is insufficient.
Connectionist AI is more mysterious, and less explicitly defined. These systems are more flexible, adapting more readily to new information. Connectionist systems consist of more or less elaborate networks of units. The units can be active or not. They signal the units connected to them. Information enters the network and activates it, causing the units to adjust their sensitivity depending on the reactions of neighbors. The strength of connectionist AI is pattern recognition, which is a holistic rather than stepwise approach to problem solving.
75 The text was used in the AI courses I attended at the University of Cincinnati and is also one recommended on the American Association for Artificial Intelligence (AAAI) web page: www.aaai.org . 168 Chapter Six: Embodied & Embedded Epistemology
With connectionist (or “neural”) architectures, we see redundant, flexible structures with distributed responsibility. Higher centers do not control or store all of the information; they just know how to retrieve it when necessary. There is no central representation on this model: "Even where adult cognition looks highly logical and propositional, it is actually relying on resources
(such as metaphors of force, action, and motion) developed in real-time activity and based on bodily ability" (Clark 1997, 154). Moreover, representation can vary; it does not need to be consistent across different areas. The information contained in representation can be transmitted in other ways, or by representation at the local level. One way to argue against strict representationalism is to appeal to "the presence of continuous, mutually modulatory influences linking brain, body, and world" (Clark 1997, 163). Examples of mutually modulatory dynamics are found in jazz music, group conversation, team sport, and partner dancing. The component
“parts” are constantly adjusting to each other to get the “job” done. They work out what needs to happen in real time without waiting for orders from something outside the system. In widely connected webs where components are not so easily isolated, representation is much less useful.
Boden characterizes one tension in AI as the question of "whether the Cartesian presuppositions of (most) AI should be replaced by a neo-Heideggerian approach" (Boden 1995,
96). "Situated robotics" seems closely akin to the kinds of projects championed by Clark and
Haugeland, and is much less enamored with representation (Boden 1995, 97). She writes:
Heideggerian critiques of AI aren't new. But they're now being mounted by people sympathetic. . . to situated and evolutionary robotics and to A-Life's study of 'animats.' These people see organisms as dynamic systems closely coupled with their environment. Instead of positing internal representations of an objective external world, they speak of whole systems embedded in, and adapted to, their own particular 'worlds' (1995, 99).
Recent sources in the AI literature do lean towards the Heideggerian. In one 2000 edition of the journal, IEEE Intelligent Systems & their applications, the editors surveyed a prestigious group
169 Chapter Six: Embodied & Embedded Epistemology of influential AI scholars and scientists to find out what the new research trends were. From
MIT Randall Davis and Howard Shrobe commented on the state of work in representation:
Another way to come at this [determining appropriate modalities of representation] is to say that perhaps vision (and sensory modalities) ought not to be treated as input channels whose information must be condensed and turned into textual symbolic representations before intelligent processing can occur (Hearst & Hirsh 2000, emphasis added).
It is becoming more acceptable to view intelligence as embodied and as located in the world.
Additionally, I will stress, that intelligence cannot be understood as existing in individuals isolated from their social communities. Let me elaborate on this claim by drawing more deeply from the work of philosopher Andy Clark
DRAFTING MODELS OF COGNITION
a good deal of actual thinking involves loops and circuits that run outside the head and through the local environment. Extended intellectual arguments and theses are almost always the products of brains acting in concert with multiple external resources. These resources enable us to pursue manipulations and juxtapositions of ideas and data that would quickly baffle the un-augmented brain. In all these cases, the real physical environment of printed words and symbols allows us to search, store, sequence, and reorganize data in ways alien to the onboard repertoire of the biological brain . . . . (Clark 1997, 207, citations omitted).
Andy Clark claims that persons are fundamentally and deeply interconnected bodies, minds, and environments. He says that we are embedded and embodied, to the extent that our thinking often requires the use of physical tools such as pen and paper. “Embedded” denotes interconnection with others, as opposed to existing as isolated individuals. We are tied to others by virtue of, for example, the use of language.76 Memberships in a nation or society, work and living communities, a culture, or even extended family all represent embedded relations. Each of
76 Code 1993; Lakoff & Johnson 1999; Longino 1990; Polanyi 1964. 170 Chapter Six: Embodied & Embedded Epistemology us is embedded in multiple communities, a point I have made above in chapter two. For Clark,
“embeddedness” describes a relationship with the physical environment, as illustrated above. As
Clark emphasizes, the nature of physical being is a state of embeddedness.
“Embodiment” is shorthand for a variety of ways a body, i.e., a person, fits into the world. It means just what it sounds like: real people have bodies, not just brains. It is the role the body plays in cognition that is the key point here. This is the same point raised as early as the 1930s and ‘40s by Michael Polanyi, Martin Heidegger, and Maurice Merleau-Ponty, albeit under different motivations. I discuss their contributions below. More contemporary theorists, such as Clark, have developed ideas of embodiment in different ways than those earlier theorists.
I find Andy Clark’s argument about embodiment and embeddedness, as detailed in his
Being There: Putting Brain, Body, and World Together Again, compelling for at least three reasons. The first is that the theory fits with current empirical research from psychology, neuroscience, and computer science. Clark, as well as John Haugeland (2000; 1998), and the philosophers Patricia and Paul Churchland (1998), for example, reference and build upon this kind of research. Haugeland (2000) deems it important to consider empirical research. Mind
Design II, a collection of essays he edited, are almost all “’scientific’ in that they are technically sophisticated and concerned with the achievements and challenges of concrete empirical research” (p. 1) although later in his introduction he admits that “scientific research into the kinds of systems that might achieve intelligence in this way—embodied and embedded mind design—is still in an early phase” (p. 27). The Churchland’s work is also heavily grounded in modern scientific research. For example, they state that “the neurosciences have revealed much about the brain at a wide variety of structural levels” and that “only research can decide how closely an artificial system must mimic the biological one to be capable of intelligence”
171 Chapter Six: Embodied & Embedded Epistemology
(Churchland and Churchland, 1998, 56-7). It seems irresponsible to ignore what contemporary scientific research has to say about cognition; a naturalistic project would not do so.
Secondly, Clark's argument is compelling because it resonates personally. Clark could be describing me: an embodied, embedded Ph.D. candidate immersed in arguments, texts, quotes, and commentary, all strewn across scraps of paper, legal pads, and computer files, some of which are taped to office shelves and backs of doors so that I can see and track the relationships of the subsections of my own argument. I need all of these devices and others—“external scaffolding” or “wideware”—to make better sense of what I read and write.
Clark’s view is not necessarily as Western culture-bound as it might first appear, either.
Cultures that do not rely as much or at all on written material still use external scaffolding, including other people, to remember and process information beyond what any individual herself could do. Even in our highly writing-intensive culture, I need other people to help me write—to help me clarify, expand, and reformulate what knowledge I do have. Recognition of the role of such social relationships in cognition is therefore needed for a complete picture of intelligence.
A fuller account of our social embeddedness is highly compatible with Clark's position. For example, Clark describes a “situated” brain, which is “a brain at home in its proper bodily, cultural, and environmental niche" (Clark 2001, 257, underscore added). Thus a third reason to employ Clark's theory is that it is compatible with a social epistemology.
Let me elaborate on this notion of “scaffolding” and how it is social. Clark uses the label
“scaffolding” to describe a tremendous range of tools including language, elements of culture
(rituals and myths), computers, pens and paper, and compasses. These tools help by offloading memory and by serving as reminders. Relying on external support, scaffolded action constitutes a "broad class of physical, cognitive, and social augmentations—augmentations that allow us to
172 Chapter Six: Embodied & Embedded Epistemology achieve some goal that would otherwise be beyond us" (Clark 1997, 194-5). These tools allow us to plan using an abstract, general view, in order to temporarily function above certain details.
Language is one of our most valuable tools. It helps us communicate ideas, but it is also a tool "that alters the nature of the computational tasks involved in various kinds of problem solving" (Clark 1997, 193, emphasis added). Language molds us and in turn we change it.
Importantly, "it also enables us to reshape a variety of difficult but important tasks into formats better suited to the basic computational capacities of the human brain" (Clark 1997, 193). This is significant because, Clark argues, it lets our brains "tackle otherwise intractable classes of cognitive problems" (p. 194). If we can identify useful processes or properties in the environment we can exploit them to our own ends; we can make possible that which otherwise remains intractable.
Another example of a tool is “coaching speech,” directed either towards one's self or towards another. Such language,
functions so as to guide behavior, to focus attention, and to guard against common errors. In such cases, the role of language is to guide and shape our own behavior—it is a tool for structuring and controlling action, not merely a medium of information transfer between agents (Clark 1997, 195).
Self-directed speech in particular is "a crucial cognitive tool" because it can help us focus on the most important tasks, and figure out how to solve whatever problems it presents us, by using whatever means might be available. It can help get the primary task done without distraction from, or in some cases, overwhelmed by, other concerns. One coaches oneself through a difficult situation, putting other problems on “the back burner” in order to concentrate enough energy for the difficult task. Coaching speech helps us alter ourselves by changing our thinking or the environment, and that helps us solve problems. This strategy works in part because "by
173 Chapter Six: Embodied & Embedded Epistemology using real external representations we put ourselves in a position to use our basic perceptual and motor skills to separate problems into parts and to attend to a series of subproblems, storing intermediate results along the way . . . " (Clark 1997, 199, emphasis added). Again, external devices help us where we lack natural ability; the brain is "empowered. . . by the availability of a real-world arena that allows us to exploit other agents, to actively seek useful inputs, to transform our computational tasks, and to offload acquired knowledge into the world" (Clark
1997, 87, emphasis added).
It is easy to see how a social epistemology makes itself at home with Clark’s position. It is only in a social setting that knowledge grows. For example, my critics remember objections that I would sooner forget, but together we "evolve" a stronger position, when they require elaboration of some point I have made. This is a particularly relevant practice in science.
Growth of knowledge cannot be done alone, in a solitary brain. That is why the notion of equipoise, discussed in chapter four, is so important. Thus, evaluating the practices of communities is necessary, and that includes an assessment of the role of gender in those practices.
MICHAEL, MARTIN, & MAURICE
A few words of background are in order regarding the work of Michael Polanyi, Martin
Heidegger, and Maurice Merleau-Ponty. Heidegger's contribution begins with his concept of
Dasein or "being there," by which he means to indicate that humans are active participants in the world. Reacting against his predecessors, who championed a detachment in the world and privileged the “objective” and detached methods of science, Heidegger pursued questions about the nature of Being, specifically about ontology. Humans exist, he understood, in webs of
174 Chapter Six: Embodied & Embedded Epistemology meaning. For instance, we are bonded to the tools that we use in way more profound than is commonly acknowledged. How we use a tool reveals a web of relationships, including: what I use the tool for, how that activity fits in with my other activities and plans, and how all of that interacts with the various ways other people exist, act, and live in the world. One concern of his involves the artificial separation of scientists from their subject matter and the ensuing diminishment of what it means to be a human situated in the world. Thus Heidegger’s work provides partial grounding for this current discussion of embeddedness.77 A terrific contemporary example of Heidegger’s emphasis on a scientist’s active participation is Evelyn
Fox Keller’s study of the research of Barbara McClintock. McClintock spent years on extremely detailed and precise examinations of corn, work that at first was dismissed as idiosyncratic and too “touchy-feely”, but later earned her a Noble prize in genetics. Understanding her methods and her successes, requires understanding how her work fits into many webs of meaning.
This chapter uses Heidegger’s idea that people actively participate in the world, which in some cases involves the use of tools, then develops an even broader conception of person that incorporates various artifacts of technology into the very notion of person. In other words, the notion of person is expanded out into the world of things. Tools themselves become an extension of the person, such that the boundaries between person and tool no longer need as sharp a demarcation. Examples that force us to reconceptualize the notion of person include things like pacemakers and genetic therapies. Another example is nanotechnology, which promises the ability to insert extremely small censors and regulators into our bodies to help us
77 See Gelven (1970); Grene (1957); Inwood (1995); Krell (1977). 175 Chapter Six: Embodied & Embedded Epistemology stay healthy. One such nanotechnology device under research uses real-time sampling of blood and medication for regulating blood sugar or insulin levels.78
My incorporation of technology and persons, differs than Heidegger’s point since
Heidegger had a “distaste for technology” (Inwood, 346). As I understand it, and to put it very roughly, Heidegger wanted to turn away from—or at the least was highly cautious of— technology because he thought that it puts people in the wrong relationship with the world.79
Technology according to this view encourages people to see themselves as the center and most important part of the world: “Within and beyond the looming presence of modern technology there dawns the possibility of a fuller relationship between man and Being—and hence between man and all that is—than there has ever been” (Lovitt, xxxvii). In contrast, I am cautiously embracing technology, seen in more neutral terms, which is similar to the approach of both Clark and Haugeland.
Embodiment and embeddedness are also stressed in the essays of Michael Polanyi.
Polanyi’s concept of “tacit” knowledge is explicitly embodied. Tacit knowledge underlies scientific knowledge because it is fundamental to all knowledge: “When we make a thing function as the proximal term of tacit knowing, we incorporate it in our body—or extend our body to include it—so that we come to dwell in it” (Polanyi 1967, 16). Polyani, a trained chemist who later wrote as a philosopher of science, viewed tools as extensions of the body, like
Heidegger. Not only does knowledge arise out of one’s body, not everything known can be stated explicitly. Tools used in science—theories and instruments—are not independent of their users. Writes Polanyi: “To rely on a theory for understanding nature is to interiorize it. For we
78 Abraham P. Lee “From BioMems to NanoMedicine: The Merging of Diagnostics and Treatment" talk given at the Human Genome Odyssey Conference, April 2001, University of Akron. See http://www3.uakron.edu/2001/ for more information. 79 Glazebrook 2000, 207-24; Lovitt 1977, xxx-xxxiii. 176 Chapter Six: Embodied & Embedded Epistemology are attending from the theory to things seen in its light, and are aware of the theory, while thus using it, in terms of the spectacle that it serves to explain (1967, 17). Notably, Polanyi does not conclude that embodied knowledge damages objectivity. In this way, Polanyi’s view harmonizes with work in contemporary feminist philosophy.
Polanyi stresses that scientists must participate personally in their work and that science must be free to pursue its own questions and define its own problems. According to Polanyi, we do not understand how scientific discoveries come about. In other words, the context of discovering scientific problems and solutions is wide-open, without formal rules for determining which problems or solutions are legitimate and which are not. Finding a good problem, having insight to find possible solutions to those problems, and possessing the ability to recognize solutions to a problem are all forms of art rather than an exact science. Unlike more positivist- leaning theorists who want to dismiss emotional attachment to science as contaminating, Polanyi believes scientists are passionate towards, as well as critical of, their work. Indeed, this is why science works. This is precisely what Evelyn Fox Keller (1983) showed in her examination of
Barbara McClintock’s work. Alison Jaggar argued in “The Myth of the Dispassionate
Investigation” (1989) that emotional connection to one’s work is absolutely crucial for continued research in science, for otherwise the tediousness of the work would turn researchers away.
Nancy Tuana (1996) also champions the family doctor—the general practitioner—as a superb model for scientists, precisely because an emotional connection to a patient (and by analogy to one’s research subject, whomever or whatever that may be) uncovers symptoms, or connections between symptoms, events, and lifestyle, that otherwise might go unnoticed. The passion in combination with the critical concern protects objectivity. I made this argument earlier in chapter two. Polanyi writes,
177 Chapter Six: Embodied & Embedded Epistemology
to some degree we shape all knowledge in the way we know it. This appears to leave knowledge open to the whims of the observer. But the pursuit of science has shown us how even in the shaping of his own anticipations the knower is controlled by impersonal requirements. His acts are personal judgments exercised responsibly (1967, 77).
I take this to mean that individual scientists must act with integrity and commitment. The point is about a scientist’s intent to search for answers, not about certainty. For Polanyi, it also definitely involves thrill, passion, and imagination.
My previous discussion of competition is also relevant here. I take Polanyi’s position to be closer to mine than to David Hull’s. For example, I do not think the mechanism for advancement in science is Hull’s adversarial competition, but rather something more akin to what Polanyi refers to as “the principle of mutual control” whereby “each scientist is both subject to criticism by all others and encouraged by their appreciation of him” (1967, 72). The scientists regulate themselves, not by competing for prizes (money, awards, recognition, etc.) but by assessing each other. I assess those who work “around me” and they check their own neighbors. Together these “chains of overlapping neighborhoods” form the basis for “mediated consensus” (Polanyi 1967, 72, 73). Power and control is spread out across the community, thus ensuring a certain objectivity out of subjectivities.
In many ways Polanyi’s work turns out to be a somewhat surprising precursor to much that appears in feminist science scholarship. It is surprising in part because he is a chemist discussing chemistry and physics, which are often taken to be the least subjective sciences. It is also surprising because he worked during the heyday of logical positivism, an intellectual movement notorious for wanting to keep anything remotely subjective out of science. But
Polanyi clearly accepts that an emotional component is at play in the work of science. He also
178 Chapter Six: Embodied & Embedded Epistemology clearly believed that authority in science was shared across the group of scientists and should not be directed towards any one goal, especially something state-sponsored. Repeatedly this comes out as a protest against state-planned science, and against the executions of non-compliant scientists in Soviet Russia in the 1930s. Because in a free environment all scientists have a shared stake in the outcomes of science, the authority is distributed across them and regulated by consensus. Given the huge amount of background necessary to master before becoming a scientist, including a specialized knowledge of scientific instruments, not everyone has access to science. Polanyi does not comment on whether or how outsiders might critique science, but I do not believe that he would reject the idea of outside review, given his own negative remarks about
Soviet science. Helen Longino’s work on the social nature of science (Longino 1990, 2002) is particularly compatible with this outlook of Polanyi’s.
Furthermore Polanyi rejects classic philosophical dualisms (mind-body; objective- subjective), which generate much criticism from many feminist scholars, due in part to the denigration of the body and the passions perpetuated by those dualisms. Polanyi embraces philosophical positions like the view that observation is theory-laden, even though he does curiously state that this is more likely to be a concern in the biological rather than the physical sciences: “observation, strictly free from valuation, is possible only in the sciences of inanimate nature” (1967, 51). Tacit knowing underlies all observation scientific or otherwise and in that theories are like tools: “To rely on a theory for understanding nature is to interiorize it” (1967,
17). These are some of the very same points scrutinized by feminist philosophers of science.
The work I am discussing here blurs the lines between subjectivity and objectivity in science. Together it all makes room for work in social epistemology, as a critique of social practices in science, such as my examination of competition in science from chapter five.
179 Chapter Six: Embodied & Embedded Epistemology
Furthermore much in Polanyi’s position grounds my overarching argument regarding the importance of feminist philosophy of science in computing and AI. Feminist philosophy of science provides a rigorous assessment of computing with a particular interest in concepts of intelligence. This interest, harkening back to Heidegger, leads to more basic questions about persons, which inform the creation of artificial intelligence and other sophisticated computer programs, and which involves an assessment of how people benefit or are harmed by such ideas and projects. Even though Polanyi does not expressly address feminist concerns, his outlook seems sympathetic to those concerns. I will therefore build his views into my comprehensive thesis.
There is one more intellectual lineage concerning embodiment to examine before returning to the contemporary discussions of Clark and Haugeland: that of Maurice Merleau-
Ponty. Polanyi himself references Merleau-Ponty (as well as Thomas Kuhn and N. R. Hanson) in his 1964 revision of Science, Faith and Society, and there are many interconnections between them. According to Merleau-Ponty the “body-subject” is neither just physical nor devoid of spirit. It is the body as a whole that gives meaning to what we perceive. Perception is pre- conscious and pre-personal (Kwant 1963, 11). That is, the mind does not fully control perception, although reasoning can be somewhat involved. Those perceptions cannot be further reduced: “For Merleau-Ponty the realities appearing to use are already a structure, a form, a
Gestalt” (Kwant 1963, 3). For Merleau-Ponty, making meaning relies on perception and since that is fundamentally tied to embodiment, meaning ultimately arises from embodiment (Dreyfus
& Dreyfus 1964, ix-xxvii). Interpreting Merleau-Ponty strictly one could conclude, as does
Hubert Dreyfus (1979) that without embodiment there cannot be perception so there can be no meaning-making, i.e., no intelligence.
180 Chapter Six: Embodied & Embedded Epistemology
Regarding meaning, people must work to create it in the world, according to Merleau-
Ponty, since for him (and other existentialists, such as Sartre and Kierkegaard) it is not given to us by God, for example. There is no single interpretation of the meaning of any given situation.
Here lies another point sympathetic to feminist philosophy of science regarding the necessity of a plurality of views in the sciences. Sandra Harding (1998) makes very similar claims about the nature of knowledge, thereby underscoring the need for support of diversity of views within science so as to have the resources to uncover more than one limited aspect of reality. Donna
Haraway writes about vision as a metaphor for knowledge, because vision “can be good for avoiding binary oppositions” (1997, 283). Her argument is very much grounded in the embodiment of knowers. For Haraway, all vision is embodied, which makes all vision limited.
Analogously, all knowledge, being embodied, is limited. No one of us has perfect vision; no one of us has perfect knowledge. We need each other to develop a full account of our knowledge.
Contemporary feminist use the phrase “situated knowledges” or “situated knower” to stress that perspectival difference is real, and not at all trivial when it comes to epistemology. We are each situated differently and none of us has a God’s eye view. Thus, Haraway requests “a doctrine of embodied objectivity that accommodates paradoxical and critical feminist science projects:
Feminist objectivity means quite simply situated knowledges” (1997, 284).
One feminist concern—taking the body seriously by refusing to overlook the embodied experience—is thus addressed here by incorporating the work of Polanyi, Heidegger, and
Merleau-Ponty with that of Clark and Haugeland. Reconceptualizing rationality, intelligence, and cognition is a goal of this chapter. This reconceptualizing led to the positing of a social account of epistemology. In turn, the social account can be understood as an aspect of what it means for intelligence to be embedded. I argue that a conception of intelligence steeped in an
181 Chapter Six: Embodied & Embedded Epistemology embodied and embedded epistemology makes more sense and promises more success than alternative conceptual frameworks for constructing artificial intelligence. Intelligence, to conclude, in necessarily connected to the outside world, both the physical environment and the human social community. Research that tries to incorporate the environment beckons, whereas projects which fail to account for either embeddedness or embodiment as dimensions of intelligence cannot fully capture what they are supposed to. Additionally, a social epistemology provides a particularly satisfying a way to explain, to give a specific example, the success of real experts and the ramifications of such an incorporation on real expert systems. A social account gives a better explanation than a more traditional account, because the social view can locate assumptions that are hidden but necessary to the experts’ functioning.
This ties into another feminist concern regarding how to make theory work in the world.
It is not always easy and the implementation itself can raise new complications. I think the important point here is to try the ideas and refine them with real feedback. This approach is in keeping with the naturalistic view and generally with the scientific method of putting forth a hypothesis for testing and subsequent refinement. Additionally, while one might understand and agree with certain theoretical points arising from work in feminist philosophy of science, that does not necessarily mean that one could apply the theory—put it to work in a research project, for instance—without help.
CONCRETE FEMINISM
In her 2001 “Dismantling the Self/Other dichotomy in science: Towards a feminist model of the immune system” Lisa Weasel, a biologist working in immunology, charges that although
182 Chapter Six: Embodied & Embedded Epistemology there has been a fair amount of theoretical work in feminist science studies, not very much has been provided in terms of concrete examples of critical feminist science. She writes:
Despite the wealth of critical feminist writings on science, far fewer feminist approaches to the sciences have ventured to take on the challenge of articulating concrete examples of actual science that might be seen as consistent with feminist goals (Weasel 2001, 28).
That is her emphasis—finding real science that can serve as an exemplar for the instantiation of feminist values, goals, and epistemologies. Weasel's model comes from immunological research that challenges the standard and dominant self/other model. Her goal in offering this alternative is a very pragmatic one: "to demonstrate how feminist theory can provide scientists with creative resources for rethinking the kinds of questions that they ask and the answers that they find" (Weasel 2001, 29). It is important, claims Weasel, to provide models on the concrete level, not just in theory, given that few practicing scientists are familiar with the claims of feminist science scholars and probably would not know how to incorporate the criticisms anyway. If we are serious about wanting science to change, we must offer suggestions for how scientists might take such steps.
I hope to provide some material for creative reconceptualization of the practice of AI. I have been arguing that intelligence is based on cognition, which is embedded and embodied, not isolated and disembodied as Descartes’ and his intellectual successors would have us believe. In the next chapter I analyze a contemporary AI project, Cyc, making specific, concrete suggestions for improving the project and for proceeding in AI. To further live up to Weasel's challenge, though, I would need to review and assess other research, starting in robotics, from the current
AI research world.80 This chapter remains theoretical: I assess the actual practice of science
80 I took a step towards that goal by participating in the Grace Hopper Celebration of Women in Computing conference in October 2002, where I led a discussion on gender in AI work. See www.gracehopper.org . 183 Chapter Six: Embodied & Embedded Epistemology against the requirements of this embedded and embodied theory of cognition, and then against some more general tenets of feminist theory.
Beyond particular research projects or research programs are specific daily practices.
While providing models of research is good, providing models of practices is also productive.
For example, looking at laboratory organization and practices, along the lines of Bruno Latour and Steve Woolgar's 1979 Laboratory Life or Sharon Traweek's 1988 Beamtimes and Lifetimes, has revealed some assumptions about gender and/or embodied and embedded cognition, and allows us to ask if particular practices exploit, enhance, or hinder the capabilities of embodied, social knowers. If one thinks that social situatedness can have an impact on one's work, then the organization of a lab, for example, might make a difference.
The point I want to stress is this: if intelligence is seen as located exclusively in the mind, social structures that promote competition between those minds in order to produce the best ideas makes more sense than if intelligence is understood to be more fluid and dynamically connected among the mind, body, and world. The normative aspect of cognition at work here privileges rationality, to the exclusion of the body, and thus is free to ignore how social arrangements can impact upon cognition. But if it is true that we are embodied, social knowers, then environments that do not encourage a full use of our capabilities must be selling us—and science—short. As such they can be challenged on normative grounds. Furthermore, if we are embodied and embedded knowers, then any artificial reconstruction or creation of intelligence must take that into account.
Cyc, the project I will discuss in depth in the next chapter, is one of those endeavors that is failing (among other reasons) because it has been built upon the assumption that cognition is basically symbolic processing. Doug Lenat, the founding researcher at Cycorp, seems to think
184 Chapter Six: Embodied & Embedded Epistemology that “cerebral methodology” is not embodied, but is complex and intellectual (Lenat, Miller &
Yokoi 1995, 46; Moody 1999, 2). Lenat’s foundation therefore runs in the opposite direction of what I have been arguing. I describe the project in much more detail below and I also elaborate on a critique of Cyc, a culmination of all that has been presented in the above chapters.
Yet even if we could clearly define what constitutes a feminist computer science project, the question about whether a research project must be explicitly labeled or identified as feminist to be so remains unclear. I do think that people, for example, can be feminist and not ever explicitly identify themselves, and so can projects. Weasel's approach might at first seem a bit unconventional:
While some may believe that a feminist science must come from a consciously articulated feminist commitment or can only arise once we live in a feminist world, I tend to disagree. Although today's scientific laboratories and classrooms may not be filled with feminist scientists, it is up to those of use who dream of such a future to begin assembling the rough outlines of what such a vision could hold in store. There is no better place to begin our search for a feminist science than with scientific theories themselves (Weasel 2001, 41).
I agree with Weasel that a concrete research project to pursue or hypothesis to test is more promising than the merely abstract when one is trying to change the practice of science. By starting with projects that look feminist, even if they are not intended to be, we have something to build upon. For my purposes here this means canvassing for projects that acknowledge embodiment and embeddedness projects. At that point we can refine, retool, and rework them according to a variety of parameters. But at least we will have something to suggest to the scientists which approximates feminist research. This is similar to Clark's suggestion that initial data for robots must come from somewhere, and once we have it we can check it against real- world parameters (Clark 1997, 97). Let me sketch the outline of an area in which such a project might start, and illustrate in it why an embodied and embedded epistemology works better than a
185 Chapter Six: Embodied & Embedded Epistemology more narrowly rational one. The area is telepistemology, which applies to research in human- computer interaction (HCI). I suggest some of the work there can inform research in AI more broadly.
TELEPISTEMOLOGY: KNOWLEDGE AT A DISTANCE
An epistemology arising from the social can be vulnerable to breakdowns in communication, which can impede the successful development of knowledge, if the normal venues for verification are disturbed.81 Telepistemology is a small field of study focusing on issues in techno-communication. This is a new area of knowledge studies, one that investigates the stability and integrity of knowledge created and/or approached from a distance. Today this usually involves the Internet. Telepistemology tries to address issues of authenticity and authority as well as those of access and agency (Goldberg 2000, 3). Kenneth Goldberg in The robot in the garden: Telerobotics and telepistemology in the age of the internet, compiled a variety of essays to frame the study. To take one example, Hubert Dreyfus points out that one of the difficulties of relying on distanced knowledge is the reduction in information normally used for authentication. The feedback we get from in-person communication comes from and is processed by the body. Without these information channels, we cannot have much knowledge.
Some of the processing might be innate, but it might also be, as Clark and Haugeland urge, that our cognitive abilities are literally diminished when we are too far removed from the situation:
". . . we won't seem to be bodily present at the site in question because we won't sense ourselves as getting a maximal grip on the object of our concern" (Dreyfus 2000, 58). In other words, we
81 For example: looking into someone's eyes to see if she is telling the truth; watching body movements for signs of discomfort that might signal that something is wrong with what is being said; asking questions and getting real-time responses; looking for natural signs such as shadows to see if they align with the alleged position in time and space 186 Chapter Six: Embodied & Embedded Epistemology cannot function as well at a distance because we cannot get enough information, at least that it how it seems now.
Several scholars from the 1997 Women, Work and Computerization conference make related claims. B. Becker points out that miscommunication is easier in the absence of a body and worse, problems regarding veracity arise when nonverbal communication is missing. She writes: “we are in touch and touched by them [other speakers] through our body without being totally aware of what happens between us and what we know already about the other one before and besides speaking to him or her” (Becker 1997, 212). Her solution is to approach virtual communication as one would interpret literature, and to be aware that we each apply our own
“desires, wishes, and hopes in the fictional persons and literal heroes in the same way as we seem to do it in these virtual environments” (p. 212). Reinforcing the view that the body is fundamental to perception, communication, and intelligence, Kerstin Dautenhan (1997) states that the body is a point of reference for understanding, which is threatened in and by cyberspace.
Citing both Mark Johnson’s The Body in the Mind and empirical evidence, she writes: “and all what nature (and studies in natural sciences confirm this) tells us is that understanding cannot be separated from a ‘living body,’ an embodied mind” (Dautenhan 1997, 213). Her views are both embodied and embedded and she concludes that technology “can and should support this social networking aspect” (Dautenhan 1997, 214).
Judith Donath notes both positive and negative aspects of virtual communication. An advantage is that people feel free from “normal” prejudices and are able to interact with others in new and exciting ways. The disadvantages could bring about more harm on balance, though.
“Yet social cues are not simply vehicles for prejudice,” she writes, “they play an essential role in the formation of community and in our comprehension of social interactions” (Donath 199, 217).
187 Chapter Six: Embodied & Embedded Epistemology
Even more interesting is her observation about the construction of virtual worlds: “How social cues will evolve in these new worlds is in part a question of how the worlds are designed” (p.
217). The concern to include the body in telecommunications echoes Haugeland’s urge to develop a conception of intelligence integrated with the body. It previews how the ideas structuring Cyc limit it. Yet, with some careful additions, one of the problems resulting from
Cyc’s conceptual basis, might be mitigated. I give an extended analysis of my worries about Cyc in the next chapter.
KNOWLEDGE RIGHT HERE
Andy Clark partially anchors his position about “how thought itself is materially possible” in evolutionary terms: brains evolved to help make bodies move. Accepting his view is difficult due to the strength of ruling metaphors about our own selves, particularly our minds as rulers of our bodies. From an evolutionary perspective, however, central planning systems are
"revealed as slow, expensive, hard-to-maintain luxuries--top-end purchases that cost-conscious nature will generally strive to avoid" (Clark 1997, 51). Alternatively, physically connected
(embodied) body and mind, and that unity embedded in its environment make more sense:
But real-time, real-world success is no respecter of this neat tripartite division of labor [among perception, cognition, and intentional action]. Instead, perception is itself tangled up with specific possibilities of action--so tangled up, in fact, that the job of central cognition often ceases to exist. The internal representations the mind uses to guide actions may thus be best understood as action-and-context-specific control structures rather than as passive recapitulations of external reality. (Clark 1997, 51).
A "rational" reconstitution of cognition, which would be needed to model a system that was run by a central planner, in contrast, overlooks the obvious benefits of finding workable solutions in
188 Chapter Six: Embodied & Embedded Epistemology the real world, what Clark calls "opportunistic strategies." Such strategies exploit features of the world as useful shortcuts to creating a new solution from scratch.
Current work in genetics strongly suggests that evolutionary forces favor wholesale substitution of workable modules (of DNA) rather than completely novel solutions created from the ground up (Shapiro, UC Colloquium, May 2001). Lakoff and Johnson also make evolution- based claims. For example, they offer an evolutionary history for reason, as evidenced by structures common to all animals, including humans. Structures that we use in reasoning are used for perception and movement in animals. More to the point, supporting the claim that it is cheaper in evolutionary time to make do with what is already there rather than reinvent the wheel
(or neural path or whatever):
Brains tend to optimize on the basis of what they already have, to add only what is necessary. Over the course of evolution, newer parts of the brain have built on, taken input from, and used older parts of the brain. Is it really plausible that, if the sensorimotor system can be put to work in the service of reason, the brain would build a whole new system to duplicate what it could do already? (Lakoff and Johnson 2000, 43).
Clark reiterates this process at the level of (embodied) cognition: "new cognitive garments seldom are made of whole cloth; usually they comprise hastily tailored amendments to old structures and strategies" (1997, 81). He urges the use of "tinkering" instead of
“engineering” when thinking about our evolution and posits "evolutionary holism" as the proper way to think about the development of intelligence in which "complex wholes will usually be developed incrementally over evolutionary time, and that the various intermediate forms must themselves be whole, robust systems capable of survival and reproduction" (Clark 1997, 88). It is another small step to the conclusion that human society works the same way; we literally do not reinvent the wheel if one of us already possesses such knowledge.
189 Chapter Six: Embodied & Embedded Epistemology
If intelligence really works under conditions of embodiment and embeddedness, then artificial intelligence as a discipline can only be successful if it abandons the old way of thinking about thinking (i.e., intelligence). An AI project modeling a central processor type of intelligence might not ever get beyond the abilities of a bug--if it is that lucky. It will fail completely if, as Clark surmises, "central cognition often ceases to exist". Yet even if central processing were to fail we won't fall apart or cease to exist because something else will take on the jobs that we, in our reconstructive modeling, frequently assign to the central processor.
Some AI projects will surely have some fantastic abilities--perhaps in graphics interfacing or computation--but those abilities only scratch the surface of an intelligent agent.
To preview an argument below, a focus on rational agents may be a poor use of resources since there is enough evidence, some of which I have shown, to question the supremacy of rationality as a human characteristic. In addition, demoting rationality is justified according to feminists since historically there has been great disparity in the assessment of rationality of persons, far too often made along the lines of gender, race, and/or class. One of Clark's claims is that abstracting problems from their environment actually makes the solution too difficult to find because it is in that real-time, embedded situation that answers to problems arise naturally and simply, using what is literally on-hand, instead of 'thinking' of solutions and then trying to find some way to instantiate them. Connections to the actual world allow agents to simplify themselves.
EXTENDING THOUGHTS
One thing I have learned from feminist philosophy (and feminists generally) is to embrace embodied experience. This means paying attention to them everywhere; in this case I
190 Chapter Six: Embodied & Embedded Epistemology am reviewing computing and AI. Although some critics, such as Hubert Dreyfus, seem to reject the possibility of computers becoming “intelligent,” I think he is ultimately mistaken about what intelligence is, who has it, and what the consequences of those situations are. Dreyfus seems to believe that AI will not work if it does not begin with a body, since, following Merleau-Ponty and Heidegger, perceptions begin within the body and build on top of it. In contrast, I think that pursuing research in AI without regard for embodiment and embeddedness is going to be harder to justify, given the evidence at hand about intelligence to the contrary.
Are there risks to pursuing the path of embodied and embedded intelligence? For example, is personal identity endangered by such projects? Clark predicts such a concern:
If cognitive and computational processes are busily criss-crossing the boundaries of skin and skull, does that imply some correlative leakage of personal identity into local environment? Less mysteriously, does it imply that the individual brain and the individual organism are not proper objects of scientific study? (Clark 1997, 82).
I do not believe there is cause for concern. In fact an understanding of persons as bodies-minds- environments does not imply anything too disturbing, it just means that we need to step back and reconsider the unit of analysis. Researchers will need to study the aspects of the person--the brain and the body--in connection with environments. Conversely, researchers and theorists will need to consider the environment both as physical and socially organized units of knowers. To say that an agent is embedded means that knowers are embedded in their physical world, so we should look to the immediate physical environment for clues when researching cognition, but we should also look to the social unit(s) as well. There is extensive feminist scholarship on embodied knowing as well as the idea of being richly and deeply connected to one's environment
(political and physical); Clark’s position is highly compatible with those theories.
191 Chapter Six: Embodied & Embedded Epistemology
In fact, Clark is repeating claims feminists have been making for a while: our lives are rich, complicated, and messy. Maria Lugones (1996) is one of many examples of feminists who have written about multiple roles we all have and the fact that individuals belong to a large number of communities, some of which overlap, some of which contradict, and some of which seem wholly incompatible. Yet we all manage to exist and “travel” within these different worlds
(Lugones 1994) without totally losing “ourselves.” These are the social worlds in which epistemological agents are embedded. Clark recognizes that it is hard for anyone to keep track of all of these worlds, and that such circumstances are true for virtually everyone. We will have to make room for our knowers who live in multiple worlds and have multiple roles when trying to move from research in human to that in artificial intelligence. Clark writes, "Of course, the bad news about messier, more biologically realistic and interactive solutions is that they are not just hard to discover but also hard to understand once we have them" (Clark 1997, 92). More to the point are the hints about collective human cognition that directly links to the work of Helen
Longino, Lynn Hankinson Nelson, Sandra Harding and others:
We can now appreciate how such migrations [changes in the contours of the problem space] may allow the communal construction of extremely delicate and difficult intellectual trajectories and progressions. An idea that only Joe's prior experience could make available, but that can flourish only in the intellectual niche currently provided by the brain of Mary, can now realize its full potential by journeying between Joe and Mary as and when required (Clark 1997, 205-6).
Knowledge is everywhere. It is much more elusive yet more ubiquitous than we have noticed previously. This means that we necessarily need others to create knowledge—there is too much for anyone person to track, even about her own self, with respect to how intelligent beings exist intimately with the world, as well as with other persons, and within their own bodies. Does this carry more broadly into AI? I think that it does.
192 Chapter Six: Embodied & Embedded Epistemology
I have already argued that the body is inseparable from a person and thus from her intelligence. That, combined with the less than stellar performance of GOFAI, plus the recent successes of newer AI movements (e.g., Robotics) leads me to conclude that this is the avenue to pursue. It is better to embrace the new directions and incorporate careful thinking and appropriate review rather than let the technology develop in isolated labs. This is an extremely important point in our modern global capitalist economy—another dimension of our embeddedness—because the technology will very likely be pursued, regardless of the political, social, or moral consequences, if a profit can be made. Witness the interest in genetic technologies, including genetically modified food and organisms and human cloning, by some who are threatening/promising to go ahead with the research, regardless of the law, ethical concerns, or public moral outrage.
Research into “new” AI also seems to have the capacity to address some of my feminist concerns, especially those about how different lived experiences, with respect to bodies or social contexts impacts cognition. Taking those differences seriously seems more just and more realistic (thus fulfilling naturalistic concerns of above discussions) than other rationalistic models that assume the mind to be separate from the body, and model their artificial intelligence on that dualistic framework.
The Cyc project has been built upon the assumption that cognition is basically symbolic processing (Adam 1998; Lenat, Miller, & Yokoi 1995; Moody 1999). The project is failing, in part, due to this assumption. Regarding Cyc's symbolic orientation Andy Clark underscores "its extreme faith in the power of explicit symbolic representation" (1997, 2), but has little tolerance for claims that Cyc will be able to create artificial intelligence. In part this is due to its brittleness (its inability to react to its environment) and this leads Clark to conclude: "Thus, the
193 Chapter Six: Embodied & Embedded Epistemology
CYC project, taken as an attempt to create genuine intelligence and understanding in a machine, is absolutely, fundamentally, and fatally flawed" (Clark 1997, 4, emphasis added). This is because it does not begin from embeddedness.
Although Cyc is failing to live up to its original goals of creating intelligence (or, at the least constructing the entire framework of common sense knowledge), there are still many researchers working under the classical AI view, i.e., that intelligence will come about primarily through computation. I am calling for increased scrutiny of this view. Note that this does not mean that work in classical AI should be completely abandoned. Embodied AI seems promising, so at a minimum it would be good if we could explore new approaches individually or in concert with the old ones, resulting in projects which Margaret Bodin refers to as hybrids (1995, 97).
We don’t need to reinvent the wheel, we just need to look at how the wheel is just a part of the larger cart of embodied and embedded epistemology. In the next chapter I look at Cyc, showing its weakness with respect to my feminist philosophy of science framework, suggest ways the very same project might be improved to meet my feminist standards, and speculate on what further investigation into AI would be needed to draw even stronger conclusions about the value of feminist philosophy of science scholarship in the realm of computing and artificial intelligence.
Supported by and incorporating the work of Haugeland, Clark, Polanyi, Heidegger, and
Merleau-Ponty and others, I have claimed that intelligence cannot be purely rational, but is necessarily embodied. Reconceptualizing intelligence also involved the positing of a social account of epistemology. In turn, the social account developed into an understanding of intelligence as embedded. Such ideas of embodiment and embeddedness relate both to artificial intelligence and epistemology broadly. It is misguided to approach any AI project assuming the
194 Chapter Six: Embodied & Embedded Epistemology existence of a mind-body dualism, but in the past this is exactly how AI was conceived.
Problems were selected and solutions pursued all within the constraints of formal symbolic processing. Those projects succeeded, but somewhat limitedly and on a much smaller scale than originally predicted. AI has not developed nearly as fast or moved as far as its original creators boasted. The apparent success comes from the assumption that rationality is the only component of intelligence that requires modeling. AI has suffered, I argue, because it has been laboring under the assumption of a dualism between rationality and embodiment with regard to intelligence.
TURNING TO THE FUTURE
I have shown how work in AI, epistemology, and feminist philosophy of science overlaps. Connecting at the intersections of embodied and embedded epistemologies, this work is at the forefront of contemporary research in each of these areas. For instance, in August 2001 the International Joint Conference on Artificial Intelligence (IJCAI) invited Manuela Veloso of
Carnegie Mellon University to speak on her research on teams of intelligent software agents, based on multiagent and multirobot systems and their ability to learn.82 Notice that robotics itself is a highly situated (a.k.a. embedded and embodied) discipline, and that networks of agents are clearly an instantiation of embedded knowing. Moreover, as if Clark, Haugeland, Lakoff and
Johnson, or the Churchlands themselves were directing the research, Professor Veloso and her team "have actively researched on the integration of reasoning, perception, and action in teams of agents that need to face adversarial environments" (p. 2, emphasis added).
82 See "Invited Speakers" list at www.boeing.com/nosearch/ijcai/invited.htm . Visited 28 May 2001. 195 Chapter Six: Embodied & Embedded Epistemology
The realm of philosophy is oriented similarly to the direction of research in contemporary
AI. In a recent (Winter 2001) issue of Hypatia, the premier journal of feminist philosophy, and the one in which Weasel's paper appears, there is a call for papers on Feminist Science Studies.
There was also a call for papers (due August 2001) for a special issue of feminist epistemology in the journal, Social Epistemology: A Journal of Knowledge, Culture and Policy.
Interdisciplinary work, as called for in these journals, promises to bring about ideas that are unlikely to arise in any one of these areas. Clark gives a specific account as to why this might happen:
The sheer number of intellectual niches available within a linguistically linked community provides a stunning matrix of possible inter-agent trajectories. The observation that public language allows human cognition to be collective [reference to Churchland] thus takes on new depth once we recognize the role of such collective endeavor in transcending the path-dependent nature of individual human cognition . . . By allowing such results to migrate between individuals, culturally scaffolded reason is able to incrementally explore spaces which path-dependent individual reason could never hope to penetrate (Clark 1997, 206).
An important part of furthering my feminist analysis is to identify both the theoretical and practical implications of feminist philosophy of science applied to work in computer science and
AI. I have made some of theoretical points here. In the next chapter I review in more detail actual projects in computer science and AI.
196
CHAPTER SEVEN
COMPUTING WOMEN: FEMINIST PHILOSOPHY OF SCIENCE AND AI
An important part of this analysis is identifying both the theoretical and practical implications of feminist science criticism applied to work in computer science and artificial intelligence. One immediate result of the stereotype of women as less capable in mathematics or technology is that many women do not pursue work in mathematics, science, engineering, or the computing sciences. I cited statistics above in chapter three which indicate that this lack of women is a serious problem with regard to the well-being of nations, institutions, as well as to those individuals who have lost opportunities to pursue creative, professional, and rewarding work. In this final chapter I reinforce the connections between feminist science criticism and an actual computing science project. The AI project, called CYC (for encyclopedia), is just one example, but it is important to talk at some length about a real example, to illustrate the variety of ways ideas about gender and other relations involving power are present in something as seemingly neutral as a computer program. The discussion will illustrate how values play a constitutive, not merely contextual, role in computing science. What happens in this project can happen more broadly in computing and science. This analysis is meant to be not abstractly critical, but constructive in the spirit of Lisa Weasel’s suggestions about what feminist science could be (Weasel 2001). The discussion here is one part of a broader movement for analyzing a range of components of computer science including processes, institutions, methods, outcomes, etc. Assuming a role Londa Schiebinger (1999) suggests for feminist criticism, I do this with respect to women primarily but not to the exclusion of men. Chapter Seven: Computing Women
Today computing, like medicine, has the chance to impact people faster and more directly than other scientific disciplines, whose findings can be more remote from our daily lives.
The impact of values embedded in computing or medicine is potentially quite powerful and influential because those disciplines affect many people. More importantly, their effects are often subtle and unnoticeable. In 1962 Rachel Carson wrote Silent Spring, about the danger of cumulative effects of poisons on the environment and on humans. She showed how a small amount of a poison might not be deadly to an individual organism, but over time and repeated exposure it can build up and become harmful to other beings and whole systems. Sometimes the danger does not arise itself until well after the application of the pesticide or herbicide; sometimes it does not manifest itself in the same physical location. Thus the effects were hard to notice.
I am suggesting that there is a similar danger of gender bias dispersing through knowledge absent any review of the potential harm or discussion of alternative ways of understanding our position in the world. Unanalyzed ideas about gender could result in a situation analogous to another one discussed by Carson, namely, biological magnification, whereby a poison gets stored in an organism and builds up. When another eats that organism, the concentrated level of poison is also consumed. This can repeatedly occur moving up the food chain until the poison concentrates at a level that is toxic to that which consumed it. My worry with regard to gender in computing is that conceptions and images of women and girls found in computer programs will assist in delivering poison—in the form of harmful ideas about gender—to a substantial number of people. When those small doses are combined with negative views about gender from other sources such as religious, legal and educational systems, the mass media, and popular culture, an atmosphere inhospitable to women and girls is created.
198 Chapter Seven: Computing Women
Unanalyzed gender bias “feeds” into programs which are “consumed” by adults and children and creates a situation parallel to one Carson warned us about, when she wrote that
“children are more susceptible to poisoning than adults” (1962, 23). Carson was not necessarily opposed to the use of pesticides, but she clearly believed that we should be judicious in our application of them, not applying them ignorantly or without considering long-term consequences and/or alternative (more natural) solutions. Likewise in computing, it is not possible to eradicate all stereotypical views such as ideas that not all women desire to have children or that some men would prefer to have primary child-care responsibilities. On the other hand, ignoring the effects of beliefs about gender on computing and the sciences misses an opportunity to enable a series of changes in the general outlook.
How might ideas about gender get incorporated into a computer program? One way is indirectly through ideas about rationality employed in artificial intelligence. As considered in chapters one and six, rationality is the subject of a set of feminist critiques in philosophy as well as in philosophy of science. Thus, it should be of no surprise that it turns up as an issue for feminist scholarship in computer science and AI, as discussed in the previous chapter.
Neither the processes behind nor results of research and development in computing sciences are value-neutral; as such they are subject to critical review. Numerous directions are possible for computing research and application. For example, some person or institution, like the military, might need a computing professional to create hardware to survive a military invasion in a desert in order to track troop movements and have reliable communications among those troops. A different client could request that another professional focus on the best possible graphical interface. A third professional might develop procedures for processing huge amounts
199 Chapter Seven: Computing Women of data as fast as possible with only a stripped down human-computer interface. Yet another client might have reason to develop a system that processes commands from more than a single source or by just one operator.
The needs I have just listed all have actual solutions, some of which differ at the software level, others in their hardware. The point of listing this sample of needs is to show that there are many ways to meet a need, but the need must first be articulated. Importantly, the needs that are not voiced cannot be addressed. Through analysis of the stated and implied goals associated with artificial intelligence, I have been identifying some of the needs that have been ignored.
Importantly, for my position, it is social factors that play a significant role in allowing some needs, but not others, to be addressed. Gender is one of those factors.
ARTIFICIAL INTELLIGENCE
In the previous chapter, I discussed various aspects of artificial intelligence. Let me review briefly. There are several ways to define artificial intelligence. Russell and Norvig’s
1995 introductory AI text, Artificial Intelligence: A Modern Approach, defines AI as the study and creation of systems that think or behave rationally, or ones that think or behave like humans.
Lest one dismiss this definition as idiosyncratic, note that the authors explicitly mention that they have surveyed eight other introductory textbooks in the field in developing their definition (1995,
4-5). As further support
Until the mid-1980’s, AI researchers assumed that an intelligent system doing high-level reasoning was necessary for the coupling of perception and action. In this traditional model, cognition mediates between perception and plans of action (Brooks 2003).
This quotation was taken from the webpage of Rodney Brooks, an MIT researcher in AI. When artificial intelligence is defined as the study of building intelligent or rational agents, the terms of
200 Chapter Seven: Computing Women the research implicitly favor cognitive aspects: artificial intelligence. In this view, embodied knowledge is contrasted with rational knowledge codified in propositional form. Consider the emphasis in this quotation from the text: “Humankind has given itself the scientific name homo sapiens—man the wise—because our mental capacities are so important to our everyday lives and our sense of self” (Russell and Norvig 1995, 3, emphasis in original). Surely, as I argued in chapter six, our mental capacities are important, but not to the exclusion of embodied knowledge, which is suspiciously absent from this definition as something necessary even to daily activities. Neither does the definition encompass the embedded aspects of intelligence, also discussed in the previous chapter.
This approach assumes that work in artificial intelligence can be successful by studying mental capacities without regard to embodiment. Doug Lenat’s CYC, a knowledge-engineering project located in Austin, Texas at Cycorp, is an example of a particular AI project that also
83 seems to make this assumption. The long-term goal for a project like CYC is to have the program learn on its own, and to increase its understanding so that it can solve problems such as identifying which group of humans is gathering information about it. Learning is one of the
84 central requirements for an intelligent agent and Lenat does intend that CYC learn to be intelligent. On the positive side, CYC might be doing well in including some aspects of embedded knowledge.
EXPERT SYSTEMS
Expert systems (ES), such as medical diagnosing programs, encode the knowledge of a human expert. Often human experts are the only ones who can effectively solve a given
83 See www.cycorp.com for more information. 84 See Russell & Norvig 1995, 4-5 for a description of the characteristics of intelligence, as outlined by Alan Turning, in what has come to be called the Turning Test; a description of a more modern version is there also. 201 Chapter Seven: Computing Women problem, in part because even under conditions of uncertainty they can make good decisions based on their experience and training. For example, doctors often make diagnoses and recommend treatment based on their diagnoses even though they have little evidence, sometimes because it takes too long to get the information or in other cases because the test to get the information is too expensive. Generally this is successful for doctors and their patients.
However, there are many situations, in medicine and in other areas, in which it is preferable for a computer to make the decision. These include situations in which a reliable or safe decision must be made under emergency conditions and time restrictions, such as when a power plant unexpectedly comes off-line and must be quickly restarted to avoid tripping the local power grid, which could possibly cause a widespread power failure. In other cases, the program is used to address circumstances that are too dangerous or too tedious for a human. For example, the
PROSPECTOR program helps locate hard-to-access mineral sources (Rich & Knight 1991, 548).
Another advantage to having expertise encoded in a program is that expert’s knowledge can be made available in more than one location, which reduces the burden on that person.
Unlike most humans, however, expert systems are restricted in their scope, working only in narrow domains. A narrow scope makes the system brittle, meaning that it can only answer questions about subjects in its knowledge domain, and if it is asked to answer a question outside of that domain it will immediately break—or at best give a wildly wrong answer. It is that tendency to break down which makes it seem brittle. The narrowness, however, is necessary because of the sheer amount of information involved in even the most specialized area. A program trying to cover too much ground would be extremely unwieldy and too difficult to code.
So an expert system say, for chemical analysis, cannot offer assistance in medical diagnosing because no ES can reason about things outside of its range of knowledge, even relatively simple
202 Chapter Seven: Computing Women things that persons regularly take for granted. Importantly, brittleness in a system can lead to disaster. One of Lenat’s favorite examples illustrates the problem: "a skin disease diagnosis system is told about a rusty old car . . . [and] concludes it has measles" (Lenat, Guha, et al, 1990,
32). The ES tries to answer the query with the information it has, but because it is such limited information, with respect to the whole world of information, it gives answers arising from that limited knowledge base, answers which look ridiculous to any person with even a bit of common sense. Thus, the primary goal of Lenat’s CYC program is to overcome brittleness by formalizing and encoding common sense, thereby solving one of the major problems for expert systems.85
Lenat views common sense as a substratum that serves to facilitate more complex reasoning by providing the basic or obvious information needed to make bridges and connections from one program or domain to another. The rationale behind formalizing common sense is that the information contained in the substratum can help find or make connections between information contained in expert systems and other kinds of programs, e.g., word processors, spreadsheets, and Internet browsers. One short-term application of the CYC program would be to help integrate a word processor with a data base program. Part of what CYC is supposed to do is to assist with natural language processing, such that a user could ask the computer program to gather data from a data base, based on questions posed in everyday language, for use in the word processor program. For example, if the computer “knows” that there are different kinds of doctors then when you ask for a list of addresses of physicians, a query to an expert system for the names of doctors will not generate addresses of professors, even though both groups are referred to as “doctor” in everyday language. This is a relatively simple example; Lenat’s ambition for CYC is on a much grander scale. The connections or bridges from one program or system to another make the information more accessible to the user and allows the computer to
85 It is notable that Lenat’s mentor at Stanford, Edward Feigenbaum, was a pioneer in expert systems research. 203 Chapter Seven: Computing Women do more for the user such as more sophisticated searching and automatic updating. Overcoming brittleness would likely ease the application of expertise to a wider range of problems, since systems could then be combined to solve complex problems.
EMBODIED AND EMBEDDED AI
As argued above, rationality is not a neutral term. In artificial intelligence research—the field which gave rise to CYC—there has been an unacknowledged and sometimes unnoticed split between embodied and propositional knowledge, where propositions are supposed to represent
“rational” knowledge. Expert systems may be “rational,” but only in a formal sense. That is, they can follow rules, even very sophisticated ones, but that is all that they can do. The
“expertise” of an ES is very limited. Part of the limitation is due to a restricted scope. Thus, querying an expert system about something outside its scope of knowledge will fail to produce a relevant answer. It is ironic, really, to claim that expert systems are rational given that they are built upon expertise that itself is rarely rule-based. Real human experts often have skills and knowledge that they cannot explicitly codify, a state of affairs which seems to indicate that they are not fully rational. Such a conclusion—that real experts are not really rational—obviously cannot withstand scrutiny.
Work related to embodiment—the role that the body plays in epistemology—has been relegated in the past to robotics. It has not been seen as a central component to AI. A feminist critique declares that there are additional grounds for concern when it is the knowledge of women and others who lack social power and status, that is devalued or ignored as such (Adam
1998; Dalymiya & Alcoff 1993; Grosz 1993; Peiris, et al 2000). Their knowledge is often understood to be non-propositional (as embodied, skilled, or tacit knowledge) in form. In
204 Chapter Seven: Computing Women contrast, propositional knowledge, which is seen as the medium for rational knowledge making, is often associated with masculinity and social power and is prized as “real” knowledge (Van
Oost 2000). The result of this dichotomy, which I assert is harmful at worst and false at best, is that the CYC project and by analogy other AI projects are ill prepared to understand and incorporate non-propositional knowledge. That means that they are epistemically incomplete.
Such projects put mistaken theory into practice. The beliefs and assumptions behind this project—what was discussed earlier in terms of epistemic context—do not apply to CYC only, but
I will limit my remarks here just to CYC.
The good news is that this outlook seems to be changing as an increasing number of researchers are moving into work on robotics and projects involving multiple users and multiple agents, all of which seem to fit very well with an epistemology which is both embodied and Need citations. embedded. The change is promising because a socially embedded, embodied epistemology, as discussed in detail in chapter four, better explains the success of real experts and real expert systems than a more traditional account. It explains better because the social view helps to locate assumptions that are hidden but necessary the functioning of such systems.
I will use the approach of an embodied and embedded epistemology, which I have been developing throughout this dissertation, as the basis of critique of Lenat’s CYC project. My examination of CYC pays special attention to criticisms that highlight the failure of the CYC team to incorporate non-propositional knowledge in the process of “capturing” “common sense”.
Feminist science scholarship is especially helpful in discerning the difference between these types of knowledge, because of its ability to asses the epistemic roots as well as the political and ethical dimensions of the problem, as shown in chapter two.
205 Chapter Seven: Computing Women
TRANSMITTING BIAS
One problem for AI results when values of the programmer are carried through to the program opaquely. To varying degrees the feminist science scholars I rely on concur that this happens all the time in science. Bias does exist in scientists and scientists do transmit their own views of the world into their science. Bias is not necessarily a bad thing. But how the wider scientific or social community scrutinizes, or fails to scrutinize, particular biases can be criticized. For example, unchecked stereotypical assumptions could be transmitted into a final product, if there are no standard safeguards against such occurrences. Assumptions can be harmful because of their transparency or invisibility to those who construct, use, and are portrayed stereotypically by the system. Programmers therefore have a lot of unnoticed power.
In the same way that language can be harmful when racist assumptions are subtly embedded within it, certain AI assumptions “infect” the programs with which they are connected.
Computer scientist and feminist critic Alison Adam describes it this way: “CYC’s models of the world are hegemonic models—unconsciously reflecting the views of those in powerful, privileged positions” (1998, 86). Before assessing how CYC works, let me explain in more detail how the beliefs of a programmer might become embedded in a program. Cecile Crutzen and
Jack Gerrissen (2000) write specifically about object-oriented programming and how it lends itself to propagating hegemonic views.
Crutzen and Gerrissen explain object-oriented (OO) programming, and its use of classes, by reference to films. The characters in films do act and interact, but they are still defined and scripted by the playwright. The analogy is between the characters in the film and objects in the computer language. If one forgets that these are scripted characters one might not realize that the
206 Chapter Seven: Computing Women characters could act differently in real life. That is how a single viewpoint could come to be accepted as the view. They write:
at the same time these scripts open up the potential of the dominance of views concerning how and why the world has changed and how this OO [object-oriented] play should change the interaction of the ‘audience’ with the represented reality. But software engineers are not aware of that potential of subjectivity . . . this supposed neutrality and objectivity are precisely the dominance and power that software engineering exercises (2000, 130, references omitted).
To review from the introductory chapter, object-oriented (OO) programming, such as what is done with the C++ language, allows the programmer to use abstraction in her program.
This allows the programmer, for example, to write code without knowing too much detail about the data. This can be quite helpful because it allows one to reuse entire sections of code, defined as classes. But this can also be a detriment because of the ease with which the code can be reused. Instead of trying to find a suitable means for capturing and expressing the data, the programmer can pull up something that is predetermined and static. The structure of the code preserves whatever was originally encoded in the data: “OO conserves and closes the meanings of the past and does not open up the future as a projection of our possibilities which we, as beings in the world of every day, could anticipate on [sic]” (Crutzen & Gerrissen, 128). The point I want to emphasize is that these are not ethically or politically harmless decisions.
Lisa Bloom (1994) argues in “Constructing Whiteness: Popular Science and National
Geographic in the Age of Multiculturalism,” that even though National Geographic claimed to be an unbiased and even progressive source of information, including pictures of peoples around the world, in fact the magazine was both intentionally and unintentionally perpetuating racism.
Her thesis is that ideas about science and technology work “in tandem with the trope of racial otherness to preserve a static and homogeneous notion of whiteness. . .” (p. 18). For example,
207 Chapter Seven: Computing Women when “we” photograph “them,” or when Americans bring new technology to the “locals” and they, the locals, are dark-skinned, then we, the Americans, being different than them, must all be the same, somewhat united in relation to an “other”. The distinction of ‘other’ is very subtle, yet very powerful and the longer it goes unnoticed, the longer its reign of influence.
Similarly, CYC risks perpetuating certain views about the world, views that are hard to challenge since they are hidden within the structure of the program and will not be challenged by users or developers. The inferencing goes on in the background, so assumptions used by the programmers might escape notice. CYC’s common sense might very well “believe” certain traditional ideas about women, gender, sexual orientation, etc., and then make inferences based on those “beliefs.” Without any strong challenge a homogenizing effect occurs, solidifying the original stereotype among users of the program. This problem could be addressed, perhaps, if the programmers could make public the methods they use for ensuring that the concerns raised by stereotyping have been addressed in the programming process or in the ability of the user to trace assumptions back to their source.
Is this characterization overstating the case? As just another computer program could
CYC have a widespread impact on society? Notably, computer technology has revolutionized the way we live in the same way that the industrial revolution did. According to the ABC Nightly
News, Alan Greenspan believes that growth in technologies is in a large part the driving force behind the American economy (January 14, 2000). Computer software is clearly a component of the technology industry. If Lenat has his way, CYC would be installed on every new computer in much the same way that Microsoft Windows® is pre-loaded on many computers. It is hoped that CYC will be of sufficient value that there will be millions or more copies of it in use, the same way that there are millions of copies of Microsoft Word® (Moody 1999).
208 Chapter Seven: Computing Women
If CYC were to gain that kind of exposure, its influence would be tremendous. Moody observes that Lenat has “plans that might ultimately affect just about anyone who interacts with computers in the course of their mundane existence—which means, of course, just about everyone.” Of course it really is not everyone, only those privileged enough to have computer access.
THE (COMMON SENSE) WORLD ACCORDING TO CYC
The conceptual foundation of CYC suffers from several interrelated problems. First, it posits the existence of a knowable, unified common sense. Second, it disguises a particular version of common sense as the universal Common Sense. By doing so it creates a homogenizing effect on how groups are viewed by perpetuating stereotypes. Third, by relying on a single common sense, CYC implicitly encourages ignorance of other common senses and their potentially valuable contributions.
To begin with, it is not clear that common sense can even be fully known. What is it and how would we know it if we had it? That it is continually evolving is an important consideration because that makes it much harder to fully capture. GAP I have been arguing that all knowledge involves mind, body, and world; and common sense is no different. It includes rational knowledge that can be represented by sets of propositions, but it also includes ways of getting around in the world, which frequently are adjusted in the moment. For example, common sense helps you figure out how close to stand when speaking to someone in a foreign country. You might learn that by reading a guidebook, but you adjust it for the particular situation, with that particular person. It is something you sense. It is not something that is fully knowable without taking that embodied aspect into consideration.
209 Chapter Seven: Computing Women
Common sense then is something that is known in common and known in common, in turn, means what is known by all or most members of the group. Across different groups, the common sense can vary. Going back to the example of conversing with a foreigner in her own country, the common sense she relies on is, in my scenario, different that the common sense I rely on in the setting of my own country. Common thus does not mean universal or monolithic.
There are many common senses, not just one. A common sense should not be confused with the common sense.
The point about the range of common senses is important in a practical sense. Laboring under the view that The Common Sense exists encourages everyone to ignore alternative views.
That, in turn, potentially omits helpful or new knowledge. Disguising a particular understanding as a universally applicable common sense is disingenuous at best and destructive at worst. It can
‘disappear’ gender as well as class, race, and other potentially important factors by subsuming all of them under a single conception. Adam (1998) cites Lenat to illustrate this problem. Just as common sense itself is to be taken for granted, so it seems are those who possess such knowledge, ‘be they a professor, a waitress, a six-year old child, or even a lawyer’” (Adam 1998,
85). She goes on to say “Given such a variety of type of subject, even within one culture, it would not be difficult to argue for different views of consensual knowledge for each of these subjects” (Adam 1998, 85). CYC ignores the minority views, the quieter voices, and allows the majority voice speak for everyone (Adam 1998, 39, 89).
This is not to suggest that every person on earth must have a say in what is understood as common sense, but that proper precautions must be taken to ensure a good mix of views have been incorporated into a common sense. This is absolutely crucial with knowledge which is to be codified and packaged in the center of a product. This is especially critical where the original
210 Chapter Seven: Computing Women work of identifying what is common sense and what is not is secured from inspection. For example Alison Adam raises the concern that CYC fails to make allowances for people with disabilities. How so? Presumably, common sense for the hearing impaired varies tremendously from those who can hear. As a result any inferences from the basic information supplied by common sense to action could vary as well. The concern here is that if those other views are not included in the basic knowledge base, all the inferences arising from that knowledge base are tainted or incomplete and not universal. Yet, at the same time, the project is billed as complete and universal.
This is not just putting forth a unified common sense, but a particular view as the important one to consider. It should be of concern to feminists that these other views are being subsumed. A related problem is that the common sense of a particular group, for example the
CYC programmers, might hold biased views about members of another group, such as the very poor. The dissemination of such views further reinforces them in the same way that the embedded racist assumptions in the National Geographic example further racist views.
The guise of universality overlooks potentially valuable contributions to the general common sense pool. For example, views of the very poor are not included since that body of common sense knowledge is not useful to the CYC project, because poor people are not going to buy the product. Embodied knowledge is not included either, although there could be attempts to do so, as there are in other projects, notably in the field of robotics. Apparently embodied knowledge is not as important or as “real,” as propositional knowledge is, according to the paradigm in which CYC has invested. Perhaps the exclusion of these other views would be reasonable if the project were billed differently: “A Common Sense for materially comfortable business people who use often use computers to help them solve problems.” Perhaps CYC
211 Chapter Seven: Computing Women common sense is just short for that long description, or perhaps the project leadership just does not consider those other views worthy. Of course this is a lot to ask of one program which started fresh from the ground up. The point here is that we cannot allow an incomplete program to sell itself as complete, because while most of the harm is probably minor, there could very well be some serious repercussions at some point, particularly if the inferencing takes place
“unsupervised” by the user, which is what happens in the computer program because the program takes care of all that inferencing for the user: that is exactly its purpose.
What is at stake in not representing these ignored voices? Is there any real harm perpetuated? Some examples should clarify my concern. A system that only incorporates technological medical knowledge and excludes knowledge of traditional practices implies that the traditional practices and beliefs are not worth formalizing. In doing this—picking out some information as worthy and discarding other—the programmers are subtly shaping the identities of themselves and potential users, a concern Adam (1998, 39) voices. Some knowledge might come to qualify as truth but the rest of it is excluded as just superstition, mere belief, old wives’ tales, myth or folk belief (Dalmiya & Alcoff 1993).
Another example of the power of programmers concerns medical diagnosing expert system. For a programmer to assume that the users who are health professionals and the subjects, the patients, are similar to the programmers themselves can create serious problems.
Take for example, variation in patient responses as discussed by Boom, Chavez-Oest, and Boom
(1995). They discuss the use of expert systems in the medical arena, cautioning against including expectations for patients without incorporating data local. Not all patients react the same way, so a patient response should not be projected as universal in character. Instead local data must be consulted to make sure that the program works for a specific population. The
212 Chapter Seven: Computing Women programmers have the choice to include local data or to make the user aware of the importance of incorporating local data, but first they must recognize that this power lies in their hands.
Surely not every belief can be included as common sense; some of our knowledge is specialized and so is not common, some is not knowledge but is trivial or informative or opinion.
When information is formalized, someone or some group is deciding what to include under the label ‘common sense.’ Power is embedded in the programming process. This power is invisible, however, and that makes it hard to show the potential for harm resulting from it. Common sense is ubiquitous, yet simultaneously extremely hard to define. Those facts make it hard to show that certain knowledge is or is not definitively important to include in the CYC project. Again, the worry is that unfairly biased views about a variety of characteristics of persons will become established as part of common sense, in the same way that racism became established within
National Geographic or that data in an ES can reflect the programmers’ own situation or characteristics. The Cyc programmers are involved in a social epistemology project of sorts, but the social group is too small to be sufficiently representative or to effectively guard against bias.
Futhermore, the programmers cannot substitute or translate or feed their own embodied knowledge into this program. CYC’s ignoring embodiment raise deep problems for its conception of common sense precisely because sense is embodied. Embodied knowledge is foundational to propositional knowledge (Adam 1998; Collins 1990; Dreyfus 1992). Because this epistemological priority exists, CYC does not have a good chance of being successful because it is missing a fundamental component in its knowledge construction. CYC and other AI work, is particularly ignorant of embodied knowledge (E-knowledge) in deference to propositional knowledge (P-knowledge). Recalling the discussion of the earlier chapter on embodied and embedded epistemology, embodied knowledge is closely related to skilled
213 Chapter Seven: Computing Women knowledge, knowledge that arises as the result of practicing a skill of task until one can do it
“without thinking.” This definition makes it seem possible for E-knowledge and P-knowledge to be separate. The view is that one starts with propositional knowledge, i.e., instructions that can be written down such that the knowledge is fully transferable to the reader. As one develops the skill one moves away from those rules.
Another sense of E-knowledge is that which is the fundamental basis of all knowledge, that by which all other knowledge is gained and interpreted. This is the sense employed by
Adam (1998; 1995) and others (Dreyfus 1992; Lakoff & Johnson 1999). Not all these authors use these terms in exactly the same way. As argued in earlier chapters, I take information about one’s embodied and embedded position to be fundamental to epistemology; glossing over the fact that not all knowledge is propositional in character severely hampers the CYC project and any programs incorporating the CYC module. CYC is vulnerable to gaps in its knowledge since certain aspects of epistemology are neglected. By excluding embodied knowledge, CYC potentially limits the contributions of persons who for whatever reasons are not well-versed in presenting the knowledge they do have in propositional form. In other words, a human user may not be able to “translate” what she knows qua embodied agent into something propositional, and in that respect, Cyc might be more harmful than helpful, if humans are forced to limit themselves to dealing only with propositional knowledge.
So is CYC doomed if it neglects embodied knowledge? The fact that E-knowledge is not even considered for inclusion in the common sense project is problematic in itself, but is the criticism overstated? No, it is not. In its current form, CYC is a tool, not an artificial mind. A proponent might argue that CYC is not creating new knowledge; it is reorganizing and analyzing the knowledge that is entered by persons. So, it does not need to have the actual embodied
214 Chapter Seven: Computing Women experience in order to have knowledge. A critic could respond by saying that it might find new connections or find more connections among the propositions at a faster rate than a person, but it is hampered by its inability to gain knowledge as an embodied agent. Without access to embodied knowledge, Cyc cannot be thinking or learning on its own. Therefore, it cannot be said to have artificial intelligence.
In the beginning the AI community was skeptical of Lenat’s project. No one had pursued his method because it seemed inelegant and overwhelming in scope. Begun in 1984, the CYC project was “a pioneer system in the world to attack the problem [the commonsense problem] in the scope and current scale of about 1 million rules (reduced from 1.5 million by compaction)”
(Munakata 1995). Lenat’s training at Stanford University and his own successes in knowledge- based systems, helped him convince many people, including financiers, that the project had merit and a viable future. Today, however, most professionals seem to consider the CYC project past its prime. Although CYC can be credited with some impressive results, they are not fulfilling
Lenat’s promises of revolutionizing AI, because he ignores embodiment and insufficiently utilizes social embeddedness. It may be, though, that Lenat has relinquished his claim regarding
CYC as an artificial intelligence project, if indeed he ever thought that (Hiltzik 2002).
RESCUING CYC
CYC’s goal is to fill in the underlying strata that humans rely on to figure out problems and increase their knowledge. This would reduce brittleness, allowing expert systems to do more work for us. Including a diverse set of views would increase the scope of CYC’s knowledge base. Strengthening CYC’s epistemic foundation would provide more chances to reduce brittleness by finding connections among a variety of information sources. For example, cross checking a medical problem against a diverse set of common sense perspectives about disease,
215 Chapter Seven: Computing Women treatments, causes, etc., could facilitate the discovery of unexpected solutions. Whereas relying on a mainstream, “unified” common sense might overlook a non-standard solution. If we continue to think of common sense as unified then there will be no justification or motivation to search beyond the standard views for alternative explanations.
So, if the goal of CYC is to reduce brittleness, more information not less is required.
Thus, CYC should include more voices, especially underrepresented ones. The basic or obvious information needed to make bridges and connections from one program or domain to another plausibly includes the common sense of groups who do not command much social power. In addition, the practice of disclosing the “location” (background beliefs and commitments) of the programmers should continue. This allows solutions to be traced back to their component beliefs and ideas for testing against the original problem specifications. This encourages responsibility on the part of the programmers and allows users to perform careful checks by hand should the need arise. In fact, CYC is constructed in a way to handle this method.
There is a solution to the problem of using a unified common sense and to the problem of using multiple common senses that contradict each other; isolate them within their own little worlds called microtheories. CYC permits, even expects, its knowledge to be structured into lattices of reified, formalized contexts. These microtheories represent different points of view, levels of detail, differences in culture or nationality, age differences, time periods, corporate cultures, (Guha & Lenat 1994, 136; Lenat 1995b, 82; Stork & Lenat 1999, 11).
Incorporating the feminist-friendly practice of disclosing the cultural and other
“locations” of programmers would fortify CYC against some of the concerns I have been raising.
This means identifying or “tagging” a microtheory to its developer(s). That would allow inferences to be traced back to their component ideas stemming from a particular person, which
216 Chapter Seven: Computing Women allows them to be tested against the original problem specifications. Tagging encourages responsibility on the part of the programmers, allows the development process to predict possible conflicts or unusual readings and can warn users about such conflicts, permitting users to perform checks by hand by going back to the source should the need arise. Although CYC does not currently include widely diverse perspectives, if those deficiencies were straightforwardly acknowledged, they could be included at some later date without damaging the structure of the program. Particularly when the deficiencies are not acknowledged and thus not open for discussion, we should worry that CYC might be harmful to some segments of society.
The strengths of the microtheories notwithstanding, the lack of embodied knowledge, as argued above, is a serious deficiency. Simply adding in embodied knowledge is impossible. The best that the CYC project can do now is to take the work to date as one stage of a much larger work in progress. Scaling back the expectations of the project from one involving artificial intelligence to something more like a tool is in order. CYC cannot be called a knowledge project in the fullest sense of the word “knowledge,” given the arguments I have made throughout this dissertation. Without a complete basis for knowledge, Cyc cannot truly be said to be capturing artificial intelligence.
AI RESEARCHERS WANTED
Using the findings of scientific research on intelligence and cognition broadly is as important as having diversity among the researchers themselves. Computing and AI are another source of information and in some cases, testing, about how we generate knowledge. A naturalized epistemologist should look to computing for information that would help to correct beliefs about our epistemic strengths and weaknesses. Researchers like Rodney Brooks at MIT are concerned with “understanding human intelligence through building humanoid robots”
217 Chapter Seven: Computing Women
(Brooks 2003). His entire basis for research has shifted away from the standard view to one that he calls behavior-based robotics, of which “the cornerstone. . . is the realization that the coupling of perception and action gives rise to all the power of intelligence and that cognition is only in the eye of an observer” (Brooks 2003). Empirical research like that of Brooks helps us reevaluate our own conceptions of intelligence, which in turn helps us move towards the successful creation of AI. Especially with something complex like intelligence, which is controlled or at least influenced by so many unknown variables, it is impossible not to simplify in order to create an artificially intelligent agent. The empirical work thus must start small. As the work builds on itself, it must continue to be informed by multiple perspectives, empirical data, as well as logical analysis.
There is a responsibility that comes with research in AI. There is also a responsibility that comes with creating a product that potentially could reach tens of thousands of computer users. The charge is this: the research needs to be thorough and researchers need to be cognizant of the power that they wield. Representing a product, based on a particular strand of theorizing about how computers could come to think and learn on their own, as universally applicable is disingenuous at best and harmful at worst. If the product merely represents the seller’s interests, while simultaneously announcing that it is universally representative, something is wrong. There is clearly a conflict of interest here: selling a product is not wholly compatible with developing a solution—even a partial one—for creating artificial intelligence.
The product might be a contribution to the wider project, but then it would appear incomplete to a buyer, who then might shop elsewhere. Instead, the need to sell the product could compromise the research as well as the communication about the findings of the research. Any legitimate attempt at working out a theory of artificial intelligence needs to account for relevant data, i.e.,
218 Chapter Seven: Computing Women other research about how people do learn, the role of the body, how memory works, how language functions in the social group, etc. Legitimacy also requires a fair, open assessment of one’s research—something which is not often done in our capitalistic system. Since it could be done, though, it is all the worse when it is not done. For example, the biotech company, Geron
Corporation, has its research reviewed by an ethics panel.86 The company will not blatantly disregard the recommendations of the panel, even though that decision might result in some loss of profits for the company.
The CYC project, as far as it has come, has not done enough. As a potential component in a wide range and great number of computer programs, the emphasis has been on perfecting the inference engine and programming the knowledge database. This has not left room for another layer of evaluation regarding what representations this program promulgates or how those representations are generated. Nor is Cycorp alone in this failure. Not often enough does a company consider how its product will affect those who are represented negatively by it.
If the language of science—one of the most objective realms—is not free from values, and if the allegedly most objective of scientific fields (i.e., physics) has subjective influences
(e.g., Traweek 1988), then computer programs are not immune either, even those which are written by very smart people and those that work well in their own domains. This is not cause for despair, although it might be cause for reorientation towards science and philosophy of science, as I have argued as a feminist philosopher throughout this dissertation. The “common” in “common sense” references the social. What is common is shared by members of the group and what we know we know with the assistance of others, and it is by explicit inclusion of the social that objectivity in science is protected. Another sense of social epistemology is
86 See. www.geron.com and go to “Who We Are,” then “Advisors,” then “Ethics Advisory Board” for more information. 219 Chapter Seven: Computing Women knowledge about the group. One of the major premises of this project on Computing Women is that social epistemology is essential to science. Knowledge is produced by embodied agents in a social context. Those aspects of social epistemology make it possible to further science without allowing it to be highjacked by the values of any one person or any one group. Starting AI projects from embodiment prevents an overemphasis on rationality as the basis for intelligence.
Accepting this argument about the importance of an embedded, embodied epistemology might be as hard as accepting that there is an alternative to competition in science, but there are many good reasons for accepting both, ranging from a naturalistic argument about what kinds of evidence are relevant to epistemology, to what feminist science critics have said about how values do permeate science and what can be done to safeguard science, as I have shown throughout this dissertation on Computing Women.
220
CONCLUSION: FUTURE PATHS
Our fundamental human nature is challenged daily by new technology. Rodney Brooks’
AI Lab at MIT is researching “implantables,” which are technologies “that interface our nervous systems directly to silicon” (Dooher 2002, 82). The idea is similar to the one behind a hearing aid (cochlear implant). Brooks thinks that in time we will incorporate more robotic technology into our bodies, and more genetic technology into robots. As technology becomes literally more entwined with our physical bodies by means of pacemakers, artificial body parts, microelectrical- mechanical devices (MEMs), biologically engineered treatments and cures for diseases, the use of genetic engineering and possibly human cloning for creating families, it is even more important to be clear about what aspects of human nature are most important. Linda Martin
Alcoff, editor of Epistemology: The Big Questions, introduces the section “How Is Epistemology
Political?” with the following statement about the selected essays. She writes that each critic
“argues for a renewed, and deeper, understanding about what epistemology is doing, how it is affected by its social context, and how it in turn produces political effects” (1998, 385). I am especially interested in the connection between epistemology and the social, political, and ethical values within it. The intersections among epistemology, gender, and justice serve as a central motivation for this project. Ultimately I think these questions develop into conversations about who we are, namely, embodied and socially embedded creatures, and how we come to have knowledge.
My dissertation, Critical Values: Feminist Philosophy of Science and the Computing
Sciences, is an examination of the intersections between epistemology, philosophy of science, and feminist theory. My thesis is that feminist philosophy of science creates new and valuable ways of looking at the sciences by using gender as a category of analysis, or a lens through Conclusion which to critically assess and constructively build projects in science, as well as in the philosophy of science. I am employing feminist philosophy of science—using that gendered lens—to examine the computing sciences. This project examines computing sciences as a way to create a platform for exploring the dimensions and contributions of feminist philosophy of science. This is not merely a critique of philosophy of science or a feminist review of computing, but a positive project in its own right, examining the epistemological structure of scientific inquiry, including the nature of objectivity, epistemic agency and the composition of an epistemic community, the importance of those epistemic communities, and the role of values in science.
Feminist philosophy of science is feminist because it takes gender and other relations of power to be significant in the creation of knowledge. Feminist projects share histories and certain goals and aims, although certainly there is not just one feminist project or one underlying feminist theory. There are actually quite a range of feminist projects which draw on various other traditions both within and beyond philosophy. For example, some projects like those of
Donna Haraway lean more heavily on postmodernism, while others—philosopher Helen
Longino and her work in the biological sciences, as well as computer scientist Alison Adam and her critiques of AI—are very much at home with science and its technical and empirical research.
Note that the emphasis is on gender, not on the feminine and not just on women. Some writers have chosen to use the phrase “gender studies” rather than “feminist theory” to avoid such misconstruals. I use “feminist” because it does locate my project within a specific body of work in contemporary American philosophy, but in no way does this signal a focus on women to the exclusion of men. The lens is directed toward gender in order to improve the lives of all
222 Conclusion persons, including women. Categories like gender are often good starting points for assessing a state of affairs, or patterns of practice, or hidden assumptions. Due to the history of underrepresentation, exclusion, and sometimes outright harm to women in our society generally, as well as in science, this feminist project is committed to analyzing the epistemological foundations of science, foundations which, inadvertently or not, might disadvantage women.
That leads us to the problem of underrepresentation of women in some disciplines of science. The numbers are particularly low in the computing sciences. By computing sciences I mean to include all of the subdisciplines: computer science and engineering, information technology and information systems, artificial intelligence (AI), robotics, etc. Concerns about the low level of participation by women in computing, and a focus on gendered aspects of computing which might explain those low levels, serve as the basis of broader and more general critiques of computing. Using gender as a lens for investigation has led me to concerns about how computing is taught, what kinds of projects are appropriately pursued in computing, why one might want to pay attention to the composition of the computing profession, and which epistemological commitments are encoded in those projects and teachings.
Although many within computing base their arguments for the increased participation of women on simple claims of equality or on future productivity claims (taking the form of either more competent bodies or more diversity to better relate to clients or customers), I argue that those claims may be shortsighted. We cannot rely simply on strategies that will add more women without also examining the nature of the discipline; we cannot simply “add more women and stir”. Christina Björkman’s (2002) work in gender studies and computing has been invaluable for assisting my thoughts on the nature of computing and some of its epistemological commitments. Such commitments can be found in a discipline’s guiding metaphor, which has
223 Conclusion many values embedded within it. Those values guide the teaching, development, and practice of the discipline. For example, Lynn Stein (1999) argues that computing interpreted through the metaphor of a sequentially ordered series of steps to a specific goal, i.e., computation, abstracts too much away from the real processes of computing. She suggests as a replacement the metaphor of a community of interacting entities.
Feminist philosophy of science, then, starts with these basic tenets of feminist theory and feminist philosophy and asks questions about the influence of gender on the content and methods of science. Here the focus of such questions falls on computing as a whole, its products, and its creators. It is important to analyze computing because it is so quickly developing into a foundation for much of technology, much of science, and in turn becoming foundational to modern Western societies.
The questions also call us to re-examine the philosophy of science. In this dissertation I focus on a subset of epistemological issues in the philosophy of science, particularly the role of objectivity and values in science, and epistemic agency, including the embodiment of knowers.
In many ways, too, these issues connect epistemology to concerns of social justice or ethics, which is also important for a feminist project.
We all are “socially located” and as such have multiple identities from which we come to understand the world; scientists are no different. They come to know the world by virtue of their cognitive abilities which are limited by both their physical bodies and their social, political, and cultural contexts. This social location is called embodiment. Investigating embodiment involves investigating the concept of rationality. Rationality has come under scrutiny by many feminist theorists because it is so often juxtaposed with embodied or skilled knowledge. This mind-body
224 Conclusion dualism, in combination with other conceptual dualisms such as rational-emotional and male- female, forms an interlocking set of binary opposites such that male-rational-mind link together and female-emotional-body link together.
This is problematic for several reasons. One, the dualisms are artificial: females aren’t necessarily emotionally sensitive, males aren’t always relying exclusively on purely rational, non-embodied knowledge, and so on. In other words, the dualisms inaccurately essentialize the differences between males and females. Two, when the dualisms are arranged hierarchically, the associations with females traditionally have been devalued. Historically in Western philosophy and in AI, knowledge arising from mind and rationality has been privileged over knowledge arising from body and emotion.
My project takes issue with many of these associations and assumptions. First, there seems to be significant evidence that knowledge does arise from the body and is very important.
In addition to surveying the work of Mark Johnson, Andy Clark, Michael Polyani, and others, to explore how embodiment impacts epistemology, I also discuss the role of emotion in the work of science. Emotional engagement with one’s research is not something that detracts from the quality of the work; in fact it is almost a prerequisite: being emotionally connected to one’s work keeps a researcher interested, motivated, and attentive.
To return to AI, if embodiment is important to epistemology, then trying to create artificial knowers in AI without regard to embodiment seems troubling, particularly as many of the embodied aspects which seem to be ignored are ones which tend to be associated with women. Concerns about embodiment remain somewhat of an open question for me and will need to be revisited as I learn more about relevant research in cognitive science, and as that research itself develops. It is also a place for some future research with practicing computer
225 Conclusion scientists to learn more about the current state of the discipline. If AI should scale back its ambitions from creating artificial intelligence to “merely” creating specific and powerful tools for assisting humans, it would be much less open to attack. As it stands, I think it is selling itself short by insisting on its ability to create artificial intelligence.
Embodiment, defined earlier, also has a social context. I argue that social context is an important consideration for epistemology. Even more strongly, scientific knowledge arises from a community of knowers. The community itself is the epistemic agent. I employ the concept of equipoise to explain what it means for a community to be an epistemic agent. Briefly, again, equipoise is a state of uncertainty on the part of individual or groups of researchers with regard to scientific knowledge (Freedman 1987). None of these researchers can claim to “know” that the treatment (or hypothesis) works until there is consensus within the broader scientific community. Consensus is not reached by mere political or social maneuvering, but is anchored to scientific evidence and reasoning, and verification of results. It is often announced via publication or sometimes presentation.
Given that scientists differ in their social locations and employing the concept of equipoise helps me argue that biases are not detrimental to science but actually strengthen its methodologies. Sandra Harding uses “strong objectivity” to make this point. Without contributions from a wide variety of researchers, themselves with a variety of social, ethical, political and other commitments, science runs the risk of pursing projects that harm certain groups. That creates inferior science because it is empirically inadequate. Questions about empirical accuracy are epistemological questions. At bottom many of the fundamental questions about science and about technology are epistemological in nature, including questions about confirmation, data and evidence, and the nature of an epistemological agent in science or
226 Conclusion technology. Scientists are able to read data and use it to support hypotheses because of the social framework which gives them access to language, to education, to knowledge of the past, and to others who help to clarify, refine, and verify knowledge claims.
Social epistemology provides major support for my feminist analysis of the computing sciences. Knowledge is particular, not universal or abstract, with social and historical dimensions. Given the power of science in the modern world, its status as objective knowledge is not trivial. Objectivity is supposed to free science from manipulations having to do with political agendas, individual gain, or personal prejudices. Scientific credibility seems to rest squarely on an understanding of objectivity that allegedly protects the public and the science itself from such abuses.
I argue that if we want to secure objectivity in science, the way to do it is in and across communities. The community verifies evidence and reasoning, certifying it as knowledge.
Thus, scientific knowledge results from socially negotiated standards of evidence, of reasoning, of experimentation, etc., tested against and anchored in the empirical world. Standards for good and effective science are discussed and established at the meta-level, in the abstract, beyond any particular project. Using gender as a tool of analysis assists in uncovering assumptions and inaccuracies embedded in those standards as they pertain to specific projects or entire disciplines, which result in harm or neglect to people.
The example of man-the-hunter vs. women-the-gatherer hypotheses in anthropology illustrates both how gender symbolism works in theorizing and the impact of social context on epistemology. These two hypotheses, both of which fit the available data, were constructed by different communities of scientists—a predominantly male group and a predominately female group, respectively. Briefly, these two hypotheses try to integrate anatomical and behavioral
227 Conclusion information into a coherent story which matches the physical data in order to account for the evolutionary development of humans. The androcentric account emphasizes the alleged contributions of males, explaining tool development and thus development in intelligence and sociability as the result of males’ innovations in hunting. The gynocentric account, in contrast, attributes developments in tools for digging, carrying, and preparing food, and thus development in intelligence, to the supposed work of females as they tried to improve their abilities to gather food.
Researchers in computing and AI can learn from the arguments that feminists make, namely, they should not assume that the views of the researcher/programmer are necessarily shared by others. In other words, researchers should not uncritically represent their beliefs as universally true or universally applicable. Alternatively, researchers should try to consider what power relations are at work, who benefits, and who might be disadvantaged by these views. I argue that the AI project called Cyc runs a significant risk of presenting its product as something universally applicable when it truly is not. Perhaps it could be, but more careful research, perhaps informed by some of the work in this dissertation, is required to do so.
228
REFERENCES
Adam, A. (1993). Gendered knowledge – Epistemology and artificial intelligence. AI &
Society, 7, 311-22.
------. (1994). Who knows how? Who knows that? Feminist epistemology and artificial
intelligence. In Adam, A., Emms, J., Green, E., & Owen, J. (Eds). IFIP Transactions A-57,
Women, Work and Computerization: Breaking old boundaries—building new forms (pp.
143-155). Amsterdam: Elsevier/North-Holland.
------. (1995). Embodying knowledge: A feminist critique of artificial intelligence.
The European Journal of Women’s Studies, 2, 355-77.
------. (1998). Artificial knowing: Gender and the thinking machine. New York: Routledge.
------. (2000). Gender and computer ethics in the internet age. The CPSR
Newsletter, 18 (1)
http://www.cpsr.org/publications/newsletters/issues/2000/Winter2000/index.html
------. (2000b). Information systems: We still need a feminist approach. In Balka, E., & Smith,
R. (Eds.), Women, work and computerization: Charting a course to the future (pp. 102-
110). Boston: Kluwer Academic Publishers.
Adam, A., Emms, J., Green, E., & Owen, J. (Eds). (1994). IFIP Transactions A-57, Women,
Work and Computerization: Breaking old boundaries—building new forms. Amsterdam:
Elsevier/North-Holland.
Albright, J. (2000, March 9). Gender Pages of CPSR [Compter Professionals for
Social Responsibility]. http://www.cpsr.org/program/gender/ References
Alcoff, L. (Ed). (1998). Epistemology: The big questions. Malden, Mass: Blackwell Publishers,
Inc.
Alcoff, L., & Potter, E. (Eds.). (1993). Feminist epistemologies. New York:
Routledge.
------. (1993b). Introduction: When Feminisms Intersect Epistemology. In Alcoff, L., & Potter,
E. (Eds.), Feminist epistemologies (pp. 1-14). New York: Routledge.
Alper, J. (1993, April). The pipeline is leaking women all the way along. Science, 260,
409-11.
Anderson, E. (1995, Summer). Feminist Epistemology: An Interpretation and a Defense.
Hypatia 10:3, 50-84.
Andrews, L. (2001). Future perfect: confronting decisions about genetics. New York:
Columbia University Press.
Angell, M. (2001). The ethics of clinical research in the third world. In Mappes, T. and
DeGrazia, D. (Eds), Biomedical Ethics 5th ed. (pp. 265-269). New York: McGraw-Hill.
Antony, L. (1993). Quine as Feminist: The Radical Import of Naturalized Epistemology. In
Antony, L. & Witt, C. (Eds.). A Mind of one’s own: Feminist essays on reason &
objectivity (pp. 185-225). Boulder, Colorado: Westview Press.
Antony, L. & Witt, C. (Eds.). (1993). A Mind of one’s own: Feminist essays on reason &
objectivity. Boulder, Colorado: Westview Press.
Asquith, P. & Kitcher, P. (Eds.). Proceedings of the Philosophy of Science Association 1984
(pp. 489-524). East Lansing: Philosophy of Science Association.
Bahls, C. & Schachter, B. (2002, March 4). Interdisciplinary research gets formal. The
Scientist. 16 (5), 17.
230 References
Baier, A. (1993). Hume: The Reflective Women’s Epistemologist? In Antony, L. & Witt, C.
(Eds.). (1993). A Mind of one’s own: Feminist essays on reason & objectivity (pp. 35-48).
Boulder, Colorado: Westview Press.
Balka, E. (1986). Calculus and coffee cups: Learning science on your own. Resources
for Feminist Research, 16 (3), 11-12.
------. (1998). Feminist technology assessment: Reflections on theory and practice.
Atlantis, 22 (2), 112-122.
Balka, E., & Smith, R. (Eds.). (2000). Women, work and computerization: Charting a
course to the future. Boston: Kluwer Academic Publishers.
Baltes, P. & Staudinger, U. (Eds.). (1996). Interactive minds : life-span perspectives on the
social foundation of cognition. Cambridge, New York: Cambridge University Press.
Bannan, J. (1967). The philosophy of Merleau-Ponty. New York: Harcourt, Brace & World,
Inc.
Barinaga, M. (1993, April). Feminists find gender everywhere in science. Science, 260,
392-3.
------. (1993b, April). Is there a 'Female Style' in science? Science, 260, 384-91.
Barral, M. (1965). Merleau-Ponty: The role of the body-subject in interpersonal relations.
Pittsburgh, Pennsylvania: Duquesne University Press.
Barnette, R. (1996). Review of What computers still can’t do: A critique of artificial intelligence
by Hubert Dreyfus (1993). Retrieved from the World Wide Web 1 Feb 1996,
www.valdosta.peachnet.edu:80/~rbarnett/phi/dreyfus.html .
Becker, B. (1997). To be in Touch or not? Some Remarks on Communication in Virtual
Environments. In Grundy, F., Kohler, D., Oechtering, V., & Petersen, T. (Eds). Women,
231 References
work and computerization: Spinning a web from past to future. Proceedings 6th IFIP WG
9.1 conference. (pp. 211-2). Berlin: Springer-Verlag.
Begley, S. (1997, April 21). The science wars. Newsweek, pp. 54-7.
Benditt, J. (Ed.). (1993, April). Gender and the culture of science: Women in science
'93. Science, 260, 383-432.
Berman, H. (1975). Restitution: A review of Rosalind Franklin and DNA. In Zucker, A. (Ed).
(1996). Introduction to the philosophy of science (pp. 383-384). Upper Saddle River:
Prentice Hall.
Björkman, C. (2002). Challenging canon: The gender question in computer science. (Licentiate
Dissertation, Blekinge Institute of Technology, Karlskrona, Sweden, 2002).
Björkman, C. & Trojer, L. (2001). Computer science and its paradigmatic basis—broadening
understandings through gender research from within. In Björkman, C. (2002).
Challenging canon: The gender question in computer science (pp. 77-94). (Licentiate
Dissertation, Blekinge Institute of Technology, Karlskrona, Sweden, 2002).
Bloom, L. (1994). Constructing whiteness: Popular science and National Geographic in the age
of multiculturalism. Configurations, 1, 15-32.
Boden, M. (Ed.). (1977). Artificial intelligence and natural man. New York: Basic Books, Inc.,
Publishers.
------. (1990). The philosophy of artificial intelligence. Oxford: Oxford
University Press.
------. (1995, Winter). “AI’s Half-Century.” AI Magazine, 96-99. Retrieved from the World
Wide Web May 28, 2001: www.aaai.org .
------. (1996). The philosophy of artificial life. Oxford: Oxford University
232 References
Press.
Boom, R. A. , Chavez-Oest, J. A., and Boom, T. (1995). The Importance of Local Data Bases
in Medical Expert Systems: TICITL. MEDINFO 95 Proceedings (pp. 873-875).
Borg, A. & Whitney, T. (2002, June). The Grace Hopper Celebration. inroads (the SIGCSE
Bulletin) 34:2, 181-2.
Borgmann, A. (2000). Information, nearness, and farness. In Goldberg, K., (Ed). The robot in
the garden: Telerobotics and telepistemology in the age of the internet (pp. 91-107).
Cambridge, Mass: The MIT Press.
Bradley, D. (2000). Too few at the top: Women in science. HMSBeagle: The BioMedNet
Magazine. Issue 90 (http://news.bmn.com/hmnbeagle/90/notes/adapt?print=yes ) (10
Nov).
Braver, R. (1999). Tuning into science. CBS News. Retrieved from the World Wide Web
March 15, 1999: www.cbs.com/common/flat/flat_story_135320.html
------. (1999b). Equal lab time for girls. CBS News. Retrieved from the World Wide Web
March 15, 1999: www.cbs.com/common/flat/flat_story_135264.html
Brenkert, G. (1998). Marketing to Inner-city blacks: PowerMaster and moral responsibility. In
Hoffman, W., Frederick, R., & Schwartz, M. (Eds). (2001). Business Ethics: Readings and
Cases in Corporate Morality, 4th ed. (pp.394-403). New York: McGraw-Hill.
Brown, E. (2001). Distributive justice and the impact of emotional responses. (Master’s Thesis.
University of Cincinnati).
Brown, J. (2001). Who rules in science? An opinionated guide to the wars. Cambridge, Mass.:
Harvard University Press.
233 References
Brown, K. (2000, February 18). A new breed of scientist-advocate emerges. Science 287, 1192-
1195.
Brown, L. (1994). Subversive dialogues: Theory in feminist therapy. New York: Basic Books.
Brooks, R. (2003). Homepage of Rodney Brooks.
http://www.ai.mit.edu/people/brooks/index.shtml [visited 13 March 2003].
------. (1991). Intelligence without representation. In Haugeland, J. (Ed). Mind design II:
Philosophy, psychology, artificial intelligence, revised ed. (pp.395-420). Cambridge,
Massachusetts: The MIT Press.
Buchanan, E. A. (2000). Strangers in the 'Myst' of video gaming: Ethics and
representation. The CPSR Newsletter, 18 (1).
http://www.cpsr.org/publications/newsletters/issues/2000/Winter2000/index.html
Bush, C. G. (1983). Women and the assessment of technology: To think, to be; to unthink, to
free. In Rothschild, J. (Ed.), Machina ex dea: Feminist perspectives on technology. New
York: Pergamon.
Cajete, G. (2000). Native science: natural laws of interdependence. Santa Fe: Clear Light
Publishers.
------. (2001, Spring). Native science assists a science major or graduate program: An
Interview with Dr. Gregory Cajete. APA newsletter, 00 (2), 18-20.
Calandra, B. (2001, Nov 26). The science of collaboration. The Scientist. 15, (23), 34.
Camp. T. (2002, June). The incredible shrinking pipeline. inroads (the SIGCSE Bulletin) 34:2,
129-134.
234 References
Campanella, T. (2000). Eden by wire: Webcameras and the telepresent landscape. In Goldberg,
K., (Ed). The robot in the garden: Telerobotics and telepistemology in the age of the
internet (pp. 25-47). Cambridge, Mass: The MIT Press.
Caplan, A. (1996). Exemplary reasoning? A comment on theory structure in biomedicine. In
Zucker, A. (Ed). Introduction to the philosophy of science (pp. 265-273). Upper Saddle
River: Prentice Hall.
Carson, R. (1962). Silent Spring. New York: Houghton Mifflin Company.
Chhibber, S., Apostolakis, G. E., & Okrent, D. (1994, Jan). On the use of expert judgments to
estimate the pressure increment in the Sequoyah containment at Vessel breach. Nuclear
Technology, 105.
Churchland, P. M. & Churchland, P. S. (1998). On the contrary: critical essays, 1987-1997.
Cambridge, Mass: MIT Press.
Clark, A. (1997). Being there: putting brain, body, and world together again. Cambridge,
Mass.: MIT Press.
------. (2001). Where brain, body, and world collide. In Edelman, G. & Changeux, Jean-Pierre
(Eds.) The Brain (pp. 257-280). New Brunswick, New Jersey: Transaction Publishers.
Collins, H. M. (1990). Artificial experts: Social knowledge and intelligent machines.
Cambridge: MIT Press.
Code, L. (1993). Taking subjectivity into account. In Alcoff, L., & Potter, E. (Eds.). Feminist
epistemologies (pp. 15-48). New York: Routledge.
Cohn, C. (1996). Sex and death in the rational world of defense intellectuals. In Laslett, B. et al
(Eds). Gender and scientific authority (pp. 183-215). Chicago: The University of Chicago
Press.
235 References
------. (1996b). Nuclear language and how we learned to pat the bomb. In Keller, E. F., &
Longino, H. (Eds.). Feminism and science (173-186). New York: Oxford
University Press.
Cohen, J. (2000, Jan 14). Designer labs: Architecture discovers science. Science, 287, 210-212.
Cohoon, J. (2002, June). Recruiting and retaining women in undergraduate computing majors.
inroads (the SIGCSE Bulletin) 34:2, 48-52.
Comfort, N. (2001). The Tangled Field: Barbara McClintock's Search for the Patterns of
Genetic Control. Cambridge, Mass: Harvard University Press.
Conboy, K., Medina, N., & Stanbury, S. (Eds). (1997). Writing on the body: Female
embodiment and feminist theory. New York: Columbia University Press.
Conley, F. K. (1998). Walking out on the boys. New York: Farrar, Straus and Giroux.
Craig, A., Fisher, J., Scholary, A. & Venables, A. (2000). Women Supported Collaborative
Work. Supplement to Balka, E., & Smith, R. (Eds.). (2000). Women, work and
computerization: Charting a course to the future. Boston: Kluwer Academic Publishers.
Cuny, J. & Aspray, W. (2002, June). Recruitment and retention of women graduate students in
computer science and engineering: Results of a workshop organized by the computing
research association. inroads (the SIGCSE Bulletin) 34:2, 168-174.
Cuomo, C. (1998). Feminism and ecological communities: An ethic of flourishing. New York:
Routledge.
Curd, M. & Cover, J.A. (1998). Philosophy of science: The central issues. New York: W. W.
Norton & Company.
Dalmiya, V. & Alcoff, L. (1993). Are “old wives’ tales” justified? In Alcoff, L., & Potter, E.
(Eds.). (pp. 217-244). Feminist epistemologies. New York: Routledge.
236 References
Dautenhahn, K. (1997). The Physical Body in Cyberspace: At the Edge of Extinction? In
Grundy, F., Kohler, D., Oechtering, V., & Petersen, T. (Eds). Women, work and
computerization: Spinning a web from past to future. Proceedings 6th IFIP WG 9.1
conference. (pp. 213-6). Berlin: Springer-Verlag.
Davies,V., & Camp, T. (2000). Where have women gone and will they be returning:
Predictions of female involvement in computing. The CPSR Newsletter, 18 (1),
http://www.cpsr.org/publications/newsletters/issues/2000/Winter2000/index.html
Davion, V. (1987, Summer). Do good feminist compete? Hypatia 2:2, 55-63.
Davis, E. (1990). Representations of commonsense knowledge. San Mateo: Morgan
Kaufmann Publishers.
Donath, Judith. (1997). Body language without the body: Situating social cues in the virtual
world. In Grundy, F., Kohler, D., Oechtering, V., & Petersen, T. (Eds). Women, work and
computerization: Spinning a web from past to future. Proceedings 6th IFIP WG 9.1
conference. (pp. 217-8). Berlin: Springer-Verlag.
Dooher, K. (2002, April). Lord of the robots [Interview with Rodney Brooks]. Technology
Review 105: 3, 78-83.
Dreyfus, H. (1972). What computers can’t do: The limits of artificial intelligence.
New York: Harper & Row.
------. (1979 revised edition). What computers can’t do: The limits of artificial intelligence.
New York: Harper & Row, Publishers, Inc.
------. (1992). What computers still can’t do: A critique of artificial reason (revised
ed.). Cambridge: MIT Press.
237 References
------. (2000). Telepistemology: Descartes’s last stand. In Goldberg, K., (Ed). The robot in the
garden: Telerobotics and telepistemology in the age of the internet. (pp. 48-63).
Cambridge, Mass: The MIT Press.
Dreyfus, H. & Dreyfus, P. (1964). Introduction to Sense and non-sense. In Merleau-Ponty, M.
(1964). Sense and non-sense. Translated by H. Dreyfus & P. Dreyfus. Chicago:
Northwestern University Press, ix-xxvii.
Dreyfus, H. & Dreyfus, S. (1986). From Socrates to expert systems: The limits of calculative
rationality. In Mitcham, C. & Huning, A. (Eds.). Philosophy and Technology II:
information technology and computers in theory and practice, (pp. 111-130). Boston:
Kluwer.
Dworkin, A. (1997). Life and death: Unapologetic writings on the continuing war against
women. New York: The Free Press.
Economist Technology Quarterly (2002, Mar 16). AI by another name. Economist 362: 8264,
22-25.
Edelman, G. & Changeux, J. (Eds.). (2001). The Brain. New Brunswick, New Jersey:
Transaction Publishers.
Edwards, R. (2002, 27 Dec). Group Claims First Human Clone. Morning Edition, National
Public Radio. Audio on-line retrieved 28 Dec 2002 at
http://discover.npr.org/features/feature.jhtml?wfId=892667
Eisenhart, M., & Finkel, E., with Behm, L., Lawrence, N., & Tonso, K. (1998).
Women’s science: Learning and succeeding from the margins. Chicago: The
University of Chicago Press.
238 References
Enns, C. (1997). Feminist theories and feminist psychotherapies: Origins, themes, and
variations. New York: Haworth.
Erb, U. (1997). Exploring the excluded: A feminist approach to opening new perspectives in
computer science. In Grundy, F., Kohler, D., Oechtering, V., & Petersen, T. (Eds).
Women, work and computerization: Spinning a web from past to future. Proceedings 6th
IFIP WG 9.1 conference. Berlin: Springer-Verlag.
Estrich, S. (1987). Real rape. Cambridge: Harvard University Press.
Eubanks, V. (2000). Paradigms and perversions: A women's [sic] place in cyberspace.
The CPSR Newsletter, 18 (1).
http://www.cpsr.org/publications/newsletters/issues/2000/Winter2000/index.html
Fales, E. (1996, September). Plantinga's case against naturalistic epistemology.
Philosophy of Science, 63, 432-451.
Farris, J., & Platnick, N. (1989). Review of the book Science as a process: An evolutionary
account of the social and conceptual development of science by David Hull. Cladistics, 5,
295-310.
Feibleman, J. K. (1966). Pure science, applied science, and technology: An attempt at
definitions. In Mitcham, C. & R. Mackey (Eds.). (1983). Philosophy and
technology: Readings in the philosophical problems of technology (pp. 33-41). New York:
The Free Press.
Fetzer, J. H. (1993). Philosophy of Science. New York: Paragon House.
Feyerabend, P. (1963). How to be a good empiricist—A plea for tolerance in matters
epistemological. In Curd, M. & Cover, J.A. (1998). Philosophy of science: The central
issues (922-949). New York: W. W. Norton & Company.
239 References
------. (1988). Against method, rev. ed. New York: Verso.
Fisher, A. & Margolis, J. (2002, June). Unlocking the clubhouse: The Carnegie Mellon
experience. inroads (the SIGCSE Bulletin) 34:2, 79-83.
Fine, A. (1998). The viewpoint of no-one in particular. Proceedings of the American
Philosophical Association, 72, 2, 9-20.
Fiske, D. (2000, Aug 2). New Approaches to Laboratory Design. Architecture Week.
http://www.architectureweek.com/2000/0802/design_1-2.html
Flax, J. (1996). Psychoanalysis and the philosophy of science: Critique or resistance? In Zucker,
A. (Ed) (1996). Introduction to the philosophy of science (322-329). Upper Saddle
River: Prentice Hall.
Freedman, B. (1987). Equipoise and the ethics of clinical research. In Mappes, T. and DeGrazia,
D. (Eds). (2001). Biomedical Ethics 5th ed. (pp. 259-265). New York: McGraw-Hill.
French, M. (1981). Shakespeare’s Division of Experience. (New York: Simon & Schuster), 21-
33.
Fricker, M. & Hornsby, J. (Eds). (2000). The Cambridge companion to feminism in philosophy.
New York: Cambridge University Press.
------. (2000b). Introduction. In Fricker, M. & Hornsby, J. (Eds). (2000). The Cambridge
companion to feminism in philosophy (pp. 1-9). New York: Cambridge University Press.
Frye, M. (1983). Some reflections on separatism and power. The politics of reality: Essays in
feminist theory. Trumansburg, NY: The Crossing Press.
Gabbert, P. & Meeker, P. (2002, June). Support communities for women in computing. inroads
(the SIGCSE Bulletin) 34:2,62-65.
240 References
Galpin, V. (2002, June). Women in computing around the world. inroads (the SIGCSE Bulletin)
34:2, 94-100.
Gary, A. (2001, Spring). Ms. feminist philosopher’s manners’s guide to respectful critical
sisterly behavior. APA Newsletters 00, 2, 88-9.
Gender Working Group, UN Commission on Science and Technology for Development. (1995).
Recommendations to the United Nations. On-line at gab.wigsat.org/recomm.htm [visited
14 October 2001].
Gelven, M. (1970). A commentary on Heidegger’s “Being and Time”. New York: Harper &
Row, Publishers.
Gigerenzer, G. (1996). In Baltes, P. & Staudinger, U. (Eds.). Interactive minds: life-span
perspectives on the social foundation of cognition (pp. 319-46). Cambridge,
Massachusetts: Cambridge University Press.
------. (1999). How good are fast and frugal heuristics? In Shanteau, J., Mellers, B., & Schum,
D. (Eds.). Decision science and technology: reflections on the contributions of Ward
Edwards (pp. 81-104). Boston, Mass.: Klumer Academic,
------. (2000). Adaptive thinking: rationality in the real world. New York: Oxford University
Press
Gilligan, C. (1982). In a different voice: Psychological theory and women's
development. Cambridge: Harvard University Press.
Ginzberg, R. (1987). Uncovering gynocentric science. In Zucker, A. (Ed). (1996). Introduction
to the philosophy of science (pp. 364-375). Upper Saddle River: Prentice Hall.
Glazebrook, T. (2000). Heidegger’s Philosophy of Science. New York: Fordham University
Press.
241 References
Gleick, J. (1987). Chaos: Making a new science. New York: Penguin Books.
Glymour, C. (1992). Android expistemology: Computation, artificial intelligence and
the philosophy of science. In Salmon, M., Earman, J., Glymour, C. et al. Introduction to
the philosophy of science (pp. 364-403). Englewood Cliffs: Prentice Hall.
Goldberg, J. (1989). Anatomy of a scientific discovery: The race to discover the secret of
human pain and pleasure. London: Bantham Books.
Goldberg, K., (Ed). (2000). The robot in the garden: Telerobotics and telepistemology in the
age of the internet. Cambridge, Mass: The MIT Press.
------. (2000b). Introduction: The unique phenomenon of a distance. In Goldberg, K., (Ed).
The robot in the garden: Telerobotics and telepistemology in the age of the internet (pp. 1-
10). Cambridge, Mass: The MIT Press.
------. (2000c). Telerobotic knowledge: A reliabilist approach. In Goldberg, K., (Ed). The
robot in the garden: Telerobotics and telepistemology in the age of the internet (pp.127-
142). Cambridge, Mass: The MIT Press.
Goldberger, N. (1996). Introduction: Looking backward, looking forward. In Goldberger, N.,
Tarule, J., Clinchy, B., & Belenky, M. (Eds.). Knowledge, difference, and power: Essays
inspired by Women’s Way of Knowing (pp. 1-24). New York, NY: Basic Books.
Goldberger, N., Tarule, J., Clinchy, B., & Belenky, M. (Eds.). (1996). Knowledge, difference,
and power: Essays inspired by Women’s Way of Knowing. New York, NY: Basic Books.
Gorovitz, S. & MacIntyre, A. (1996). Toward a theory of medical fallibility. In Zucker, A. (Ed)
Introduction to the philosophy of science (pp. 246-254). Upper Saddle River: Prentice Hall.
242 References
Grady, D. (2003, Feb 25). 50 years later, Rosalind Franklin’s X-ray fuels debate. New York
Times. On-line at http://www.nytimes.com/2003/02/25/science/25FRAN.html [visited 25
Feb 03].
Greene, J. & Haidt, J. (2002, Dec). How (and where) does moral judgment work? TRENDS in
Cognitive Sciences, 6(12), 517-523. Retrieved from the World Wide Web 18 Jan 2003:
http://news.bmn.com/magazine/article?pii=S1364661302020119
Grene, M. (1957). Martin Heidegger. London: Bowes & Bowes.
Grene, M. (1969). Introduction. In Polanyi, M. Knowing and Being (pp. ix-xvii). Chicago:
The University of Chicago Press.
Grimwade, A. (2002, Feb 18). Open Societies Need Open Access. The Scientist 16[4]:10.
Grossman, W. (1998, August). Access denied. Scientific American, 38.
Grosz, E. (1993). Bodies and knowledge: Feminism and the crisis of reason. In Alcoff, L., &
Potter, E. (Eds.). (1993). Feminist epistemologies. New York: Routledge.
Grundy, F., Kohler, D., Oechtering, V., & Petersen, T. (Eds). (1997). Women, work and
computerization: Spinning a web from past to future. Proceedings 6th IFIP WG 9.1
conference. Berlin: Springer-Verlag.
Guha, R. V., & Lenat, D. B. (Fall 1990). CYC: A midterm report. AI Magazine, 11(3), 32-59.
------. (1994, July). Enabling agents to work together. _Communications of the ACM, 37(7),
127-142.
Gurak, L. (2001, Feb). ‘Is this the party to whom I am speaking?’: Women, credibility and the
Internet. The Women’s Review of Books, XVIII, 5, 5-6.
Gürer, D. & Camp, T. (2002, June). An ACM-W literature review on women in computing.
inroads (the SIGCSE Bulletin) 34:2, 121-127.
243 References
Gürer, D. & Cuny, J. (2002). Recruiting and retaining women in computer science: Where are
we? Where do we go from here? Proceedings of Grace Hopper Celebration of Women in
Computing Conference CD [compact disk]. Palo Alta, CA: The Institute for Women and
Technology.
Hacking, I. (1999). The social construction of what? Cambridge: Harvard University Press.
Håpnes, T. & Sørensen, K. (1995). Competition and collaboration in male shaping of
computing: A study of Norwegian hacker culture. In Gill, R. & Grint, K. (Eds.). The
Gender-Technology Relation: Contemporary Theory and Research (pp. 174-191). Bristol,
PA: Taylor & Francis Inc.
Haraway, D. (1985). Primatology is politics by other means. In Asquith, P. & Kitcher, P.(Eds.),
Proceedings of the Philosophy of Science Association 1984 (pp. 489-524). East Lansing:
Philosophy of Science Association.
------(1997). The persistence of vision. In Conboy, K., Medina, N., & Stanbury, S. (Eds).
Writing on the body: Female embodiment and feminist theory (pp. 283-295). New York:
Columbia University Press.
Harding, S. (1986). The science question in feminism. Ithaca: Cornell University Press.
------. (1991). Whose science? Whose knowledge? Thinking from women's lives.
Ithaca: Cornell University Press.
------. (1996). Multicultural and global feminist philosophies of science: Resources and
challenges. In Nelson, L. H., & Nelson, J. (Eds.). Feminism, science, and the philosophy
of science (263-287). Norwell, MA: Kluwer Academic Publishers.
------. (1998). Is science multi-cultural? Postcolonialisms, feminisms, and epistemologies.
Indianapolis: Indiana University Press.
244 References
Harding, S. (Ed). (1993). The ‘racial’ economy of science: Toward a democratic future.
Indianapolis: Indiana University Press.
Harding, S. & M. B. Hintikka (Eds) (1983). Discovering reality: feminist perspectives on
epistemology, metaphysics, methodology, and philosophy of science. Boston: Dordrecht.
Harris, R. (2000). Squeezing librarians out of the middle: Gender and technology in a
threatened profession. In Balka, E., & Smith, R. (Eds.). Women, work and
computerization: Charting a course to the future (pp. 250-259). Boston: Kluwer Academic
Publishers.
Harvey, E. & Okruhlik, K (Eds). Women and Reason. Ann Arbor, Michigan: University of
Michigan Press.
Haugeland, J. (Ed). (1998). Having thought: essays in the metaphysics of mind. Cambridge,
Mass: Harvard University Press.
------. (1998b) “Mind Embodied and Embedded” In Haugeland, J. Having thought: essays in the
metaphysics of mind. Cambridge, Mass: Harvard University Press.
------. (2000). Mind design II: Philosophy, psychology, artificial intelligence, revised ed.
Cambridge, Mass: The MIT Press.
Hearst, M. & Hirsh, H. (2000). “AI’s greatest trends and controversies.” IEEE intelligent
systems & their applications. Retrieved from the World Wide Web 28 May 2001:
www.aaai.org/Pathfinder/html/overview.html .
Heidegger, M. (1977). The question concerning technology and other essays. (W. Lovitt,
Trans.). New York: Harper & Row, Publishers.
Herring (2000) Gender differences in CMC: Findings and implications. The CPSR Newsletter,
18 (1). http://www.cpsr.org/publications/newsletters/issues/2000/Winter2000/index.html
245 References
Hert, P. (1997). Social dynamics of an on-line scholarly debate. The Information
Society, 13, 329-360.
Hiltzik, M. (2002, March). A.I. Reboots. Technology Review 105:2, 46-55.
Hochschild, A. (1997). The Second Shift. Avon Books.
Hoffman, W., Frederick, R., & Schwartz, M. (Eds). (2001). Business Ethics: Readings and
Cases in Corporate Morality, 4th ed. New York: McGraw-Hill.
Holmstrom, N. (1999). Do women have a distinct nature? In Pearsall, M. (Ed). Women and
Values: Readings in recent feminist philosophy, 3rd ed. (pp. 50-59). Cincinnati: Wadsworth
Publishing Company.
Holton, G. (1998, April 27). Different perceptions. Scientific American. On-line at:
www.sciam.com/explorations/1998/051898women/gender.html [visited 14 October 2001].
Homiak, M. (1993). Feminism and Aristotle’s rational ideal. In Antony, L. & Witt, C. (Eds.). A
Mind of one’s own: Feminist essays on reason & objectivity (pp. 2-17). Boulder,
Colorado: Westview Press.
Honderich, T. (Ed). (1995). The Oxford Companion to Philosophy. New York: Oxford
University Press.
Howard, K. (2000, July). The Bioinformatics Good Rush. Scientific American, 58-63.
Huff, C. (2002, June). Gender, software design, and occupational equity. inroads (the SIGCSE
Bulletin) 34:2, 112-115.
Hull, D. (1988). Science as a process: An evolutionary account of the social and
conceptual development of science. Chicago: University of Chicago Press.
------. (1997). What's wrong with invisible-hand explanations? Philosophy of
Science, 64, S117-S126.
246 References
Inwood, M. (1995) “Heidegger, Martin” in Honderich, T. (Ed). The Oxford Companion to
Philosophy. New York: Oxford University Press.
Jagger, A. (1983). Feminist politics and human nature. Totowa, NJ: Rowman
& Littlefield Publishers.
------. (1989). “The myth of dispassionate investigation” In Zucker, A. (Ed). (1996).
Introduction to the philosophy of science (pp. 352-362). Upper Saddle River: Prentice
Hall.
Joeres, R. & Laslett, B. (Eds). The Second Signs reader: feminist scholarship, 1983-1996.
Chicago: University of Chicago Press.
Johnson, D. & Miller, K. (2002, June). Is diversity in computing a moral matter? inroads (the
SIGCSE Bulletin) 34:2, 9-10.
Johnson, D. & Nissenbaum, H. (eds). (1995). Computers, ethics & social values. Englewood
Cliffs, NJ: Prentice Hall.
Johnson, M. (1999). Embodied Reason. In Weiss, G. & Haber, H. (Eds.). Perspectives on
Embodiment (pp. 81-102). New York: Routledge.
Jonietz, E. (2001, April). Population, Inc. Technology Review, 50-55.
Kahneman, D. Slovic, P. & Tversky, A. (Eds.) (1982). Judgment under uncertainty: Heuristics
and biases. New York: Cambridge University Press.
Kaiser, J. (2000, February 18). Ecologists on a mission to save the world. Science 287, 1188-
1192.
Kantrowitz, B. (1994, May 16). Men, women & computers. Newsweek, 48-55.
Keller, E. F. (1983). A feeling for the organism: The life and work of Barbara
McClintock. New York: Freeman.
247 References
------. (1985). Reflections on gender and science. New Haven: Yale University Press.
------. (1996). Language and ideology in evolutionary theory: Reading cultural norms into
natural law. In Keller, E. F., & Longino, H. (Eds.). Feminism and science (154-172).
New York: Oxford University Press.
------. (2000). The century of the gene. Cambridge: Harvard University Press.
Keller, E. F., & Longino, H. (Eds.). (1996). Feminism and science. New York: Oxford
University Press.
Kershner, R. (ed.). (1993). A Portrait of the Artist as a Young Man. (New York: Bedford Books),
296-7.
Kilbourne, J. (2000). Killing us softly 3. Videorecording. Northampton, MA : Media Education
Foundation.
Kim, J. (1998). What is ‘Naturalized Epistemology’? In Alcoff, L. (Ed). Epistemology: The Big
Questions (265-284). Malden, Mass.: Blackwell Publishers.
Kitcher, P. (1993). The advancement of science: Science without legend, objectivity
without illusions. New York: Oxford University Press.
Knoppers, B. (2002, January 18). Genetic Information and the Family. In HMS Beagle: The
BioMedNet Magazine (http://news.bmn.com/hmsbeagle/current/notes/feature1), Issue 118.
Koertge, N. (Ed). (1998). A house built on sand: exposing postmodernist myths about science.
New York: Oxford University Press.
Kohlstedt, S., & Longino, H. (Eds.). (1997). Osiris: Women, Gender, and Science, 12.
Kolata, G. and Petersen, M. (2002, July 10). Hormone replacement study a shock to the medical
system. The New York Times.
248 References
http://www.nytimes.com/2002/07/10/health/10HORM.html?todaysheadlines [Visited 10
July and 8 October 2002].
Kornblith, H. (1994). A conservative approach to social epistemology. In Schmitt, F. F. (Ed).
Socializing epistemology: The social dimensions of knowledge (pp. 93-110). Lanham, MD:
Rowman & Littlefield, Publishers.
Korsmeyer, C. (1977). The hidden joke: Generic uses of masculine terminology. In Vetterling-
Braggin, M., Elliston, F., & English, J. (Eds). Feminism and Philosophy (pp. 138-153).
Totowa, NJ: Rowman and Littlefield.
Krell, D. F. (Ed). (1977). Martin Heidegger: Basic Writings from Being and Time (1927) to
The Task of Thinking (1964). New York: Harper & Row, Publishers.
Kuhn, T. (1970). The structure of scientific revolutions (2nd ed.). Chicago: University of
Chicago Press.
Kuosa, T. (2000). Masculine world disguised as gender neutral. In Balka, E., & Smith, R.
(Eds.). Women, work and computerization: Charting a course to the future (pp. 119-126).
Boston: Kluwer Academic Publishers.
Kwant, R. (1963). The phenomenological philosophy of Merleau-Ponty. Pittsburgh,
Pennsylvania: Duquesne University Press.
Lakoff, G. & Johnson, M. (1999). Philosophy in the flesh: The embodied mind and its challenge
to Western thought. New York: Basic Books.
Lanius, C. (2002). Girls and Technology: Where Do We Stand in 2002? Proceedings of Grace
Hopper Celebration of Women in Computing Conference CD [compact disk]. Palo Alta,
CA: The Institute for Women and Technology.
249 References
Laslett, B. et al (Eds). (1996). Gender and scientific authority. Chicago: The University of
Chicago Press.
Latour, B. & Woolgar, S. (1979). Laboratory life: the social construction of scientific facts.
Beverly Hills: Sage Publications.
Laudan, L. (1984). Science and values: The aims of science and their role in scientific
debate. Berkeley: University of California Press.
Laurence, L. & Weinhouse, B. (1994). Outrageous practices: The alarming truth about how
medicine mistreats women. New York: Fawcett Columbine.
Lazowska, E. (2002, June). Pale and male: 19th century design in a 20th century world. inroads
(the SIGCSE Bulletin) 34:2, 11-12.
Lenat, D. B. (1995a). Steps to sharing knowledge. In Mars, N.J. I. (Ed.), Towards very large
knowledge bases: knowledge building & knowledge sharing. Amsterdam, The
Netherlands: IOS Press.
------. (1995b, September). Artificial intelligence. Scientific American, 80-82.
------. (1995c, November). CYC: A large-scale investment in knowledge
infrastructure. Communications of the ACM, 38 (11), 33-38.
Lenat, D. B. & Feigenbaum, E. A. (1991). On the thresholds of knowledge. Artificial
Intelligence, 47, 185-250.
Lenat, D. B., & Guha, R. V. (1990). Building large knowledge-based systems:
Representation and inference in the CYC project. New York: Addison-Wesley
Publishing Company.
------. (1994, July). Enabling agents to work together. Communications of the ACM 37:7.
250 References
Lenat, D., Miller, G., & Yokoi, T. (1995, Nov). CYC, WordNet, and EDR: Critiques and
Responses. Communications of the ACM, 38, 11, 45-48.
Leopold, A. (1949). The land ethic. In Pierce, C. & VanDeVeer, D. (Eds). (1995). People,
penguins, and plastic trees: basic issues in environmental ethics, 2nd ed. (pp. 142-151).
Cincinnati: Wadsworth Publishing Company.
Lewis, R. (2002, Feb 18). Race and the Clinic: Good Science? The Scientist 16[4]:16.
http://www.the-scientist.com/yr2002/feb/lewis_p16_020218.html [visited 30 Sept 2002].
Liu, M.L. & Blanc, L. (1996). On the retention of female computer science students.
Proceedings of the 1996 ACM SIGUCCS, 32-35.
Lloyd, G. (1983). Public reason and private passion. Metaphilosophy, 14, 308-26.
Longino, H. (1987). The ideology of competition. In Miner, V. & Longino, H. (Eds.).
Competition: A Feminist Taboo? (pp. 248-258). New York: The Feminist Press at the
City University of New York.
------. (1987b). Can there be a feminist science? In Zucker, A. (Ed). (1996). Introduction to the
philosophy of science (pp. 377-381). Upper Saddle River: Prentice Hall.
------. (1990). Science as social knowledge: Values and objectivity in scientific
inquiry. Princeton: Princeton University Press.
------. (1994). The fate of knowledge is social theories of science. In Schmitt, F. F. (Ed).
Socializing epistemology: The social dimensions of knowledge (pp. 135-157). Lanham,
MD: Rowman & Littlefield, Publishers.
------. (1995). To See Feelingly: Reason, Passion, and Dialogue in Feminist Philosophy. In
Stanton, D. & Stewart, A., (Eds). Feminisms in the Academy (pp.19-45). Ann Arbor:
University of Michigan Press.
251 References
------. (1996). Cognitive and non-cognitive values in science: Rethinking the Dichotomy. In
Nelson, L. H., & Nelson, J. (Eds.). Feminism, science, and the philosophy of science (pp
39-58). Norwell, MA: Kluwer Academic Publishers.
------. (1997). Comments on science and social responsibility: A role for
philosophy of science? Philosophy of Science, 64, S175- S179.
------. (2002). The fate of knowledge. Princeton: Princeton University Press.
Lovitt, W. (1977). “Introduction.” In Heidegger, M. (1977). The question concerning
technology and other essays. (W. Lovitt, Trans.). New York: Harper & Row,
Publishers.
Lugones, M. (1987). Playfulness, ‘World’-Travelling, and Loving Perception. Hypatia 2:2.
New York: Routledge.
------. (1996). Purity, Impurity, and Separation. In Joeres, Ruth-Ellen & Barbara Laslett (Eds).
The Second Signs reader: feminist scholarship, 1983-1996 (pp 275-296). Chicago:
University of Chicago Press.
MacKinnon, C. (1989). Toward a feminist theory of the state. Harvard University Press.
Mahowald, M. (2000). Genes, Women, Equality. New York: Oxford University Press.
Malpas, J. (2000). Acting at a distance and knowing from afar: Agency and knowledge on the
Internet. In Goldberg, K., (Ed). The robot in the garden: Telerobotics and
telepistemology in the age of the internet (pp. 109-124). Cambridge, Mass: The MIT
Press.
Mandavilli, A. (2001, Oct 12). Huge salary survey confirms gender gap in biology. In
BioMedNet News. (http://news.bmn.com/news/story?day=011012&story=2) [visited 14
Oct 01].
252 References
Mappes, T. and DeGrazia, D. (Eds). (2001). Biomedical Ethics 5th ed. New York: McGraw-Hill.
Mark, G. (1997). Flying Through Walls and Virtual Drunkenness: Disembodiment in
Cyberspace? In Grundy, F., Kohler, D., Oechtering, V., & Petersen, T. (Eds). (1997).
Women, work and computerization: Spinning a web from past to future. Proceedings 6th
IFIP WG 9.1 conference (pp. 221-2). Berlin: Springer-Verlag.
Mars, N.J. I. (Ed.), Towards very large knowledge bases: knowledge building & knowledge
sharing. Amsterdam, The Netherlands: IOS Press.
Martin, E. (1991). The egg and the sperm: How science has constructed a romance based on
stereotypical male-female roles. Signs: Journal of women in culture and society, 16.
Massachusetts Institute of Technology [MIT]. (1999). A Study on the status of women faculty in
science at MIT. Retrieved April 18, 1999 from the World Wide Web:
web.mit.edu/fnl/women/women.pdf .
Matyas, M., & Dix, L. (Eds.). (1992). Science and engineering programs: On target for
women? Washington, DC: National Academy Press.
Mbugua, R. (2002). Computer ubiquity - a mere fallacy to most women in Kenya. Proceedings
of Grace Hopper Celebration of Women in Computing Conference CD [compact disk].
Palo Alta, CA: The Institute for Women and Technology.
McGee, G. (2000). The perfect baby: Parenthood in the new world of cloning and genetics, 2nd
ed. New York: Rowman & Littlefield Publishers, Inc.
McIlwee, J., & Robinson, J. (1992). Women in engineering: Gender, power, and
workplace culture. Albany: State University of New York Press.
Merchant, C. (1980). The death of nature: Women, ecology and the scientific revolution.
London: Wilwood.
253 References
Merleau-Ponty, M. (1964). Sense and non-sense. Translated by H. Dreyfus & P. Dreyfus.
Chicago: Northwestern University Press.
Miner, V., & Longino, H. (Eds.). (1987). Competition: A feminist taboo? New York:
The Feminist Press at the City University of New York.
Mitcham, C. & Huning, A. (Eds.). Philosophy and Technology II , 111-130.
Mitcham, C. & Mackey, R. (Eds.). (1983). Philosophy and technology: Readings in the
philosophical problems of technology. New York: The Free Press.
Moody, S. (1999, Dec 24). The brain behind Cyc. The Austin Chronicle Screens. Retrieved
from the World Wide Web on January 11, 2000:
www.auschron.com/issues/dispatch/1999-12-23/screens_feature.html .
Moor, J. (1995). What is Computer Ethics? In Johnson, D. & Nissenbaum, H. (eds).
Computers, ethics & social values (pp. 1-15). Englewood Cliffs, NJ: Prentice Hall.
Morell, V. (1993a, April). Called 'Trimates,' three bold women shaped their field.
Science, 260, 420-425.
Morell, V. (1993b, April). Seeing nature through the lens of gender. Science, 260, 428- 9.
Moulton, J. (1977). The myth of the neutral ‘Man’. In Vetterling-Braggin, M., Elliston, F., &
English, J. (Eds). Feminism and Philosophy (pp. 124-137). Totowa, NJ: Rowman and
Littlefield.
------. (1983). A paradigm of philosophy: The adversary method. In Harding, S. & M. B.
Hintikka (Eds.). Discovering reality: feminist perspectives on epistemology, metaphysics,
methodology, and philosophy of science (149-64). Boston: Dordrecht.
Mozans, H. (1974). Woman in science, with an introductory chapter on woman’s long struggle
for things of the mind. Cambridge, Mass: The MIT Press.
254 References
Nakamura, L. (2001, Feb). Head hunting in cyberspace: Identity tourism, Asian avatars and
racial passing on the Web. Women’s Review of Books, 18: 5, 10-11.
National Science Foundation (2000). Women, Minorities, and Persons with Disabilities in
Science and Engineering. http://www.nsf.gov/sbe/srs/nsf00327/pdfstart.htm [Visited 30
Sept 02].
Negroponte, N. (2003, February). Creating a culture of ideas. Technology Review. Pp. 34-5.
[see also www.technologyreview.com ].
Nelson, L. H. (1990). Who knows? From Quine to a feminist empiricism. Philadelphia:
Temple University Press.
------. (1993). Epistemological communities. In Alcoff, L., & Potter, E. (Eds.). Feminist
epistemologies (121-159). New York: Routledge.
Nelson, L. H., & Nelson, J. (Eds.). (1997). Feminism, science, and the philosophy of
science. Norwell: Kluwer Academic Publishers.
------. (1997b). Introduction. In Nelson, L. H., & Nelson, J. (Eds.). Feminism, science, and the
philosophy of science (pp. ix-xix) Norwell: Kluwer Academic Publishers.
Newton-Smith, W. H. (Ed). (2000). A companion to philosophy of science. Malden, Mass:
Blackwell Publishers.
Nichols, B. (Winter 2001). Exploring a moral landscape: Genetic science and ethics. Hypatia,
16 (1), 45-63.
Noddings, N. (1984). Caring: A feminine approach to ethics and moral education. Berkeley:
University of California Press.
NOVA (17 April 2001). Cracking the code of life. Video and other resources available on-line
at: http://www.pbs.org/wgbh/nova/genome/ [visited 19 October 2001].
255 References
Okruhlik, K. (1994). Gender and the biological sciences. In M. Curd & J. A. Cover
(Eds.). Philosophy of science: The central issues (pp 192-209). (1998). New York: W. W.
Norton & Company.
------. (2000). Feminist Accounts of Science. In A companion to the philosophy of science.
Newton-Smith, W. H. (Ed). Malden, Mass: Blackwell Publishers.
Oldham, G. (2000). “Gender Equity in Science and Technology: Does it Matter?” Keynote
Presentation, Conference on Gender, Science and Technology, Montevideo, Uruguay. On-
line at gab.wigsat.org/oldham.html [visited 14 October 2001].
Pearl, A., Pollack, M., Riskin, E., Thomas, B., Wolf, E., & Wu, A. Becoming a computer
scientist. inroads (the SIGCSE Bulletin) 34:2, 135-143.
Pearsall, M. (Ed). (1999). Women and Values: Readings in recent feminist philosophy, 3rd ed.
Cincinnati: Wadsworth Publishing Company.
------. (1999b). Preface. In Women and Values: Readings in recent feminist philosophy, 3rd ed.
Cincinnati: Wadsworth Publishing Company.
Peiris, D. R., Gregor, P. & V, I. (2000). Women and Computing. In Balka, E., & Smith, R.
(Eds.). Women, work and computerization: Charting a course to the future. (pp. 34-41).
Boston: Kluwer Academic Publishers.
Perry, R., & Greber, L. (1990). Women and computers: An introduction. Signs: Journal
of women in culture and society, 16 (1), 74-101.
Pickering, A. (Ed.). Science as practice and culture. Chicago: The University of Chicago Press.
Pierce, C. & VanDeVeer, D. (Eds). (1995). People, penguins, and plastic trees: basic issues in
environmental ethics, 2nd ed. Cincinnati: Wadsworth Publishing Company.
256 References
Pitt, J. (2000). Thinking about technology: Foundations of the philosophy of technology. New
York: Seven Bridges Press.
Plumwood, V. (1991). Nature, self, and gender: Feminism, environmental philosophy,
and the critique of rationalism. Hypatia, 6, 3-27.
Polanyi, M. (1964). Science, Faith and Society. Chicago: The University of Chicago Press.
------. (1967). The Tacit Dimension. Garden City, New York: Anchor Books.
------. (1969). Knowing and Being. Chicago: The University of Chicago Press.
Potter, E. (1993). Gender and epistemic negotiation. In Alcoff, L., & Potter, E. (Eds.). Feminist
epistemologies (pp. 161-186). New York: Routledge.
Prey, J. & Treu, K. (2002, June). What do you say? Open letters to women considering a
computer science major. inroads (the SIGCSE Bulletin) 34:2, 18-20.
Quine, W. V. (1951). Two dogmas of empiricism. In M. Curd & J. A. Cover (1998). (Eds.),
Philosophy of Science: The central issues (pp. 280-301). New York: W. W. Norton &
Company.
------. (1969). Epistemology Naturalized. Ontological relativity and other essays. New York,
Columbia University Press.
Rahman, L. (2002). ICT training to improve health care of rural populations. Proceedings of
Grace Hopper Celebration of Women in Computing Conference CD [compact disk]. Palo
Alta, CA: The Institute for Women and Technology.
Rayl, A. (2001, November 12). Reflections from Nobel Conference XXXVII: Laureates offer
perspective on recent events and scientist’s roles in the future. The Scientist 15, 22.
(http://www.the-scientist.com/yr2001/nov/rayl_p12_011112.html, visited 19 Jan 2002).
Rhodes, R. (1986). The making of the atomic bomb. New York: Simon & Schuster.
257 References
Rich, E. & Knight, K. (1991). Artificial intelligence, 2nd ed. New York: McGraw-Hill.
Roberts, E., Kassianidou, M., & Irani, L. (2002, June). Encouraging women in computer
science. inroads (the SIGCSE Bulletin) 34:2, 84-88.
Rolston, D. W. (1988). Principles of artificial intelligence and expert systems
Development. New York: McGraw-Hill.
Rommes, E. (2000). Gendered User-Representations: Design of a digital city. In Balka, E., &
Smith, R. (Eds.). Women, work and computerization: Charting a course to the future
(pp. 137-145). Boston: Kluwer Academic Publishers.
Rosser, S. (1993). Female friendly science: Including women in curricular content and pedagogy
in science. The Journal of General Education, 42 (3), 191-220.
------. (1995). Introduction: Reaching the majority: Retaining women in the pipeline. In Rosser,
S. (Ed.). Teaching the majority: Breaking the gender barrier in science,mathematics, and
engineering (pp. 1-21). New York: Teachers College Press.
Rosser, S. (Ed.). (1995). Teaching the majority: Breaking the gender barrier in science,
mathematics, and engineering. New York: Teachers College Press.
Rothschild, J. (Ed.). (1983). Machina ex dea: Feminist perspectives on technology. New York:
Pergamon Press.
Russell, S., & Norvig, P. (1995). Artificial intelligence: A modern approach. Upper
Saddle River: Prentice Hall.
Salmon, M., Earman, J., Glymour, C., et. al. (Eds.). (1992). Introduction to the philosophy of
science. Upper Saddle River: Prentice Hall.
258 References
Sanders, J. (1995). Girls and technology: Villain wanted. In Rosser, S. (Ed). Teaching the
majority: Breaking the gender barrier in science, mathematics, and engineering (pp. 147-
159). New York: Teachers College Press.
Schacter, D. (1996). Searching for memory. New York: Basic Books.
Schiebinger, L. (1999). Has feminism changed science? Cambridge, Mass.: Harvard University
Press.
Schmitt, F. F. (Ed). (1994). Socializing epistemology: The social dimensions of knowledge.
Lanham, MD: Rowman & Littlefield, Publishers.
Schmitt, R. & Mood, T. (Eds). Alienation and social criticism. Atlantic Highlands, N.J.:
Humanities Press.
Schulman, K. A., Berlin JA, Harless W, et. al. (1999). The effect of race and sex on physicians'
recommendations for cardiac catheterization. The New England Journal of Medicine, 340
(8), 618-26.
Shaffer, N. E. (1996:Suppl). Understanding bias in scientific practice. Philosophy of
Science, 63:3:1, S89-S97.
Shah, S. (2002, July 1). Globablizing clinical research. The Nation. Retrieved from
http://www.thenation.com .
Shapiro, J. (2001, May 11). Molecular genetics and 21st century evolutionary theory: Lessons
from the past 50 years. [Talk given at 37th Annual University of Cincinnati Philosophy
Colloquium: Do levels in science relate?].
Shilts, R. (1988). And the band played on: Politics, people, and the AIDS epidemic. New
York: St. Martin’s Press.
Siegel, H. (1996:Suppl). Instrumental rationality and naturalized philosophy of science.
259 References
Philosophy of Science, 63:3:1, S116-S124.
Simonelli, R. (2001, Spring). Native science assists a science major or graduate program: An
interview with Dr. Gregory Cajete. APA Newsletters 00, 2, 18-20.
Sloman, A. (1998). “What is artificial intelligence?” Retrieved from the World Wide Web on
May 28, 2001: www.cs.bham.ac.uk/%7Eaxs/misc/aiforschools.html (link from the AAAI
web page: www.aaai.org ).
Smallwood, S. (2001, May 18). Prominent female astronomer leaves Harvard over tenure
dispute. The Chronicle of Higher Education, XLVII, 36, A14.
Sobstyl, E. (2002, Winter). Review of Richmond Campbell’s Illusions of Paradox: A Feminist
Epistemology Naturalized. Hypatia, 17, 1, 226-230.
Solomon, M. (1994). A more social epistemology. In F. F. Schmitt (Ed.), Socializing
Epistemology: The social dimensions of knowledge (pp. 217-233). New York: Rowman &
Littlefield Publishers.
------. (1997, March). Review of Edwin Hutchins' Cognition in the Wild.
Philosophy of Science, 64, 181-2.
Sonnet, G. (1998a, April 27). Where’s the Difference? Excerpted from “Choices and Successes:
Women in Science and Engineering.” Scientific American. On-Line at
(www.sciam.com/explorations/1998/051898women/impede.html )
------(1998b, April 27). Digging Deeper Excerpted from “Choices and Successes: Women in
Science and Engineering.” Scientific American. On-Line at
(www.sciam.com/explorations/1998/051898women/issues.html ).
260 References
------. (1998c, April 27). Advocating Women. Excerpted from “Choices and Successes:
Women in Science and Engineering.” Scientific American. On-Line at
(www.sciam.com/explorations/1998/051898women/arguments.html )
Spender, D. (2000). The digital life style for women. The CPSR Newsletter, 18 (1),
http://www.cpsr.org/publications/newsletters/issues/2000/Winter2000/index.html
Spertus, E. & Pine, E. (Winter 2000). Introduction: Gender in the Internet Age. The CPSR
Newsletter [Retrieved from the World Wide Web at]
http://www.cpsr.org/publications/newsletters/issues/2000/Winter2000/index.html
Stack, J., Panteli, A., Atkinson, M., & Ramsay, H. (1998). Women in computing: The ethical
responsibility of the IT industry. AISB Quarterly, 100, 53-62.
Stanton, D. & Stewart, A., (Eds). (1995). Feminisms in the Academy. Ann Arbor: University of
Michigan Press.
Staugaard, A., Jr. (1997). Structured and object-oriented techniques: An introduction using
C++, 2nd ed. Upper Saddle River, NJ: Prentice Hall.
Stein, L. (1999). Challenging the computational metaphor: Implications for how we think.
Cybernetics and Systems: An International Journal, 30: 473-507.
Stolberg, S. (2002, January 18). Bush's Advisers on Ethics Discuss Human Cloning. New York
Times. (http://www.nytimes.com/2002/01/18/politics/18ETHI.html).
Stork, D. (ed) & Lenat, D. (1999). Hal’s legacy: 2001’s computer as dream and reality.
Excerpted from From 2001 to 2001: Common sense and the mind of HAL. Available on-
line at www.cyc.com/halslegacy.html [visited 5 April 1999].
Superson, A. (2002). Welcome to the boys’ club: male socialization and the backlash against
feminism in tenure decisions. In Superson, A. & Cudd, A. (Eds). Theorizing backlash:
261 References
Philosophical reflections on the resistance to feminism (pp. 89-118). Lanham, MD:
Rowman & Littlefield.
Superson, A. & Cudd, A. (Eds). (2002). Theorizing backlash: Philosophical reflections on the
resistance to feminism. Lanham, MD: Rowman & Littlefield.
Suzuki, D. (2003, Jan 28). Human cloning turns into a freak show. Environmental News
Network. On-line at http://www.enn.com/news/2003-01-28/s_2355.asp (visited 11 Feb 03).
Swartz, N. (1994, April 9). Philosophy as blood sport. [Retrieved from the World Wide Web at] http://www.sfu.ca/philosophy/swartz/blood_sport.htm.
Symonds, J. (2000). Why I.T. doesn’t appeal to young women. In Balka, E., & Smith, R.
(Eds.). Women, work and computerization: Charting a course to the future (pp. 70-77).
Boston: Kluwer Academic Publishers.
Taubes, G. (2001, April). Your genetic destiny for sale. Technology Review, pp. 41-46.
Teague, J. (2002, June). Women in computing: What brings them to it, what keeps them in it?
inroads (the SIGCSE Bulletin) 34:2, 147-158.
Teich, A. H. (Ed.). (1997). Technology and the future (7th ed.). New York: St.
Martin's Press.
------. (2003). Technology and the future (9th ed.). Belmont, California: Wadworth/Thomson
Learning.
Townsend, G. (2002, June). People who make a difference: Mentors and role models. inroads
(the SIGCSE Bulletin) 34:2, 57-61.
Travis, J. (1993, April). Making room for women in the culture of science. Science,
260, 412.
262 References
Traweek, S. (1988). Beamtimes and lifetimes: The world of high energy physicists. Cambridge:
MIT Press.
------. (1992). Border crossings: Narrative strategies in science studies and
among physicists in Tsukuba Science City, Japan. In A. Pickering (Ed.), Science as
practice and culture. Chicago: The University of Chicago Press.
Tuana, N. (1996). Revaluing science: Starting from the practices of women. In Nelson, L. H., &
Nelson, J. (Eds.). Feminism, science, and the philosophy of science (pp. 17-35). Norwell:
Kluwer Academic Publishers.
------. (2001). Introduction. In Tuana, N. & Morgen, S. (Eds.). Engendering Rationalities (pp.
1-20). Albany, NY: State University of New York Press.
Tuana, N. & Morgen, S. (Eds.). (2001). Engendering Rationalities. Albany, NY: State
University of New York Press.
Turkle, S. (1984). The second self: Computers and the human spirit. New York : Simon and
Schuster
------. (1997). Life on the screen: Identity in the age of the internet. New York:
Touchstone.
Turkle, S. & Papert, S. (1990). Epistemological Pluralism: Styles and voices within the
computer culture. Signs: Journal of women in culture and society, 16 (1) 128-157.
Ufomata, Titi. (2003, Apr 4). The environment in African folklore: A gendered view. Talk
given at 14th Annual Institute for Liberal Studies [Kentucky State University] Conference
on Science and Culture, Frankfort, Kentucky.
Uhlir, M. (2002, Jan 26). Conflict of interest. Lidove Noviny. Reprinted in World Press Review
(2002, April), 29 (4), 46-7.
263 References
Ullman, E. (2001). Participant observer. Women’s Review of Books, 18 (5), 8.
University of Michigan. (2002, Dec 2). “U.S. Supreme Court agrees to hear admissions
lawsuits.” Press Release 2 Dec 2002. [Retrieved from the World Wide Web on 17 Jan
2003] www.umich.edu/~newsinfo/Releases/2002/Dec02/r120202.html .
------. (2003, Jan 15). “University of Michigan Response to White House Announcement on
Affirmative Action” Press Release 15 Jan 2003. [Retrieved from the World Wide Web on
17 Jan 2003] www.umich.edu/coleman.html .
Van Oost, E. (2000). Making the computer masculine: The historical roots of gendered
representations. In Balka, E., & Smith, R. (Eds.). Women, work and computerization:
Charting a course to the future (pp. 9-16). Boston: Kluwer Academic Publishers.
Velasquez, M. (1998). Business Ethics: Concepts and cases, 4th ed. Upper Saddle River, NJ:
Prentice Hall.
Vetterling-Braggin, M., Elliston, F., & English, J. (Eds). (1977). Feminism and Philosophy.
Totowa, NJ: Rowman and Littlefield.
Wagner, I. (1993). Women’s Voice: The case of nursing information systems. AI & Society 7,
295-310.
Wajcman, J. (2003). Feminist perspectives on technology. In Teich, A. H. (Ed.). Technology
and the future (9th ed.) (pp. 135-147). Belmont, California: Wadworth/Thomson Learning.
Wallace, R. (2001). Making teamwork work: The importance of diverse psychological types. In
HMSBeagle: The BioMedNet Magazine, 111. [Retrieved the World Wide Web 28 Sept
2002] http://news.bmn.com/hmsbeagle/111/note/adapt?print=yes
264 References
------. (2001b). The dynamics of team formation. In HMSBeagle: The BioMedNet Magazine,
113. [Retrieved the World Wide Web 26 Oct 2002]
http://news.bmn.com/hmsbeagle/113/note/adapt?print=yes
Watson, J. & Bragg, L. (1991). The Double Helix : A Personal Account of the Discovery of the
Structure of DNA, (reissue edition). New York: New American Library.
Weasel, L. (2001, Winter). Dismantling the Self/Other dichotomy in science: Towards a
feminist model of the immune system. Hypatia 16 (1), 27-44.
Webster, J. (1995). What do we know about gender and information technology at work? The
European Journal of Women’s Studies, 2, 315-334.
Weintraub, P. (2001, July 6). The bitter feud over LYMErix: Big pharma takes on the wrong
little Osp. HMS Beagle: The BioMedNet Magazine 106.
http://news.bmn.com/hmsbeagle/106.notes/feature3.
Weinberg, S. (2001). Facing up: Science and its cultural adversaries. Cambridge, Mass.:
Harvard University Press.
Wilson, C. (2000). Telepresence. In Goldberg, K., (Ed). The robot in the garden: Telerobotics
and telepistemology in the age of the internet (pp. 64-89). Cambridge, Mass: The MIT
Press.
Women’s Review of Books. (2001, Feb). Participant observer: Ellen Ullman talks about the
facts and fictions of high technology. The Women’s Review of Books XVIII (5), 6-7.
Worell, J. & Johnson, N. (Eds.). (1997). Shaping the future of feminist psychology: Education,
research, and practice. Washington, D.C.: American Psychological Association.
Worell, J. & Remer, P. (1992). Feminists perspectives in therapy: An empowerment model for
women. New York: Wiley.
265 References
Wray, K. B. (1999). A defense of Longino's social epistemology. Philosophy of
Science, 66, S538-S552.
------. (2000, March). Invisible hands and the success of science. Philosophy of
Science, 67, 163-175.
Wylie, A. (1992). Reasoning about ourselves: Feminist methodology in the social sciences. In
Harvey, E. & Okruhlik, K (Eds). Women and Reason (pp. 225-244). Ann Arbor, Michigan:
University of Michigan Press.
------. (Fall 1995). Doing philosophy as a feminist: Longino on the search for a feminist
epistemology. Philosophical Topics, 23 (2).
------. (1996). Precis for Gender Politics and Science: The challenge of ‘gender research’ in
archaeology. Unpublished manuscript.
------. (2000). Feminism in philosophy of science: Making sense of contingency and constraint.
In Fricker, M. & Hornsby, J. (Eds). The Cambridge companion to feminism in philosophy
(pp. 166-184). New York: Cambridge University Press.
Zucker, A. (Ed). (1996). Introduction to the philosophy of science. Upper Saddle River:
Prentice Hall.
266