SFI springBulletin 2005, vol. 20 #1

Simulated Agents in Love and Other New Insights into Human Social Dynamics Santa Fe Institute Bulletin spring 2005, vol. 20 #1

3 6 8 12 17

18 23 30 34 44

Editor’s Note SFI’s founding president, health and aging patterns. This information may George Cowan, reminded shed light on foundational relationships me recently that a goal of the Institute from the between human life span and culture. In other very beginning has been to foster better under- work, Sam Bowles uses mathematical models standing of human behavior. In this spring 2005 and statistics to explore the links between issue, we feature two articles about current SFI genetics and giving behavior. Marc Feldman’s work to that end. The projects described—far- study of the genetic structure of human popula- flung both geographically and topically—have a tions suggests that, particularly for medical pur- common element so far as each brings the rigor of poses, racial categories may not be the best way a quantitative approach to the social sciences domain. of grouping people. In Bali, SFI researcher Steve Lansing is looking at how Of course even with the use of quantitative tools, link- behavioral factors such as language, cultural rules, and ways ing behavior to genetics remains remarkably complicated of economic production affect genetic relatedness within a and often problematic. That’s where SFI’s multifaceted population. His provocatively named computer mode— approach comes in. Looking at the evolving behavior of Simulated Agents in Love (SAIL)—supplements the field- individuals in the context of their social, biological, and cog- work. External Faculty member Hillard Kaplan and col- nitive underpinnings, as well as their natural and social leagues study the Tsimane, a hunter-gatherer society in environments, should help us fathom that most complex of Bolivia. Linking medical information with data on cultural adaptive agents, our fellow human. norms, diet, housing, and social networks, his team is gain- ing a cultural/epidemiological understanding of the group’s GINGER RICHARDSON SFI Bulletin Spring 2005 New Ways of Viewing the World

TOCtable of contents

Gregor Mendel, Meet Florence Nightingale, and Other Wonders of the Nature-Nurture Debate 3 Equality and the Lack of It 6 Seeking the Body’s Innate Fountain of Youth 8 First China Summer School Promotes Interdisciplinary and International Collaboration 12 Rural/Urban Migration in China Focus of Study 13 Shanghai Workshop Explores Organizing Principles of Life 15 2005 Complex Systems Summer School in Beijing 16 Chinese Dignitary Visits SFI 16 Scientists Reunite After Four Decades 17 Cancer’s Complex Nature 18 The Character of Biological Law 23 A Complex Look at Sports 30 Logic of Diversity 34

TRANSITIONS Crutchfield and Fontana Accept Appointments 39 New International Fellows 39 SFI Mourns Lee Segel 39 New Trustees 40 New External Faculty Members 41 SFI Postdoctoral Fellows Move to New Institutions 43 Borenstein 2005 Steinmetz Fellow 43

Understanding Communication— From Interstellar to Intercellular and Beyond 44 Seeking Security in a Hostile Environment 48 Public Lectures 52

Santa Fe Institute Bulletin SPRING 2005 1 2 Santa Fe Institute Bulletin SPRING 2005 Gregor Mendel, Meet Florence Nightingale, and Other Wonders of the Nature-Nurture Debate By Lesley S. King

In the terraced rice paddies SFI Researcher and Professor of and the hilly forests of Bali, Anthropology at the University of Indonesia, a history is being Arizona Stephen Lansing is in the fore- unveiled. It’s the story of front of the research. Building on 30 migrations, love and marriage, years of study in Indonesia, Lansing and and of toiling to feed families. In short, it’s the history of a his colleagues are now embarking on a people on an island. But this is quest to “build a new microscope and told not by archaeological aim it at the emergence of patterns of DPI PHOTO artifacts as is often the case. social structure through time.” To begin / Instead it’s written in the DNA with, they’re comparing the genetic of its characters, in the chains relatedness among farmers in rice-grow- of genes that make up their ing areas (where cooperation is an engi- physical bodies. neering necessity) with those who farm the highland forests (who tend to be

This is only one of a number of sto- much more independent). By doing so, STEPHEN LANSING AND UN . J ries being uncovered at SFI as part of they hope to track how “ecological : its Behavioral Sciences Initiative, an ini- feedback can influence social structure, tiative that brings together researchers from many and note how these processes leave recoverable traces PHOTOS fields to help understand behavior. In this case, scien- in population genetic structure.” tists are looking at various ways that genetics, the sci- Zero in on a single village in Bali where, uncharac- ence pioneered by Gregor Mendel, can separate inher- teristically, half the residents live by means of irrigated ited factors from environmental ones, thereby helping farming, while the other half live by farming open to understand behavior’s many facets. The studies are, patches in the surrounding forest highlands. Lansing as SFI founder George Cowan notes, “bringing the has already noted different marriage practices in the rigor of more measurable sciences to the study of people within this village. “There are quite different social sciences,” which was one of SFI’s initial goals. patterns,” he says. The relatedness among community members in the irrigated farmers is tighter. That’s because they marry each other and stay put. They Demi Nini, the rice Goddess and Her Consort formed from have to cooperate, whereas those growing rice in the the first cuttings of rice at the upstream edge of each forests are less permanent and less cooperative. farmer’s field, Bali Through the passage of time, such differences in behavior may leave traces in the non-coding regions of the farmers’ DNA.

Santa Fe Institute Bulletin SPRING 2005 3 But patterns of relatedness within communities can Gold Mines and Giving be affected by many factors, including marriage rules, It seems odd that the subject of “giving” would be migration, language drift, and historical changes in the controversial, but it is. It wreaks havoc in our nation modes of production. Is it possible to disentangle these and in our hearts. How much of our tax dollars should processes? In true SFI form, Lansing is using computer go to poor or struggling nations, to support war, to modeling to aid in the research. Utilizing SAIL keep peace? How much time or energy should we give (Simulated Agents In Love), he can create a population to our jobs, our families? How much money to chari- of agents, give them marriage rules and DNA, let them ties? And how much should we get in return? On the mate, and see how the patterns of relatedness change dusty rural landscapes of South Africa, a network of over time. “We can look at marriage rules, migration giving and receiving has formed among people with patterns, and population size and see what effect those difficult lives. The network tells a story of duty, chari- have on relatedness,” he says. “We can simulate what ty, and love, helping explain the nature of “giving.” might have led up to the patterns we see in the data.” From an evolutionary perspective, the notion of This project brings together specialists in many “fitness” provides insight into understanding this activi- areas: geneticists Michael Hammer, Tatiana Karafet, ty. When animals feed their offspring, they perpetuate and Herawati Sudoyo; linguists Sergei Starostin, Ilia their genetic makeup, enhancing what biologists call Peiros, and Joel Kuipers, and theoreticians Joseph their “inclusive fitness.” Does inclusive fitness explain Watkins and David Krakauer. The team has already giving among household members? SFI Researcher obtained samples of dialectical variation from nearly Samuel Bowles and his colleague, Dorrit Posel,

THROUGH THE PASSAGE OF TIME, SUCH DIFFERENCES IN BEHAVIOR MAY LEAVE TRACES IN THE NON-CODING REGIONS OF THE FARMERS’ DNA.

200 villages in the islands of eastern Indonesia. The Professor of Economics at the University of KwaZulu- next step is to work with a team of Indonesian geneti- Natal, believe it does, but only in degree. They are cists and medical doctors to obtain DNA samples from studying migrant workers in South Africa, mostly communities in the region. The data will also be used those who spend long periods of time away from their for medical research on diseases such as malaria and families while toiling in gold mines. The migrants send thalassemia. substantial amounts—called remittances—home to The outcome of such research may have broad their families, typically amounting to about half of implications. Indonesia is an ideal microcosm within their urban wage. which to look at the way behavioral elements such as Bowles and Posel wanted to know: Do they send language, cultural rules, and modes of living affect pat- more if they are more closely related genetically to terns of relatedness because each island encodes a dif- their household of origin? Their households are large ferent history. The archipelago was colonized by and genetically heterogeneous and thus a good study Austronesian-speaking farmers and fishermen about group. Some of them contain no blood relatives of the 5,000 years ago. Since then, the island cultures have migrant (his wife, and her siblings, for example) while developed in different ways. It may be possible to others are closely related (his children, his parents, or reconstruct not only the broad outlines of their histo- his siblings). ries, but also the ways their communities and lan- Through the use of mathematical modeling and sta- guages changed over time, and whether these changes tistical testing, the researchers have found that genes do had different effects on women and men. If the project influence the way people give, but they provide only turns up clear patterns of association, they can be test- part of the explanation. Referring to the hypothesis that ed in other regions of the world. people favor close kin, Bowles and Posel write in a In December, Lansing gave a lecture in Jakarta paper appearing in Nature, “Data on remittances sent by about the first results from this project, which focused South African migrant workers to their rural households on Bali. The story merited a full page of coverage in of origin allow an explicit test, the first of its kind for Indonesia’s leading newspaper, and led to offers of humans.” The researchers employed a number of fac- assistance from Indonesian geneticists, archaeologists, tors in their model. They used estimates of the fitness and linguists. benefits and costs associated with the remittance; they

4 Santa Fe Institute Bulletin SPRING 2005 took into account the genetic relatedness of the migrant also (by Bowles and his colleagues) on hunters and to the beneficiaries of the transfer; and they factored in gatherers and in other “small scale societies” around the beneficiaries likelihood of contributing to future the world. The experiments show that people are often gene pools based on their age and gender. Thus they generous to total strangers, and are willing to pay to were able to estimate the level of remittance that maxi- punish those who treat others unfairly. mizes the migrant’s inclusive fitness. Cynics like to say that we are born selfish. Recent They’ve shown that, “Yes, the migrant worker will empirical research – the South African migrants, the

THE RESEARCHERS HAVE FOUND THAT GENES DO INFLUENCE THE WAY PEOPLE GIVE, BUT THEY PROVIDE ONLY PART OF THE EXPLANATION. give more at a cost to himself if that giving will benefit experiments on students, hunters and gatherers, and his own genes,” says Bowles. “But the effect is mod- others—suggests a quite different picture. Even though est,” they write. “Less than a third of the observed level people differ a lot across cultures and situations, and of remittances can be explained by our kin-altruism can be influenced by self-interest, they are nonetheless model.” In fact, the migrants give much more than often civic minded even towards non-kin. Says they’d give considering only genetic relatedness. Bowles, “We should think of designing public policies “Nobody thinks that people are trying to spread and shaping organizations to bring out the best in peo- their genes here,” says Bowles. “The motives for giving ple while guarding against the selfish types, rather than include love, obligation, and investing in future family just assuming the worst about everyone.” ties.” But these motives induce people to behave as if they cared about spreading their genes. Sickle-Cell Disease and Unity Other factors come into play here, such as the pres- Another exploration of the nature–nurture debate ence of a wife. A wife is not genetic kin, but, Bowles spans across regions as disparate as Africa and the and Posel write, the wife’s presence in the household Mediterranean and explores not how behavior affects may increase remit- genes or genes affect tances both for behavior, but how inclusive-fitness rea- best to behave sons (she cares for toward genetic pat- the migrant’s chil- terns. SFI Science dren and may bear Board member and more children) and Stanford geneticist due to altruistic Marcus Feldman and motives toward colleagues are help- non-kin. But Bowles ing penetrate the notes that while the mystery of unity in wife’s contribution human genetic to inclusive fitness makeup. They’ve helps explain remit- shown that in terms tances, most of the of racial categories, “wife present” effect cannot be explained this way. humans are more similar than was suspected. “Maybe they simply love their wives,” says Bowles. In a study of the genetic structure of human popu- Maybe so, but this raises the question: Why are we so lations, Feldman, working with colleagues Noah nice to people who are not our kin? Rosenberg and Jonathan Pritchard, both alumni of the The study dovetails nicely with recent experimental SFI Summer School, looked at the genetic makeup of work done by Bowles and by SFI Vice President for 1,056 individuals from 52 populations and found that Academic Affairs John Miller and SFI External Faculty people are less different than one would think. member Ernst Fehr. These experiments—done anony- Generally, race has been defined by phenotypes such mously and with real money—have been carried out as skin color, facial features, and hair type, but these not only on psychology majors and other students, but data show that “the within-population component of

Santa Fe Institute Bulletin SPRING 2005 5 SO FOR THE VAST MAJORITY OF THEIR DNA, AN AFRICAN OR ASIAN PERSON IS AS SIMILAR GENETICALLY TO A CAUCASIAN, AS A CAUCASIAN IS TO HIS CAUCASIAN NEIGHBOR. genetic variation, estimated in the study as 93 to 95 toward the disease. “To use genotype effectively in mak- percent, accounts for most of human genetic diversity. ing diagnostic and therapeutic decisions, it is not race Thus most genetic diversity occurs within groups, and that is relevant, but both intra- and trans-continental very little is found between them.” So for the vast contributions to a person’s ancestry,” they write. majority of their DNA, an African or Asian person is as The findings have many ramifications. “The assign- similar genetically to a Caucasian, as a Caucasian is to ment of a racial classification to an individual hides the his Caucasian neighbor. Of course, the picture is more biological information that is needed for intelligent complex than saying we’re all one, but their point is, therapeutic and diagnostic decisions,” the group writes. that particularly for medical purposes, racial categories They note that a person classified as “black” or are not the best means of grouping people. “Hispanic” could have a mixture of ancestors when Writing about this finding in Nature, Feldman and looking at his or her continent of origin. “Confusing colleagues Richard Lewontin and Mary Claire King state race and ancestry could be potentially devastating for that ancestral genetic data are far more useful. They cite medical practice,” they say. They hope that, when sickle-cell disease as an example. Often thought to be an treating diseases with genetic influences, their findings African disease, tendency toward it is actually tied to will lead to an understanding that all contributions to a ancient ancestry in a geographic region where malaria patient’s ancestry might be relevant. was endemic. Africa is one region, but so are the Feldman also hopes that the findings affect the way Mediterranean and southern India. Thus people whose disease is studied as well. He explains that today many ancestors come from those areas might be more prone research dollars are funneled into understanding dis-

Equality and the ing genetic data. The group found, Lack of It for example, that a son whose par- By Lesley S. King ents are in the top tenth of income earners has more than 20 times the In recent months, SFI Researcher chance of ending up in the top ten Samuel Bowles has been causing a percent himself compared to the son stir. He’s appeared on National born to parents in the poorest tenth. Public Radio and his research has However, it’s not a doomsday pic- been cited in The New York Times. ture they present, but instead a sci- The attention is because of ideas he entific portrait of equality and the and colleagues are presenting in a lack of it, with hopes of one day lev- new book titled Unequal Chances – eling the playing field so that this Family Background and Economic country can move closer to its ideal. Success (Princeton University Press, isn’t the land of equal opportunity “Genes are apparently part of the 2005), which presents the results of after all. picture,” says Bowles about the a workshop held at SFI. The findings He delivers this message not in unlevel playing field, “but not a very ought not to surprise anyone, but tones of antagonistic fervor, as some large part, and not for the reasons somehow they tear into the very politicians and activists tend to do. most people would think.” Bowles, heart of America’s self-love: Bowles’ Instead he uses statistical data drawn working with Herbert Gintis, SFI research is revealing that America from studies, some of them includ- External Faculty member and

6 Santa Fe Institute Bulletin SPRING 2005 ease on a molecular level, but that DNA is due to genetic differences between can explain only a tiny percentage of females and males is the same kind major behavioral differences or chronic of statement that has continually diseases. “When the focus has been so been used about races. And until we much on reducing behavior or chronic know that we have the environmen- disease to a molecular phenomenon, we tal and opportunistic issues resolved, lose sight of the other factors involved in we cannot rely on statistical correla- cultural variation, and the resources do tions to claim causation.” not go to study these,” he adds. He notes that SFI’s Behavioral Many problems exist in attempting Sciences Initiative is an important to link behavior and genetics. He cites, step in looking at such issues. “There for example, studies in the past that have been a lot of conclusions made have attempted to say the reason fewer about the role of genetics, but we women than men pursue science is due really have to look at the whole, be care- to genetic differences. Feldman says we ful not to miss other influences, such as first have to discount the importance of fine-scale geographic and lifestyle differ- societal issues, such as the fact that ences, for instance. So much of the pic- women often balance child-raising and ture can be there, and can be missed by careers, and even developmental ones. assuming that it can all be reduced to Girls often lose on the social value molecular causes.” scale by being good on the math scale, he notes. “We know those sex differ- Lesley S. King’s articles have appeared in ences exist as they do with race, but to The New York Times, Audubon, and other claim that a difference in achievement publications.

Professor Emeritus of Economics at may be important,” they write. personality, race, wealth inheritance, the University of Massachusetts, But they caution that heritability and even luck as forces affecting the Amherst, and Melissa Osborne can be affected by changing the transmission of economic success Groves, Assistant Professor of environment. And, echoing the work across generations. The study could Economics at Towson University, of Marcus Feldman and his col- impact many areas of public policy, seeks to “uncover the channels leagues, they stress that estimates of and, judging from the stir it’s creat- through which parental incomes this kind are sensitive to plausible ing, public opinion as well, especial- influence offspring incomes.” To do alternative assumptions. Surprisingly, ly since it shines an unwelcome light this, they study the income of twins. they found that the genetic inheri- on such factors as race and class, Twins provide insights into the tance of IQ plays a negligible role in identifying them as helpers or hin- nature-nurture debate because in the process. “Other similarities drances to success—things that the most cases they have been reared in between parents and children—of U.S. “melting pot” has pretended to similar environments, thereby hold- race, personality, looks, and health transcend. “It hits the funny bone of ing “nurture” constant. The status—are likely to provide a better America,” says Bowles. “People still researchers found that identical explanation of the role of genes in like to say that anyone can get to the twins’ earnings were more similar— keeping wealth in the family,” con- top. I wish it were true. ” one twin to another—than were the cludes Bowles. incomes of fraternal twins. “The The book tackles subjects far remarkable income similarity of beyond the twin studies and the identical twins compared to fraternal United States. In Unequal Chances, twins suggests that genetic effects the team looks at influences such as

Santa Fe Institute Bulletin SPRING 2005 7 Seeking the Body’s Innate Fountain of Youth By Barbara Ferry

In the Beni region—the lowlands—of Bolivia, the Most of what we know about the aging process Tsimane people still live by traditional means, employ- comes from studying modern populations. However ing hunting, fishing, and modest horticultural gardens such studies provide an incomplete picture because to obtain their food. A rare example of a traditional they leave out a vast majority of humans’ evolutionary subsistence population, they have lean bodies that history. By studying a traditional society, a group endure high rates of disease, tough workloads, and lit- about which very little is known in regards to aging tle modern medical care. Because of these factors, and the epidemiology of disease, Kaplan hopes to con- Hillard Kaplan, SFI External Faculty member and pro- tribute to a debate on the human life span that ulti- fessor of anthropology at the University of New mately seeks to answer the questions: what causes Mexico, believes the Tsimane will provide a window humans to age and die and how long can people live? into understanding the process of aging. For three It is almost certain that at least some of the process- years, Kaplan has been researching this population es leading to aging among traditional subsistence peo- alongside Michael Gurven of the University of ples differ from those afflicting people in modern California at Santa Barbara and graduate student Jeff developed nations. In developed nations, humans Winking, University of New Mexico. experience very different diets, activity regimes, and physical assaults on their bodies.

8 Santa Fe Institute Bulletin SPRING 2005 Modern hunter-gatherers are not living replicas of The starting point for the new theories on the our stone age past, having been affected by global human life course is the observation that successful socioeconomic and epidemiological forces. reproduction is not mainly a matter of producing off- Nevertheless, studying them can provide an important, spring, but rather is a matter of acquiring food and if imperfect, lens on the life histories of our ancestors allocating it to offspring so as to maximize the number and the conditions that shaped the biology of human of surviving, sexually mature adults. This perspective development and aging. Such research is urgent, places the acquisition and distribution of food at the because global change is proceeding rapidly, and virtu- center of reproductive fitness, rather than fertility. ally all human populations will soon be incorporated Humans have long childhoods, compared to other into global economic and health systems. species, and may be dependent on parents for as long The next decade will probably be the last during as two decades. From an evolutionary point of view, it which such research will be possible. Although the makes no sense for people to be able to reproduce up Tsimane were exposed to Jesuit missionaries in the late until their mid-40s if they cannot survive long enough 17th century, they were never successfully settled into for their children to become self-sufficient. Natural missions and remain relatively unacculturated. The selection seems to have favored long childhoods in Tsimane are thus an ideal population for a study of the humans, and previous theories of aging didn’t take this aging process among traditional forager-horticulturalists. into account. By linking medical information with more tradition- A second component of the Santa Fe Institute proj- al studies of cultural norms, diet, hygiene, work effort, ect is to organize a theoretical working group of biolo- activity, mobility, housing, and social networks, Kaplan gists, demographers, economists, and anthropologists and his research team hope to gain a more complete to exchange ideas and promote further development in

THIS PERSPECTIVE PLACES THE ACQUISITION AND DISTRIBUTION OF FOOD AT THE CENTER OF REPRODUCTIVE FITNESS, RATHER THAN FERTILITY.

cultural epidemiological understanding of health and the theory of the life course. The third component is to MICHAEL GURVEN : disease among the Tsimane. A long-term goal of the assemble another working group of empirical scientists research is to explain the age profile of human mortali- interested in aging research in both developed and less PHOTOS ty and the rate at which humans develop and age in developed contexts, with the goal of stimulating more terms of economic productivity, muscular strength, comparative research. endurance, body composition, disease resistance, and In addition to informal working group meetings, cognitive function. two workshops will be organized, one focusing on the- oretical results and the other on empirical applications. Caring Parents Also working with Kaplan on this SFI project are How We Grow Old—Contrasting Views economist Arthur Robson of Simon Fraser University Life expectancy in developed countries has and Ronald Lee, director of the Center on the increased dramatically in the past 100 years. Medical Economics and Demography of Aging at University of advances have led to a reduction of infant and child- California, Berkeley. Both Robson and Lee have inde- hood mortality rates and helped people live longer. pendently proposed new theories of aging with a com- These increases have reopened the debate on the mon point of departure: classical biological theories of potential length of the human life span, but there are aging are inadequate for species such as humans, differing views on how aging occurs. which engage in extensive and prolonged parental According to one view, the human lifespan is like a care. ticking clock (time bomb?) with an upper limit of According to the new theories, resource transfers about 85 to 86 years. Improved medical care and pub- from parent to offspring are critical to understanding lic health have allowed people to live longer and the role of natural selection in determining mortality caused the distribution of deaths to become increasing- rates. One goal of the project is to further develop ly compressed in the upper-age range. This ticking these theories and to work through the shared and clock view, characterized as the gerontological view, unshared features of the two approaches. treats the various outcomes of aging as resulting from

Santa Fe Institute Bulletin SPRING 2005 9 a single, unitary process, resulting in programmed tion of dietary intake in many mammals, is associated death. with a series of physiological adjustments, including But, according to an alternative view, there is no set lower growth rates, delayed age at first reproduction, upper limit to the human lifespan, and improvements and shortened life span. Human life histories show evi- in medical treatment and living conditions will contin- dence of systematic variation in response to environ- ue to produce increases in longevity. This view, called mental variation. These outcomes appear to be the the epidemiological view, treats aging as resulting from result of the interaction between changes in environ- a competing set of risks due to independent or quasi- mental conditions and genes affecting human physiol- independent processes. According to this view, there is ogy and behavior. no ticking time bomb that results in death, and as advances in medical care succeed in preventing or cur- Diet and Inflammation Count Too ing diseases, the causes of aging and death can individ- Standard epidemiological theory suggests that ually be eliminated. chronic diseases, such as heart disease and cancer, have increased in relative frequency, due to reductions in Genetic Factors Matter mortality resulting from infectious disease. Fewer peo- Although there have been dramatic gains in the sur- ple die from infectious disease, so the relative rate of vival rates of older people in the United States over the chronic disease such as heart disease and cancer has last several decades, the potential for improvements increased. Changes in diet and exercise in developed remains an open question, because our knowledge societies may also play a role. Yet it also has been about the interaction of biology, behavior, and envi- argued that aging and the onset of chronic disease may ronmental conditions in determining the aging process be accelerated in response to poorer nutrition and

ACCORDING TO AN ALTERNATIVE VIEW, THERE IS NO SET UPPER LIMIT TO THE HUMAN LIFESPAN, AND IMPROVEMENTS IN MEDICAL TREATMENT AND LIVING CONDITIONS WILL CONTINUE TO PRODUCE INCREASES IN LONGEVITY. and age-specific mortality rates is still limited. There is, increased disease loads. however, growing evidence that the aging process is Two University of Southern California researchers strongly influenced by biological control mechanisms, who have been collaborating with the team recently and that genetic variation among humans and nonhu- proposed that major decreases in mortality at the older mans is associated with differences in rates of aging. ages are due to a reduction in the level of inflammation This means that between species, and within the experienced over a lifetime. The human body’s reac- human species, there are genetic dif- tion to bacteria, viruses, and parasites is ferences that code for different rates of to engulf the agent, causing inflamma- aging: genes regulate cellular process- tion. These researchers believe that es that lead to different rates of cellu- chronic exposure to such physical lar, and ultimately whole-organism assaults has negative secondary effects aging. But the interrelation between and may cause the body to age more genes, genetic expression, and the rapidly. Defense and repair is costly to environment is not completely under- the body. With the introduction of stood. Progress in mapping the human antibiotics and public water systems, genome is likely to lead to major exposure to foreign agents, and hence, advances in our understanding of the the body’s inflammation, has decreased. genetic substrate governing the aging Kaplan’s research project is collecting process. data on diet, weight and growth, bio- There is also strong evidence, markers of inflammatory processes, however, that genetic control mecha- viral, bacterial, and parasitic infection nisms do not necessarily lead to a rates, and aging among the Tsimane to fixed pattern of aging. For example, help evaluate the different theories of environmental factors, such as restric- aging that have been proposed. In addi-

10 Santa Fe Institute Bulletin SPRING 2005 Researchers collect data on the Tsimane’s diet, weight, growth and other factors

tion to these scientific goals, Kaplan’s research is con- Most of the theories of aging have been based on tributing to student training and the participation of analyses of historical populations in societies with a traditionally under-represented groups in research. central government, such as historical Europe. By Both Bolivian and U.S. graduate and undergraduate working with the Tsimane, Kaplan and his team will students have participated in data collection and will be able to, in a sense, travel back in time and thereby continue to do so in the future. The research also bring new insight to the future. involves the community members themselves in data collection, and trains Tsimane in primary health care. Barbara Ferry is a staff writer for The Santa Fe New Mexican.

Santa Fe Institute Bulletin SPRING 2005 11 PHOTO COURTESY SFI

First China Summer School Promotes Interdisciplinary and International Collaboration By James W. Minett

This past year SFI’s International Program sponsored young researchers and to further my current work in a Complex Systems Summer School in China. As a modeling processes of language evolution. research fellow working in Hong Kong, I learned that The school was held at Qingdao University in William Wang, who supervises the work of our language Shandong Province, located on the coast of China. evolution group at the Chinese University of Hong Organized jointly by SFI and Qingdao University, with Kong, would be giving a week-long course in evolution- the cooperation of the Key Laboratory of Complex ary linguistics at the Summer School. Then I discovered Systems and Intelligence Science of the Chinese that John Holland, External Faculty member at SFI and Academy of Sciences, the school brought together aca- professor in the departments of psychology and comput- demics and students from around the world, with about er science at the University of Michigan, was to be his sixty percent of the participants coming from China. co-lecturer. I knew I would apply. Attending the school Qingdao University is situated just beyond the city would be an opportunity for me to meet a new group of center, whose restaurants, bars and beaches provided

12 Santa Fe Institute Bulletin SPRING 2005 ample entertainment throughout our eling techniques, whereas the over- After opening preliminaries, one-month stay. Downtown Qingdao seas students tended to focus their Dave Feldman, from the College of is modern, with newly built office research on applications in specific the Atlantic, kicked off the main towers and shopping malls. Other fields such as ecology, economics, business of the school with a series parts of the city reflect both the ele- and linguistics, although there were of lectures on complexity science, gant architecture constructed during exceptions in both cases. This com- summarizing a number of important the European occupation of the city plementary blend of foci was, no concepts underlying the analysis of at the beginning of the 20th century doubt, deliberate on the part of the complex systems, such as informa- and the dilapidated structures of vil- summer school organizers to pro- tion theory, entropy, and computa- lages that have been engulfed by the mote both interdisciplinary and tion theory. These lectures were expanding city, hinting at an agricul- international collaboration. helpful in providing a guide to some tural past. The city even inspired a Like previous schools, the Qingdao of the general tools that can be used couple of the projects, with one Summer School consisted of three for complex systems analysis, point- group modeling the popularity of the weeks of lectures and group project ing the way to more detailed pre- city’s bars and another studying the work, followed by one week for the sentations in the literature. self-organization of individuals mov- completion and presentation of the Also in week one was a discus- ing about in bounded spaces, such as group projects. The school focused sion by Thomas Peacock, from the the city’s buses. more on a broad coverage of The areas of interest of the stu- applications of complexity sci- dents attending the Summer School ence than on theory, although were broad. Significantly, the the school did begin with a research interests of the majority of one-week intensive course on COPYRIGHT CORBIS

those from China were on complex- fundamental aspects of com- : ity theory and computational mod- plex systems research. PHOTO

Rural/Urban Migration in China Focus of Study

SFI Science Board member Marcus Feldman from Using social network analysis and models of cultural Stanford teamed with Li Shouzhou at Xi’an. transmission and evolution, the new research will incorporate In rural China, where a strict patrilineal family system is personal social networks, characteristics of regions where dominant, preference for sons is ubiquitous. However, in migrants flow-in and flow-out, living arrangements and dura- large cities the extent of son preference has been weakened tion of stay in urban areas, frequency of returning to the by the process of modernization and improvement in the countryside and duration of stay in rural areas, individual social security system (families are no longer as reliant on socioeconomic characteristics, mass media, as well as local the son to support them in old age). Migration from the policies and regulations, etc. The models aim to describe country to the city leads to a dramatic change in lifestyle and interaction among temporary rural-urban migrants and urban social networks, which most probably influences the values residents in terms of culture and behavior and to simulate and concepts of rural-urban migrants. The original, strongly the dynamics of transmission and diffusion. male-biased culture and the corresponding behaviors are like- Development and testing of models for the cultural trans- ly to be influenced by the host culture, which is more mod- mission and diffusion during the process of rural-urban migra- ern. In particular, marriage and childbearing preferences tion in China as well as the interaction of these migrants and change, as well as other behaviors, which trickle back to the urban residents is important, both academically and practical- country, influencing those peasants who later marry in the ly. The research will help to predict the social and economic rural areas. The influence may eventually give rise to a cultur- consequences of rural-urban migration in China and perhaps al transition within the whole population and even a diminu- produce corresponding policy formulation for community tion of son preference. development.

Santa Fe Institute Bulletin SPRING 2005 13 Department of Mathematics at MIT, on pattern formation in fluid mechanical systems. Peacock’s lec- tures combined an overview of the mathematics of fluid mechanics, dis- cussing, for example, the Navier- Stokes equations, Rayleigh-Bernard Convection, and Taylor-Couette Flow, with video clips that illustrat- ed vividly the types of patterns that can emerge in such systems. Although his lectures did not pro- vide students with specific tools with which to tackle problems in their own fields, they did serve to highlight the rich behaviors that can emerge from complex systems. PHOTO COURTESY SFI In the second week of the The school also included lectures introduction to the theoretical discus- school, the lectures switched to biol- on the evolution of agriculture, sions of Holland and to ground them. ogy and the evolution of agriculture. specifically rice domestication and Holland concentrated on describing a Eric Smith, a research professor at cultivation in China, Southwest framework, based on the classifier SFI, and Satoru Miyano, from the Asia, and the Americas. It also had system, for modeling the emergence Institute of Medical Science at the lectures on network dynamics, pro- and evolution of language. University of Tokyo, spoke about viding a basic introduction to net- The Echo system, described in reaction networks in biology, with work-based tools for complex sys- Holland’s book Hidden Order, was Smith providing an overview of the tems research. originally derived to study the citric-acid cycle and Miyano describ- In the third week, William Wang dynamics of complex adaptive sys- ing how Petri nets can be used to and John Holland lectured on my tems. It is an agent-based, rule- model interactions among genes. own research area of language evolu- based system consisting of heteroge- Hao Bailin’s (Institute for tion. The style of presentation here neous, interacting agents that evolve Theoretical Physics, Chinese was a little different from the other by building up rules to encode bene- Academy of Sciences) discussion of talks, with Wang and Holland alter- ficial behaviors. This system is a sit- the K-strings method for character- nating their discussions between the uated model: agents must interact izing and comparing genetic empirical and theoretical. Wang with an environment and survive in sequences was particularly interest- focused on the empirical, sampling a it by filling reservoirs, measures of ing to me because the technique broad range of features that illustrate the extent to which they are able to might find application in the phylo- the structure and complexity of lan- fulfill certain basic requirements, genetic classification of language, guage. From an overview of FOXP2, such as acquiring food, finding shel- one of my own research interests. the so-called “language gene,” whose ter, and so on. Whenever a reservoir The opportunity to learn about such role in the phylogenetic and ontoge- is low, a bridging rule, e.g., a rule a method for potential use in my netic emergence of language has stating “I’m hungry!” is activated own field was, for me, one of the been much discussed in the recent lit- and stays on until the reservoir is re- most valuable aspects of the erature, to the historical development filled; bridging rules act to keep the Summer School. of the Chinese language, his lectures agent focused on an important task, served both to provide a stimulating such as finding food. New rules are

14 Santa Fe Institute Bulletin SPRING 2005 triggered whenever a general set of number of researchers, including Instructive as most of the lec- conditions are met. For example, a both Holland and Wang, now tures were, I found the most suc- child may learn to speak a new believe that language first emerged cessful aspect of the Summer School word “cookie,” if she is hungry, she through the re-use of existing cogni- to be the group project work that hears her mother say “cookie,” and tive, physiological, and social func- each of us undertook. The process she happens to be seeing an actual tions. Holland therefore used the of forming groups was a complex cookie. New bridging rules can be lectures to pose the question: can a system in itself: while some people triggered too, when, by chance, an simple form of shared language picked a particular topic to work on agent undertakes an action that emerge among a group of interact- and stuck with that, others moved leads to a reservoir being filled ing agents based on just domain- from group to group, both to sample when no existing bridging rule is general skills, such as shared atten- the topics being considered and to active, allowing agents to learn new tion and recognition of recurrent contribute ideas of their own. ways to focus on important tasks. patterns? Although he did not pro- Indeed, several students participated Currently, many linguists believe vide an explicit answer to this ques- in more than one project. that language is an innately tion, he did describe endowed system that evolved how a classifier sys- through the selection of language- tem might be con- specific cognitive and physiological structed to help functions. However, an increasing answer this question. COPYRIGHT CORBIS : PHOTO

Shanghai Workshop Explores Organizing Principles of Life

Last May, SFI co-sponsored an International Workshop on of inheritance to the emergence of neural maps. Biocomplexity with Fudan University, one of China’s leading Presenters at the workshop came from China, Taiwan, universities with an international reputation for academic Korea, India, Germany, and the U.S. Their primary research excellence. The workshop also served as the 40th Eastern interests ranged from laboratory synthesis and database cre- Forum for Science and Technology sponsored by the ation and mining, to public health and agriculture, to funda- Shanghai Municipal People’s Government, the Chinese mental mathematical and computational representations of Academy of Sciences, and the Chinese Academy of living systems. Engineering. The organizers were SFI International Fellow What held such an eclectic meeting together was a Hao Bailin, academician of the Chinese Academy of Sciences shared sense among the participants that there are organiz- and professor at the T-Life Research Center at Fudan, and ing principles characteristic of life that transcend these appli- Eric Smith, research professor at SFI. cation domains. What we learn about the roles of hierarchy The meeting was topically broad, including, but also and autocatalysis at the molecular level may also predict con- reaching beyond familiar fields in biocomplexity such as straints on the emergence and succession of forms of civi- genomics and proteomics, bioinformatics, and the genetic lization. Network dynamical principles discovered to make reconstruction of phylogenetic (inheritance) trees. Also repre- protein and gene regulation comprehensible continue to be sented were ecosystem and microbialecology, scaling and relevant to the structuring of ecologies, epidemics, and invariance principles, and the mathematical structure of evo- immunology. As we understand better the principles that lutionary theory and information transmission across genera- underlie the amazingly diverse yet resilient natural world, we tions. Important new directions that are reshaping classical can work to integrate engineering and medical practices fields were presented, including quantitative and computa- within that framework, discovering elegant ways to restore tional modeling of the emergence of civilizations, and the functions that integrate smoothly into the living processes study of network structure and dynamics. Finally, intrinsic responsible for creating it. system-level phenomena were discussed, from the structure

Santa Fe Institute Bulletin SPRING 2005 15 The common interest of our aim of the project work was less for a constructive blend. It is to be group was extinction, albeit from about carrying out new research and hoped that complexity science different perspectives, such as ecolo- more about gaining experience with teaching and research can continue gy, paleontology, and linguistics. We using methodologies that will be to blossom in the coming years so soon agreed to form a team to useful in future research. that the flow of knowledge between model extinction processes, a project Taking part in the Qingdao nations and across disciplines can that would support my research into Complex Systems Summer School grow. the dynamics of language extinction. proved to be a rewarding experi- James W. Minett is a research fellow The range of topics considered in ence, both for the broadened per- working on evolutionary linguistics in the projects was wide, from the for- spective to research in different the Digital Signal Processing & Speech mation of biofilms, the layer of fields that it offered and for the Technology Laboratory of the microbial organisms that grows, for opportunity to work in mainland Department of Electronic Engineering example, on one’s teeth when left China. The different general at The Chinese University of Hong unbrushed; to the modeling of con- approaches of the Chinese students Kong. For more information on his flict and warfare. Given the short —theory-based—and the overseas research, you can visit his homepage at duration of the summer school, the students—application-based—made http://www.ee.cuhk.edu.hk/~jminett/

Chinese Dignitary Visits SFI Business Network, with Ann Stagg, manager of Marketing and Business Relations, explaining how the Network con- Jiang Mianheng, son of former tributes to the environment of scientific entrepreneurialism Chinese President Jiang Zemin, visit- and innovation, in addition to providing valuable research rev- ed SFI on February 28th and partici- enue for SFI. The visit culminated in a tour of the Institute pated in a round table discussion with

and informal discussions with various researchers. PHOTO COURTESY SFI SFI researchers. Dr. Jiang is a suc- cessful entrepreneur instrumental in the development of the information technology industry in Shanghai and also a vice president of the Chinese Geoffrey West and Jiang Mianheng Academy of Sciences. A guest of the New Mexico Economic Development Board, he requested to visit SFI to learn how the Institute fosters interdisci- plinary research and maintains an environment of academic creativity and freedom. This trip followed visits to the MIT Media Labs and Harvard University, where he 2005 Complex Systems Summer School in Beijing explored how to ensure standards of academic excellence and This summer, the second Complex Systems Summer School how to evaluate the success of a bottom-up organized institu- to be held in China will be hosted in Beijing by the Institute of tion. He explained that the Chinese Academy of Sciences has Theoretical Physics and the Academy of Mathematics and been undergoing many changes and is committed to building Systems Sciences of the Chinese Academy of Sciences. The world-class research institutions. school, which will run from July 11 through August 5, will SFI researchers Jessica Flack, John Holland, David focus on both theory, with lectures on the fundamental princi- Krakauer, John Miller, Eric Smith, Geoffrey West, and Henry ples and tools of complexity science research; and applica- Wright participated in the discussion with Jiang, who was tion, including biological systems, social aspects of language especially curious about why the researchers chose to work evolution, and population dynamics. The Summer School at at SFI over other academic institutions. The group also dis- Santa Fe, held from June 6 to July 1, will focus more on the- cussed many SFI/China initiatives, ranging from research col- ory, although it too will deal will such applied topics as the laborations to the upcoming Complex Systems Summer modeling of food webs, the evolution of cancer, and the School in Beijing. The discussion also touched on the SFI structure of the World Wide Web.

16 Santa Fe Institute Bulletin SPRING 2005 PATRICK MCFARLIN : PHOTO Scientists Reunite After Four Decades

“I was browsing through a book- of complex adaptive systems (CAS). agents start with very primitive cogni- store in Berkeley and saw a book called Not long after they taught together, tive abilities. They learn, and adapt Emergence,” says Bill Wang, research Wang got an appointment at Berkeley. through doing so. “If we demonstrate professor of language engineering at “He became well known in linguistics,” that language can be learned with more the Chinese University of Hong Kong. Holland says. Then time and distance primitive abilities, then that would “I took it off the shelf and was sur- took over, causing them to lose touch change the way linguistic research is prised to see it was by John Holland.” with each other for four decades. done,” says Holland. This serendipitous instance initiated a After reuniting, they realized they This spring the work expanded with reuniting of old friends and an impor- shared many scientific interests. “Ever the help of a workshop at SFI con- tant collaboration for scientific since that course with John at vened to discuss “Language Evolution research. Michigan, I have been working at and Acquisition: Models, Networks, Wang e-mailed Holland, who is an adapting ideas from evolution theory to Robustness, and Diversity.” The work- SFI External Faculty member and pro- linguistics,” says Wang. “Hence my joy shop included Holland, Wang, and six fessor of psychology and computer sci- at discovering John’s book.” other colleagues from around the ence and engineering at the University Holland went to Hong Kong, where world. of Michigan. “We started by comparing a collaboration began, uniting his “The collaboration became even stories about our grandchildren and research in CAS with Wang’s work in more interesting since some of SFI’s then the relationship became more and language. “Many of the researchers in activities were beginning to blossom in more substantive,” says Wang. our program had heard of John and his Asia,” Wang adds. So, 43 years after “Forty-three years ago,” relates work, so it was nice to have him in co-teaching one of the first courses in a Holland, “Bill and I were at the Hong Kong,” says Wang. At the time, field that would one day become com- University of Michigan. We had the Wang wasn’t new to the work of SFI. plexity science, Holland and Wang co- first appointments in what was then In fact he was already on an advisory taught a course in language evolution called the Communication Sciences committee for Murray Gell-Mann’s proj- at the SFI Summer School in Qingdao. Program.” At that time programs such ect on language evolution. They will teach together again at the as this existed only at Michigan and In the research, Holland and Wang 2005 school in Beijing. MIT. Together Holland and Wang are using CAS to redirect the focus of “What I look for, in life and work, is designed and taught the first course in understanding language acquisition, fun,” says Holland. “And this has been the program. What’s now called the removing emphasis from genetics and really fun.” Computer Sciences Program, has since placing it on the process of learning. —Lesley S. King become one of the leaders in the study They’ve created a model in which the

Santa Fe Institute Bulletin SPRING 2005 17 Cancer’s Complex Nature by Daniel Rockmore

Much of what goes on at SFI, indeed even the very creation of SFI, is and was inspired by a desire to study and understand life, from its origins in some primordial molecular soup to the complicated web of interactions into which it has evolved. Nevertheless, many of the tools developed at SFI to plumb the mysteries of the growth and evolution of living systems can also be used to shed light on those darker processes that bring life to an end. These final phenomena range in scale from the grand to the minute: the catastrophic phase transition that could produce a “Day After Tomorrow”-like snowball earth; the cascade of extinctions modeled by the dynamics of a food-web network; the emergent behavior that signals a market crash; or the progress of disease, either across a society or within an individual or even a single cell. It is the last of these, and in particular, toward understanding the etiology of cancer, where SFI scientists are making some significant contributions.

Editor’s Note: This is an expanded version of an article that ran in the previous Bulletin.

18 Santa Fe Institute Bulletin SPRING 2005 CELLULAR MANIFESTATIONS OF CANCERS

Cancer as a Complex Adaptive System crowning achievement of this work has been their discov- At some level of consideration, the healthy human ery of a “universal growth law” for organisms, one that body is a multi-dimensional mosaic of cells, differentiated displays a three-fourths power relation between the body by their various functions (e.g., skin cells or muscle cells). mass of an organism and its metabolic rate as well as a Cancer develops as an uncontrolled reproduction of principled derivation of the evolved intricate branched abnormal cells, which can then embark on a deadly cycle systems (for example, the circulatory or pulmonary sys- of invasion and destruction of nearby tissues that spreads tems) whose fractal-like structures can be shown to opti- throughout the body. Organs become a competitive land- mize energy delivery and resource access. scape where abnormal and normal cells are actors fighting West and his group are now using these same tools to it out for resources. Should the abnormal cells gain the try to develop a physics-based model of tumor growth. upper hand, the function of the organ may be in jeop- Herein the idea is to relate tumor growth to the develop- ardy: a liver that suddenly does not have sufficient ment of the branching system of capillaries, a process healthy tissue to maintain the body’s chemical balance, or called “neovascularization” or “angiogenesis” that is lungs lacking the healthy tissue to absorb the oxygen that responsible for the delivery of energy to the surface of the sustains life or so heavy with tumor that they collapse tumor. Generally, cell survival is linked to proximity to under their own weight. blood supply, so that a better understanding of the for- Within this competitive landscape, the etiology of can- mation and development of these new resource supply cer can take on an evolutionary interpretation. Cells chains for tumor growth is a crucial component in under- reproduce, compete, and evolve with a clear advantage standing cancer. West, former SFI Postdoctoral Fellow (toward an end goal of population dominance) conferred Van Savage (now in the Department of Systems Biology on that cell type that reproduces the quickest. at Harvard University), and SFI Graduate Fellow Alex Evolutionary pressures are also induced by therapies, Herman continue to push this work forward and are, pushing a “natural selection” of those cells resistant to according to Herman, “potentially laying the groundwork treatment. The language of evolution, selection, and com- for theory-based extrapolation of experimental results in petition puts cancer research squarely into SFI’s purview. animal models of cancer to humans.” This allometric framework is also guiding the work of A Universal Law for Tumor Growth other cancer research groups. Of particular note is the When Distinguished Research Professor (and now work of Thomas Deisboeck (Harvard/MIT) who is using Interim President) Geoffrey West came on board at SFI, these ideas as a foundation for investigating tumor devel- he was interested in bringing to the investigation of living opment. The coarse-scale characterization of tumor systems an outlook that was honed over years of study of growth as a rapid (exponential) increase in volume fueled theoretical physics. In that rarified intellectual world of by the concomitant explosion in the number of conduits invisible and indivisible particles, the hallmark of success to sustain efficient blood flow share much with the basic is the principled derivation of universal laws—simple allometric assumptions. By adapting the analysis of West, mathematical formulas like Newton’s “F=MA” or et al., Diesboeck’s group derives an analogous “universal Einstein’s “E=MC2”—that apply with near, if not exact, law for tumor growth.” It is a first step in a back-and- agreement across a range of real-world phenomena. West forth dance between theory and experiment that already wanted no less from his new work in biology. It is in this seems applicable to the design of therapies. Drug design- spirit that he, SFI External Faculty member James Brown, ers can take advantage of the detailed knowledge of the and Brian Enquist, of the National Center for Ecological stages of tumor development. For example, just knowing Analysis and Synthesis, made their first big hit in the the rate at which tumor cells are both generated and lost study of allometry. They took the very physics-like at different stages in development will give the clinician a approach that fundamental principles for growth in any benchmark for the evaluation of therapeutic strategies. living form, be it microbe, marmot, or man, can be Diesboeck, et al., believe that the allometric outlook will deduced from considerations of energy and resource have “far-reaching implications” for our understanding of transport that are independent of the organism. The tumor ontogeny.

Santa Fe Institute Bulletin SPRING 2005 19 Cancer Research In Silico Robert Weinberg (MIT) in which they identify six pheno- SFI Research Professor and University of New Mexico typic cellular characteristics that appear to bear strongly Professor of Computer Science Stephanie Forrest is on malignant tissue formation. Forrest and her collabora- involved in a variety of active collaborations in cancer tors translate these hallmarks into parameters for the sim- research. Her work in this direction seems a natural out- ulation. Exploration of parameter space in the subsequent growth and synthesis of past and continuing achieve- simulations allows the scientists to chart the many possi- ments in computational biology, the study of computer ble paths in the development of tissue as it grows from viruses and their prevention, as well as computational single cell to multicelluar entity, cancerous or otherwise. modeling and simulation. This work has already begun to yield interesting insights “We’re investigating various simple hypotheses for the into how the hallmarks interact. Among these is a new dynamics of resource competition among pre-cancerous point of view on the role of angiogenesis, which is the cells,” says Forrest. Initial work with Carlo Maley of the formation and differentiation of blood vessels. University of Washington’s Fred Hutchinson Cancer This first version of CancerSim is necessarily a simpli- Research Center uses some of the tools of evolutionary fied, highly abstracted model of tumor growth. In an simulation—the same agent-based modeling that came of effort to make it more realistic, Forrest has begun working age in the SFI-led investigations of “artificial life.” with Geoffrey West and Alex Herman, hoping to incorpo- Like any good computational simulation, their work rate aspects of their allometric analysis into a next genera- creates an in-silico laboratory, not just reproducing tion of the computational model. known phenomena, but also suggesting and explaining In order to beat ‘em, you’ve got to change ‘em, and new ideas for therapies. A recent paper with another keep changing ‘em… Hutchinson researcher Brian Reid, investigates the possi- Cancer is life run amok—causing the breakdown of a bility of a new therapy. “Rather than killing off the cancer living system via the hatching of a cell that mutates to dis- cells,” says Forrest, “it instead seeks to boost the repro- play a pattern of uncontrolled growth. The population of ductive fitness of relatively benign cells, thereby allowing mutated cancer cells undergoes a “microevolution” in the them to out-compete the cancer cells in the race for dom- organ. Therein selection pressure favors those cells that inance.” can overcome the “barriers” imposed by immune systems These in silico ideas are finding an in vitro test site in or resource competition. The behavior of the population the investigation of a particular type of esophageal cancer of cancer cells, in essence, breaks the implicit social (and and its precancerous state, “Barrett esophagus,” that arises biological) contract binding together the cellular popula- in a significant fraction of those who suffer from gastroe- tions within the multicellular society that is a tissue. sophageal reflux disease. Forrest and Maley have begun One way in which cancer cells seem to win this the difficult process of tuning their general model to the microcompetition is through an ability to mutate at a data of this particular disease. This has already resulted in tremendous rate. The sheer numbers of subtle variations general insights (in the form of predictions) regarding achieved through this “genetic instability” effect a shotgun genetic factors in cancer development. approach to finding routes through and over the life-pre- Lately, Forrest’s computational approach to the study serving barriers. The minor genetic variations that occur of cancer has acquired two new collaborators, Robert across the range of cancer cells defines them as a “quasi- Abbott of Sandia National Labs and Kenneth Pienta of the species,” a notion first developed by SFI Science Board Comprehensive Cancer Center of the University of member Manfred Eigen and SFI External Faculty member Michigan. Together they have developed CancerSim, a Peter Schuster. new and improved artificial life-inspired Genetic instability is on the one hand computer simulation package for investigat- an advantage in the battle to overcome ing tumor growth. CancerSim aims to “char- the natural barriers, but can also work acterize the processes of cellular alteration against the survival of the “winning” that underlie tumorigenesis.” This three- strain which may simply mutate itself dimensional cellular automata evolves out of existence. External Faculty mem- according to a set of rules born of “The ber Ricard Solé has looked into the Hallmarks of Cancer,” a well-known paper effects of exploiting the long-term disad- written by Douglas Hanahan (UCSF) and vantage of instability. In particular, one

THEY TOOK THE VERY PHYSICS-LIKE APPROACH THAT FUNDAMENTAL PRINCIPLES FOR GROWTH IN ANY LIVING FORM, BE IT MICROBE, MARMOT, OR MAN, CAN BE DEDUCED FROM CONSIDERATIONS OF ENERGY AND RESOURCE TRANSPORT THAT ARE INDEPENDENT OF THE ORGANISM.

20 Santa Fe Institute Bulletin SPRING 2005 idea is to induce increased instability that seems to focus on the guts of the into the tumor cell population. This problem. In their simplified model, the seems to be an especially promising cell is pictured as housing the interac- avenue of therapeutic research for can- tions of three genes: a proto-oncogene cers that operate near an “instability (cancer carrier), a tumor suppressor, threshold,” defined as a level of instabil- and a “housekeeping gene.” As ity close to one in which the tumor Krakauer describes it, the proto-onco- cells would begin to fall apart. gene is the “foot on the gas pedal,” and In work with Dominik Wodarz, of the tumor suppressor is the “brake.” UC Irvine, SFI Research Professor David Ordinarily, keeping “the pedal to the Krakauer has studied the implications of metal” should put the cell on a road to genetic instability in the context of cel- ruin, signaling the beginning of “pro- lular evolution, which selects for the grammed cell death” or apoptosis, but ability to promote angiogenesis. Their what seems to happen in cancer is that models suggest that while genetic insta- the cell acquires the ability to ignore bility is necessary initially, it becomes a such signals and thereby take a joy ride disadvantage in the long run. of never-ending reproduction, whose numbers soon dwarf the population of From Populations to healthy cells. Particulars Separately, the tools of populations The competitive framework in dynamics and genetic networks have which cellular phenotypes fight each each begun to shed light on the com- other for dominance in the tissue casts plex system that is cancer, but as the cancer problem squarely within the Krakauer points out, “No one has yet realm of population dynamics. It is an done a great job of integrating the analysis that assumes that the battle two.” In particular, he believes that one within the body has already begun. As huge open question is, “What is the such, it begs the question of where and right formalism for studying the logic how did the fight begin? What is the of the relevant genetic interactions?” In set of subcellular conditions that gave rise to that first this regard we have begun to see a colony of rogue cancer cells? Our genes contain the basic resurgence of interest in Boolean networks, the forerun- data that provide each cell with instructions for growth, ner of the modern study of genetic networks. These were and Krakauer, as well as many other scientists, have first proposed almost 30 years ago by SFI pioneer Stuart

“RATHER THAN KILLING OFF THE CANCER CELLS,” SAYS FORREST, “IT INSTEAD SEEKS TO BOOST THE REPRODUCTIVE FITNESS OF RELATIVELY BENIGN CELLS, THEREBY ALLOWING THEM TO OUT-COMPETE THE CANCER CELLS IN THE RACE FOR DOMINANCE.” begun to study and model the manner in which various Kauffman (who has just been asked to head a new genes work together (or against one another) in order to Institute for Biocomplexity and Informatics at the gain new insights to the origins of cancer. University of Calgary). Krakauer’s recent work with for- Almost surely, cancer is a multigenic disorder whose mer SFI Postdoctoral Fellow Nihat Ay, has begun to understanding will require the unraveling of a complex develop a formal framework for quantifying the notion of tangle of genetic influences. Many scientists now see can- system robustness in order to enable a rigorous analysis cer as explained, in part, as a problem of aberrant gene of the performance of genetic networks. regulation, an instance of confused information flow both With a little luck and much perseverance, perhaps into and within the cell. Krakauer is quick to point out these research collaborations at SFI and elsewhere will in that many of the mutations known to occur in cancer time evolve toward a better understanding of cancer, gen- development are found in genes involved in signal trans- erating new ideas that will speed us toward a cure, fuel- duction—meaning those genes that mediate the informa- ing a progress of development at an exponential rate, an tion flow among genes within the cell, thereby influenc- ironic intellectual twin of the cancerous phenomena that ing the ways in which proteins are produced. Krakauer this research seeks to erase. and SFI Steinmetz Fellow Sabrina Spencer have begun to Dan Rockmore is professor of mathematics and computer science at study a highly simplified model of the cell, an abstraction Dartmouth College and a member of the SFI External Faculty

Santa Fe Institute Bulletin SPRING 2005 21 22 SFI@20Santa Fe Institute Bulletin SPRING 2005 The Character of Biological Law: An SFI Program Searching for Regularities in the Biosciences

By David Krakauer

“How can it be that writing down a few simple and elegant formulae, like short poems governed by strict rules such as those of the sonnet or the waka, can predict universal regularities of Nature?” Murray Gell-Mann – Nobel Prize Speech, Stockholm, 1969

In the city of Wittenberg, Tycho’s scientific output Germany, in the year 1566, the was considerable: it includes astronomer Tycho Brahe lost the De Nova et Nullius Aevi tip of his nose in a duel with a Memoria Prius Visa Stella (“On COPYRIGHT CORBIS fellow student. In place of a the New and Never Previously : nose, and for the remainder of his life, he wore a Seen Star,” Copenhagen, 1573); De Mundi Aetherei ART sculpted insert made from an alloy of gold and silver. Recentioribus Phaenomenis (“Concerning the New The duel, which was fought with another Danish Phenomena in the Ethereal World,” Uraniburg, 1588); nobleman, Manderup Parsberg, was ostensibly fought and Astronomiae Instauratae Mechanica (“Instruments for over which of them was the more skilled in mathemat- the Restored Astronomy,” Wandsbeck, 1598). The ics. When Tycho was not distracted by a loose nose (it majority of Tycho’s observations were not published frequently fell off, requiring that he carry a small snuff during his lifetime. However, the data were made box concealing an adhesive), or tending a pet moose available to his assistant, the young Johaness Kepler, celebrated for its beer habit, he was refining the instru- who had been recruited to assist Tycho in the calcula- ments that would permit him to collect the most pre- tion of planetary orbits, and who succeeded Tycho as cise and exhaustive archive of astronomical data in the royal astronomer following Tycho’s death from a burst 16th century. bladder at a royal banquet in 1601.

In a famous argument about the probabilistic nature of quan- tum mechanics with Dutch physicist Niels Bohr, Albert Einstein remarked, “God does not play dice with the uni- verse.”

Ken Knowlton, “God Does Not Play Dice With the Universe,” Unretouched Dice, 1999

Santa Fe Institute Bulletin SPRING 2005 23 COPYRIGHT CORBIS : ART

Kepler was a champion of the Copernican, helio- Kepler had to first accept in some fashion a general centric model of the solar system. He had previously theoretical framework without empirical proof spent 20 years unsuccessfully attempting to map the (Copernicanism); second, abandon prevailing scientific motion of the planets onto imaginary spheres encasing dogma (Aristotelianism); third, have access to data five perfect geometric figures: the cube, tetrahedron, with sufficient accuracy to reveal regularities (empiri- octahedron, icosahedron, and dodecahedron. However cism); and fourth, adopt mathematics as an elliptical ingenious his geometry, the data refused to fit the the- and predictive form of expression for the regularities in ory. Presented with Tycho’s high-quality data, Kepler the data (mathematical reductionism). This is a canoni- eventually determined that an ellipse with the sun at cal example of the inductive-deductive recursion, in one focus was a better fit to the trajectory of Mars which theory and data co-evolve towards improved than a circle with the sun at the center. Building on this solutions. In essence, physical theory seeks to explain breakthrough, Kepler proposed his three laws: (1) The the regular component in the behavior of objects, planets move in ellipses with the sun at one focus, through relationships among quantitative variables, (2) planets sweep equal areas in equal times in their and these regularities are called the Laws of Nature. motion about the sun, and (3) the average distance to Typically, regularities are expressed using mathematics. the sun cubed is proportional to the period squared. A challenge for biology in the 21st century is to bring The scientific biographies of Tycho Brahe and the sensibility of Kepler to the vast Tycho-like data Johaness Kepler serve to illustrate some very important resources of biological science, and in this way search themes of scientific progress related to the search for for the law-like regularities in biology. scientific laws. To arrive at his laws of celestial motion,

24 Santa Fe Institute Bulletin SPRING 2005 The Philosophy of Natural Law

One might ask why laws of nature are of interest at While these five conditions might be necessary, all. They are in part because there seems to be a strong they are not sufficient. This is because they cannot dis- correlation between law-like behavior and our ability tinguish among accidental truths, false existentials, and to perform inductive inference (Goodman 1947). In genuine laws of nature. Accidental truths are state- other words, laws help us to generalize reliably from ments such as “DNA and RNA are the universal mole- instances of an empirical phenomenon. For example, cules of biological inheritance.” Biologists have, after Newton’s laws of motion are shorthand for an effec- all, never observed an alternative molecule in nature, tively infinite number of observations on falling apples. and all terrestrial species utilize these molecules. Furthermore, laws when expressed mathematically, Necessary truths are statements such as “no object represent a form of data compression, in which a great having mass can be accelerated beyond the speed of deal of phenomenology can be captured in a terse light.” This statement is not rendered a necessary truth symbolic description (this is what Gell-Mann meant because this prohibition has never been violated, but when he spoke of sonnets and wakas). Laws also serve because it derives from a more fundamental theory of to reveal correlations among different theoretical prin- nature, and moreover, is correlated with a larger set of ciples, for example by finding equivalences among constraints forbidding related behaviors. Philosopher equations (Dretske 1977), and thereby pave the way Karl Popper has argued that laws forbid certain actions, towards superprinciples of nature such as supersym- whereas accidental truths do not. Another way to metry in physics. think about this is that laws are reinforced by observa- The philosophical debate about laws of nature tions, whereas accidental truths are left as unique makes a distinction between two broad schools of instances. Philosopher Nelson Goodman (1947) pro- thought: regularity theories and necessitarian theories. poses that law-like statements can receive confirmation Both regularity theorists and necessitarians agree on from events, whereas accidental statements cannot. the following conditions for a theory to qualify as a A further conceptual problem with laws is their law of nature (see the Internet Encyclopedia of Philosophy): anachronistic quality. Whereas laws aspire to the very highest levels of generality, they are often subsumed within subsequent theories. Newton’s laws of motion are subsumed within Einstein’s theory of relativity, and Boyle’s law can be derived from the later kinetic theory of gases. 1. The theory relates to a factual truth and not a logical one There is a historical component of (SFI is in Santa Fe, N.M., versus every positive integer has a what comes to be called a law corresponding negative value). based on the perceived generality of the theory to the scientific 2. The theory is true for every time and place in the universe. community of any given time. 3. The theory contains no proper names (laws relate to general concepts, not people or places). (continued)

4. The theory makes universal or statistical claims.

5. The theory makes conditional not categorical claims. (Given a certain mass and acceleration, then the force can be deter- mined, versus the statement “mass exists.”)

The necessatarians would append the additional condition:

6. The theory is physically necessary and could be no other way.

This final clause usually implies deriving laws from still more funda- mental principles.

Santa Fe Institute Bulletin SPRING 2005 25 Wigner on Laws of Laws in Living Systems explains. The theory provides—given a Nature and Invariance If we were to provide a short list of set of restricted conditions—a putative Principles laws of physics we might include algorithm through which living struc- tures with some locally adaptive prop- The physicist Eugene Wigner has Kepler’s laws, Newton’s laws, Boyle’s erties might emerge. The theory does suggested that events are organized law, Hooke’s law, Coulomb’s law, not state that these structures must according to the laws of nature, where- Ohm’s law, the laws of thermodynam- emerge; neither does the theory pro- as the laws of nature are themselves ics, Maxwell’s equations, Hubble’s law, vide a sense of the expected distribu- organized according to symmetry or and elements of the theory of relativity tion of structures from which the invariance principles (Wigner 1963). and quantum mechanics. What laws observed solution is drawn. The theory There is a hierarchy of theory, with could we write for living systems that provides us with a strong justification laws subsuming observations, and are not merely diverse manifestations for understanding structure by using symmetry principles encompassing of the underlying laws of physics (e.g., some evolution-independent variational laws. Symmetry, following the defini- conservation principles)? We might go principles, such as engineering or tion of the mathematical physicist with exponential growth without molecular dynamics. It is not the theory Emmy Noether, refers to some quantity resource limitation, Mendel’s laws, the of evolution that tells us how the eye expressed in an equation (such as ener- Hardy-Weinberg law, Gauss’s law of functions as it does; for this we turn to gy) remaining independent of a change competitive exclusion, elements of the optics and neuroscience. in a continuous parameter (such as Darwinian theory of evolution to So while evolution does have law- time). include genetic drift and kin selection, like properties, in that it applies to all Scientific explanations provide a and signaling with excitable media. few simple principles according to Whereas most of the items which properties of an object or event taken from the physics list can be understood. Our theories serve conform to the minimal set to eliminate the element of surprise of criteria required by both and tend to turn all observations into the regularity and necessitar- inevitabilities. Components of behavior ian perspectives, most if not not specified by the laws of nature are all on the biological list vio- called the initial conditions. Wigner late one or more of these writes, “The former are precise beyond items. For example, expo- anything reasonable, the latter virtually nential growth is the out- nothing is known about.” come of first-order rate Wigner asks, “How can we be cer- equations; Mendel’s laws tain that we know all the laws relevant depend upon a fair segrega- to a set of phenomena?” If we are tion, which in turn depends incorrect in our formulation of these upon meiosis, which laws, there is a possibility that the depends upon chromo- number of initial conditions required to somes in cells, and so on specify a behavior would be larger than multiplying contingencies. COPYRIGHT CORBIS is strictly necessary (the regulatarian Darwin’s theory of evolu- : dilemma). One way to determine this tion through natural selec- PHOTO would be to show that initial conditions tion has perhaps the best could be chosen arbitrarily. prospect of acceptance as a organisms with a past—and to a degree Unfortunately, in most cases, this is not biological law. A group of physicists at allows us to reconstruct the past when an option. This problem is particularly the Institute for Advanced Study in coupled with a model of the mechanics present in biology, where many phe- Princeton asked whether evolution of heredity—it lacks the quantitative nomena are the outcome of lengthy should be considered a universal law, power that we have come to expect evolutionary processes, and where such that it too might become a neces- from laws in physics. The question then changing environments introduce a sary outcome of the minimal theory of becomes, is there some fusion of physi- slew of new initial conditions on which “everything” that they were pursuing. cal theory (or some other approach to select. At moments such as these it is worth reiterating what Darwin’s theory such as theoretical computer science or

26 Santa Fe Institute Bulletin SPRING 2005 non-linear dynamics) with evolutionary which theories of functional mechanism represents a blueprint common to life theory that would satisfy this require- —such as those developed by engi- and is therefore suggestive of a path- ment? Is there some way of moving neers—are most appropriate when way through which life need evolve, beyond what theoretical ecologist and exploring biological structures? Can we the fundamental engine fueling the president of the Royal Society, Lord assume that our understanding of the acquisition of energy, and the mecha- Robert May has elegantly described as function of a structure is complete nism responsible for the production of “theories of contingent generality,” and enough, and the constraints sufficiently essential bio-molecules. This project the philosophers of biology, Kim well understood, that we might under- applies the methods of self-organizing Sterelny and Paul Griffiths have called stand an organism in much the same dynamical systems theory to the study “recurrent causal mechanisms”? way that we understand a computer or of thermodynamics in this pivotal reac- Most of the questions that we are a clock? Perhaps the optimality per- tion network. The idea of the project is asking at the Santa Fe Institute in our spective fails when the optimality crite- to employ these universals of biochem- pursuit of the Character of Biological rion is constantly shifting in time, and is istry, and ask whether there are ele- Law attempt to fuse data sets derived itself the outcome of the behavior of mentary physical self-organization from modern collection methods in the system? These problems arise fre- mechanisms under which they are genetics, molecular biology, and behav- quently in the study of niche construc- favored. The new approach to the ior, with new organizational theories of tion. analysis of metabolism proposed here biosystems. Now more than ever, biolo- These are the broad questions that emerges from the physics of self- gy has the benefit of a profusion of large fascinate us and call for a data-driven, organization in thermodynamically data sets, and like Kepler with his access complex systems science. Below are reversible systems. These systems dif- to Tycho’s orbits, we have no excuse but some summary statements of ongoing fer from the dissipative reactions tradi- to search for regularities in our data. Like research projects being pursued at SFI tionally studied in chemical pattern for- Kepler, we have to abandon certain cher- under this theme of laws in biology. mation, and also from the discrete ished ideas and search for an appropriate Those interested in further details are models associated with self-organized mathematical language with which to urged to explore recent publications criticality. present our new findings. and work of the Santa Fe Institute,

A CHALLENGE FOR BIOLOGY IN THE 21ST CENTURY IS TO BRING THE SENSIBILITY OF KEPLER TO THE VAST TYCHO-LIKE DATA RESOURCES OF BIOLOGICAL SCIENCE, AND IN THIS WAY SEARCH FOR THE LAW-LIKE REGULARITIES IN BIOLOGY.

For example, we might ask which which can be found through links pro- B. Energetics and Scaling variational principles should we use to vided at www.santafe.edu. In cellular biological systems, diffu- understand biological phenomena such A. Thermodynamics of sion is insufficient to transport energy as scaling laws, metabolism, or signal Metabolism and nutrients to targets, thus transport transduction? How much generality do Biology has become increasingly systems and pumps are required to conservation principles in physics have aware of the role of energetic con- move energy efficiently. The structure of in explaining typically biological phe- straints in the organization of living sys- these transport systems has a strong nomena? In other words, how much tems. One practical means of pursuing influence on the efficiency of energy empirical variation can physical princi- this issue is through the analysis of the delivery, and powerful consequences on ples account for, and at what level of thermodynamical properties of the the macroscopic, morphological variation detail? When physical principles fail, reductive citric acid cycle. While this is observed among species, their rates of can we discover, using new approach- certainly a terrestrial solution to the growth, their ecological distributions, and es, regularities in information and ener- problem of energy transduction, its even patterns of extinction. This project gy flows in systems with extensive pat- largely invariant features shared by all provides a unifying series of models with terns of feedback? There is some evi- living species are strongly suggestive which to directly address these issues. It dence from the study of genetic regula- of fundamental physical limitations. The is a remarkable fact of biology that the tion that certain network architectures citric acid cycle is made up of the same scaling relationships for mass, M, are recur across distantly related lineages 12 chemicals in all terrestrial species, power laws, yMb, with exponents, b, for reasons of robustness. regardless of their environments and that closely approximate simple multiples Another set of questions concerns their nutrient availabilities. As such, it of 1/4 (e.g., 1/4, 3/4, 3/8). For mass

Santa Fe Institute Bulletin SPRING 2005 27 dependence, three generic principles, ment of phosphates to amino acid tal change and environmental which should be viewed as derivative residues—to transfer information about response, both internally through from natural selection, are postulated: (a) the environment into the response changes in genomic and developmental networks must be space-filling in order to mechanisms of the cell. control systems, and externally, service all local biologically active regions Phosphorylation, in other words, is the through ecological interactions with in an organism; (b) their terminal units, particulate current that flows through other species? Most existing models of which interface with the resource envi- biological circuits facilitated by kinase diversification suffer from a sort of ronment (e.g., capillaries, petioles, mito- catalysts powered by ATP. We examine Goldilocks problem: they are either so chondria, cytochrome oxidase molecules) the way in which catalytic architectures rococo in their development that they are invariant in size; and (c) organisms acquire information about the environ- are of little practical use in distinguish- have evolved so that the energy (and ment, and how kinase networks can ing between different potential possibly other appropriate quantities) act as memory stores. Molecular mem- processes that may be involved in required to distribute resources and sus- ory suffers from a constant degradation diversification, or they are so simple tain them is minimized. Building upon through proteolysis and decay, which that they fail to encompass the com- these assumptions, we show how 15 raises the question, How are robust plexity of interactions between environ- orders of magnitude of variation in basal memories constructed? Furthermore, ment, and intrinsic and extrinsic driv- metabolic rate across diverse phylogenet- biological information needs to be ers. The challenge is to explore a vari- ic groups can be reduced to a factor of extracted from storage for develop- ety of models that 1) reflect a range of 20 in M. We plan to extend the theory to ment. We explore molecular compres- biologically reasonable processes cover 27 orders of magnitude from the sion, protein-based data quantization, which may be involved in the diversifi- terminal oxidase molecules of mitochon- and concurrent processing with mobile cation process; and 2) capture at least dria to the species areas relationships in process algebras. the basics of the dynamics between ecosystems. We ask whether these scal- the physical environment and D. Endogenous Versus ing principles are likely to be universal by genome/developmental change and Exogenous Sources of exploring a diversity of temperatures, net- Innovation ecological interactions. work topologies, and optimality criteria. Evolutionary innovations, like other E. Evolution of Parasitism C. Evolutionary evolutionary changes, reflect the influ- The study of the life cycles of sim- Computation & Cellular ence of genomic and developmental ple organisms such as viruses provides Signaling changes to generate new physiological, a wealth of data on those features of All biological processes make use developmental, and morphological inno- biological organization that remain con- of catalysts to expedite reactions. vations; the success of such events served across groups during evolution, Catalysts are operationally defined as within the physical and biological envi- and those features that vary. Virus life participants in reactions which survive ronment; and changes in the environ- cycles are very well documented, with reactions unmodified, and without ment that may either facilitate or retard many of the genes contributing to the which reactions would proceed at far such success. Here we examine the biological function identified, and their slower rates. Catalysts in biological intersection between organisms and protein products well characterized. reactions range from charged surfaces their environment in relation to some Thus viruses provide unique systems to protein enzymes and nucleic acid significant events in earth history. Of for studying what has been called the sequences. Simple catalysts lower free particular interest is the first appear- “logic of life” (Jacob 1974)—meaning energy barriers, whereas complex cata- ance of multicellular animals at the end that there is a minimal set of basic lysts are a vital source of information of the Proterozoic era closely associat- functions ensuring the persistence of for transforming substrates. The pro- ed with a series of ice ages commonly an evolving lineage. tein kinases are universal transducers called the Snowball Earth. The Viruses are obligate parasites. This of biological information. Prokaryotes Snowball Earth events are a special implies that the hosts with which they and eukaryotes make use of these case of a more general problem: What interact provide information critical to the enzymes—through covalent attach- is the interaction between environmen- completion of the virus life cycle. LIKE KEPLER, WE HAVE TO ABANDON CERTAIN CHERISHED IDEAS AND SEARCH FOR AN APPROPRIATE MATHEMATICAL LANGUAGE WITH WHICH TO PRESENT OUR NEW FINDINGS.

28 Santa Fe Institute Bulletin SPRING 2005 Different virus groups differ with regard and consequently make discovering and socialization. to the amount of information they law-like behavior difficult. These con- The sciences of humanity are the require from the host for completion of siderations are even more significant most recent to adopt the quantitative the life cycle. Another way of stating this when we turn to the social sciences. and compression-based approach of is to recognize that the host represents Some of the defining, positive fea- mathematical and computational theo- a catalyst of varying complexity, facilitat- tures of the 20th century were scientif- ry. Arguably those of us working in the ing the replication of the virus genome. ic discoveries (antibiotics, electrical natural sciences have a vested interest The viruses therefore provide us with an refrigeration, computers, the moon in their success. It is, after all, social ideal system with which to study the landing, fertility treatment, open-heart institutions and their occupational net- evolution of autonomy—an increasing surgery). This local progress is often works that fostered and now perpetu- independence from environmental vari- contrasted with the persistence of ate the scientific revolution. ables, and an increasing dependence on global social prejudices and economic factors encoded intrinsically. inequality. Even though natural science References and Further Reading In this project, we explore what we has often been misappropriated might call catalytic grammars of virus towards undesirable social purpose, 1. Armstrong, D. M. What is a Law of life cycles. We denote them as gram- there is little disputing that scientific Nature? Cambridge University Press, mars because one of the essential understanding has made progress. Cambridge. 1983. ways in which viruses differ from one The obvious answer to these ques- 2. Carroll, J. The Humean Tradition. The another is in the sequencing of their tions is that human society is far more Philosophical Review, 99. 185-219. 1990. basic replicatory, translational, and reg- complex than the non-human, natural 3. Dretske, F. Laws of Nature. Philosophy of ulatory operations. We consider a world, and human planning and under- Science, 44. 173-181. 1977. range of virus life cycles, and their cor- standing makes for individual variation 4. Ferguson, K. Tycho and Kepler. The responding host contributions, in an that disrupts our best attempts at dis- unlikely partnership that forever changed attempt to build a kinetic theory of covering cross-cultural statistical regu- our understanding of the heavens. Walker virus diversity based on regulatory vari- larities. The factors listed at the start of and Co. 2002. ation. This approach differs from stan- this section are impediments to 5. Goodman, N. The Problem of dard taxonomies of viruses, which use progress. Human culture is so combi- Counterfactual Conditionals. Journal of DNA data, or genomic organization, as natorially rich and self-referential that Philosophy, 44: 113-128. 1947. a criterion for classification. We use prediction becomes ineffectual. 6. Lewis, D. Counterfactuals. Harvard the kinetics of regulation to explore While the complexity defense cer- University Press, Cambridge, MA. 1973. invariant principles found among all tainly is part of the story, it does not 7. Ramsey, F. Foundations: Routledge and translational parasites. imply that we shall never discover reg- Kegan Paul, London. 1978. ularities in behavior and culture. 8. Wigner, E. P. Events, Laws of Nature and A Concluding Unscientific Indeed, structural anthropology, com- Invariance Principles. Nobel Lecture, Postscript parative linguistics, and developmental December 12, 1963. — with Apologies to Kierkegaard psychology pursued exactly this objec- In starting with physics and moving tive before falling foul of relativism. David Krakauer is a research professor into biology, we have observed that the The Santa Fe Institute is, in parallel and at SFI. extent to which science has been suc- in conjunction with the natural sci- cessful at discovering mathematical ences, actively pursuing regularities in expressions capturing regularities has the social sciences through the use of been graded. This increased difficulty quantitative models and theories. in biology in comparison to physics, These include programs looking at very reflects several factors, including the particular regularities in financial mar- contributions of non-linearities, the kets, to more general regularities found number and variability of initial condi- in all human societies, such as cooper- tions, the constancy of the environ- ation among unrelated individuals, the ment, the degeneracy of stable states emergence of institutions, the multiple for a given set of boundary conditions, network structure of social systems, and the size of the behavioral reper- the course of cultural and linguistic his- JULIE GRABER toire of individual units. In biology all of : tory, and the neural basis of learning these factors play an important role, PHOTO

Santa Fe Institute Bulletin SPRING 2005 29 A Complex Look at Sports

By Matthew Blakeslee

Thirty years ago if you’d want- ed to invest your money, you would have probably entrust- ed it to a cigar-smoking, gray- haired sage who’d been a Wall Street player for decades. He would have been a big node in the old boys’ network, steeped in the traditions and lore of stock trading that dated back, with only minor mutations, to the days of the robber barons. True, he would base his deci- sions in part on “the num- bers,” but he would base them equally as much on his person- al relationships, the rules of thumb inherited from his men- tors, and on gut instinct.

ILLUSTRATIONS: SPENCER LEVY

30 Santa Fe Institute Bulletin SPRING 2005 But in the 1990s things started to Ranking Teams— Massey said. In addition, he discussed change. Enter stage left the physicists, The Worst Way is to some of the ways even the good mod- the modelers, the geeks bearing sili- Have a Tournament” els can go wrong: sometimes they lead con, offering complex analytic methods Ranking teams is a messy exercise, to strange rankings that put bad teams for dicing up market dynamics in pow- said Ken Massey, a visiting professor of over good, or even more absurdly, rank erful new ways. These innovations mathematics at Hollins University. He teams in circular relation to one anoth- paved the way for the revved-up finan- is in a position to know. Massey pub- er, the same way paper beats rock cial world of today, with its new lishes the well-known Massey Ratings beats scissors beats paper. money-making schemes, such as cur- on his web site, primarily as a hobby. “The worst way to figure out the rency speculation and time-series fore- These ratings are used as part of the best team is to have a tournament,” he casting, which were inconceivable back formula for the BCS (Bowl said. in the days of the three-martini lunch. Championship Series) Rankings each And yet, that is what’s called for. In June 2004, a genial assortment year to determine which college foot- Massey said that at best his maximum of business people, sports luminaries, ball teams should face each other in likelihood model is able to predict and academics convened in Los the NCAA championship bowls. roughly 75 percent of games in base- Angeles to discuss whether the same Massey outlined several kinds of math- ball and about 66 percent in the NFL. sort of revolution might be brewing in ematical models that people have used The rest, he said, is where all the fun the world of sports. The meeting was to try to capture the dynamics of teams of watching and arguing about sports is underwritten by SFI Business Network or other institutions competing for the found. members Legg Mason Funds same prize. These included classical lin- Scott E. Page, SFI External Faculty Management Inc. and Credit Suisse ear statistical methods, Markov Chains member and a complex systems pro- First Boston, and hosted by the Trust (which are the crux of the Google fessor at the University of Michigan, Company of the West. The question on search engine), maximum likelihood agreed that linear models can never the table: Can game the- models (on which the Massey capture the complexity of interactions ory, complexity science, Ratings are based), and neural inherent to team sports, and spoke of or even neuropsychology networks. a better way to search for useful pat- lend useful insights into how to build “People are often confused with terns that human intuition can’t better teams, rank teams and value the BCS Rankings,” Massey said, spot. He calls it the General Manager’s players, and predict tournament out- “because they think that when the Backpack Problem: given a finite budg- comes better than the simple methods championship game is played, that’s et, your job is to look at players on a that are still the industry mainstay? the definitive answer as to which is the few key dimensions (pitch speed, bat- Many of the participants were opti- best team.” ting average, forty-yard dash time, or mistic that the answer is yes, and glim- But team sports are so high-dimen- similar criteria) and put together the mers of this new method do indeed sional, he said, the very idea of creating best team you can through an auction seem to be appearing. For instance, a one-dimensional list of teams in order competing against other teams. much on everyone’s lips was Michael of merit, not to mention the notion of “It’s a nightmare problem,” Page Lewis’s 2003 bestseller, Moneyball. singling out the team that is “best,” is said. “It’s beyond NP hard!”—which is The book follows the story of the inherently artificial. To make matters an engineer’s way of saying that in Oakland Athletics as they came up with more difficult, the data sets in sports practice you can never solve it no mat- a new and unconventional set of yard- tend to be very small and very noisy. ter how big a computer you sic on it. sticks for valuing players, and, in the Luck, weather, and injuries play big This brought Page to the center- process revolutionized the way baseball roles in game outcomes. So do hard-to- piece of his talk, the simple strategy is managed. With only a shoestring quantify psychological factors such as game called Colonel Blotto. In Colonel budget, the A’s, as they are known, put slumps and home-field advantage. Blotto, each of two opposing sides is together a first-rate team using players Even the random ordering of team given an army of equal size and must who had been passed over by all the face-offs in a tournament can introduce compete for control over a set number big shots of the Major Leagues. statistical artifacts into the rankings, of territories. Each side assigns any IF AN NFL COACH WERE TO GO BY ROMER’S CHART AND RUN ON THE FOURTH DOWN AND FAILED, IT WOULD PROBABLY BE HIS LAST SEASON COACHING, NEVER MIND THAT IT MAY BE A SUPERIOR STRATEGY OVER THE LONG RUN.

Santa Fe Institute Bulletin SPRING 2005 31 number of available troops to each ter- men affirmed that many of sports’ ritory. After all the troops have been standard procedures and strategies placed, the total number of troops in exist mainly through momentum and each territory is revealed to both sides. tradition. They also agreed that even if To Punt or Not to Punt The side with the greatest number of some of these traditions are exposed David Romer, a professor of political troops in a territory wins it. as non-optimal, supplanting them will economy at U.C. Berkeley, then turned In its basic form, there is no best be difficult. to the subject of punting. Specifically, strategy for winning Colonel Blotto; it is “In college football, people are he’s analyzed the common decision overwhelmingly a game of luck. reluctant to try new things because you faced by NFL coaches of whether to “ONE WORD HAS BEEN LEFT OUT IN ALL OF THIS SO FAR,” HE SAID: “EMOTION. FOOTBALL IS A GAME OF EMOTIONS, NO QUESTION.”

punt or go for the first down. Romer has However, Page showed how through never want to lose,” said Chow. “We programmed a complex model that ana- introducing externalities (such as inter- have to treat every game as if it were lyzes the vast, chess-like tree of contin- actions or nonlinearities) into the game, the Rose Bowl,” he added, echoing gencies sprawling before the future of thereby rendering it more complex, Romer, meaning that they are so busy any decision of whether or not to punt. meaningful strategy options emerge. exploiting a good rule that they are not Romer used the model to derive a graph As he put it, “Complexification (of exploring anything better. prescribing when a team should kick the strategic environment) leads to DePodesta echoed this in his dis- and when it should “go for it.” simplification (of the strategies).” cussion of the “management turnstile” Romer argued that his model For example, if it turns out that con- so notorious in professional sports. If an shows that NFL football teams are trolling France and Germany also gives NFL coach were to go by Romer’s chart punting much too often; they are fol- you extra “bonus” influence on and run on the fourth down and failed, lowing imitative rules of thumb that Luxembourg and Belgium, you have a it would probably be his last season coaches have always used that are strong strategic incentive to invest coaching, never mind that it may be a strategically poor. This pack mentality troops in those two countries. Negative superior strategy over the long run. also exists in markets, causing financial externalities are also possible: for According to DePodesta, almost bubbles. For instance, his graph shows instance, placing more troops in a partic- everyone in sports, from the fans to that if a team has fourth down and one ular region might paradoxically destabilize the general managers to the investors, yard to go on their own 10-yard line— it by stoking native resentment. has an extremely short focus: this 90 yards from the end zone—they Discerning and exploiting such nonlinear- year’s championship. should nevertheless go for the first ities is key to forming useful strategies. “There is no excuse accepted for down. This runs counter to the conven- By analogy, Page said, the General not winning,” he said. “The ends justi- tional wisdom, which holds that the ball Manager’s Backpack Problem becomes fy the means, as far as most sports should be sent as far away from the more tractable with a more complex people are concerned.” goal line as possible, even if it means model of players. DePodesta, who had been hired to giving the ball to the opposing team. “Instead of simply summing up a improve what had been a lackluster Romer estimates that by following his player’s vector of attributes,” he said, Dodger club, said he has found little model (and assuming that the opposing “look for interactions between them.” enthusiasm for a team development team doesn’t follow this strategy), the He argued that this should make process that takes the long view—say, average NFL team would win one more the problem of picking players and sacrificing the likelihood of grand victo- game per season than it does now—a building an effective team around a ries for several years in order to culti- considerable gain. well-thought-out strategy easier, not vate some well-chosen talent. The next two speakers were stars harder as one might expect. He also “Our obsessive focus on outcomes of the sporting world: Norman Chow, said it could lead to a more interesting at the expense of process is counter- offensive coordinator for University of “meta-game,” in which different man- productive,” he continued. “Teams that Southern California, and Paul agers pursue a range of finely minced have good process, and patience, often DePodesta, general manager for the strategies and counterstrategies. do better. But few teams are willing to Los Angeles Dodgers, and before that, do it because of the mindset that you one of the stars of Moneyball. Both

32 Santa Fe Institute Bulletin SPRING 2005 have to win a World Series or you’re with management. He predicted coach- use the information they possess in nothing.” es would soon follow suit. predictable ways. Essentially Camerer was proposing Basketball or Moneyball Emotions in the Game that a better understanding of people in Dean Oliver, author of the book Colin Camerer, a Caltech economist neuropsychological terms might be able Basketball on Paper, is the creator of who studies the cognitive basis of eco- to supply analysts with new variables RoboScout, a software program that nomic decision making, strategic think- and insights for structuring their mod- watches basketball games and ana- ing, and risk taking—in a word, “neu- els. For example, DePodesta suggested lyzes them with a keen statistical eye. roeconomics”—spoke of the need to that someday it may become possible In much the same way the Oakland A’s inject psychology into game theory. to “brain type” players in order to bet- crunched the numbers to find new, bet- Game theory involves the formal analy- ter predict what kind of training is best ter criteria for valuing baseball players, sis of situations where completely for them and how they will react to RoboScout looks for—and apparently rational individuals strive to maximize pressure or deal with success. finds—factors that can better predict their own gains by competing or coop- People like DePodesta, Oliver, and the outcome of basketball match-ups. erating with others according to an most of the other speakers believe it Oliver used the specific example of established set of rules. The theory should be possible to develop a more the Los Angeles Lakers versus the captures certain aspects of economics rigorous science of what leads to suc- Detroit Pistons NBA finals in 2004. First and strategy extremely well, Camerer cess in sports. This generated discus- he listed the most common rationales said, but is deafeningly silent on all sion regarding the emerging synthesis people had given before the finals for sorts of other factors—things like pride, between pro sports and science, and favoring the Lakers at eight-to-one. herd mentality, and emotional attach- the role that SFI might play in advanc- These included statements such as, ment—that influence economic and ing that research agenda. Sports may “The Lakers can turn up the heat any strategic decisions in real life. seem an odd research subject—many time they want to,” “They have the USC’s Chow touched on this issue think it’s not serious—but it is awash in two best players,” and so on. Next he earlier in the day. “One word has been data, and the rules are well defined. It ran through the rationalizations people left out in all of this so far,” he said: is a good environment for researchers had offered after the Detroit Pistons “Emotion. Football is a game of emo- to cut their teeth, providing insights won, such as, “The Lakers didn’t want tions, no question. The human element into complex systems with human it badly enough,” and “Detroit con- is so important…These are twenty- actors. Sports metaphors are often trolled the tempo.” year-old kids playing in front of a crowd invoked by politicians and business Oliver quipped at this point, of fifty thousand people. Some of them people, and there’s a reason. Features “Everybody has to be right both before are afraid, some of them are nervous, of the sporting world: heterogeneity, and after.” some of them are giddy. Some players strategy, adaptation, creation and He then went on to deflate all of choke under pressure while some real- implementation of rules of thumb, the those arguments, calling them superficial. ly thrive.” Managing all that is at least rise and fall of dynasties (think Rome, “Both of these sets of reasons are as important in his job as coordinating think Enron, think Cowboys), also apply immeasurable,” he said, “and therefore plays, he said. to the business and policy worlds. unmanageable. If you can measure it, Camerer agreed, commenting, “A you can manage it; if not, not.” five hundred-page book on game theo- In analyzing the same games, ry won’t even have ‘emotion’ in the Roboscout came up with several non- index. And here’s the coach telling us standard observations that would have it’s the most important thing.” predicted the Pistons’ victory. For According to Camerer, game theory example, RoboScout observed that the not only leaves out emotion, it also Lakers couldn’t get easy buckets, the assumes things about people that Matthew Blakeslee is a Pistons forced the Lakers inside, and aren’t true. For example, game theory freelance writer the Pistons controlled the offensive assumes we all chase the logical con- living in California. glass. (You can read more about clusions of our strategies to infinity, RoboScout at www.82games.com.) while in reality most people can only Oliver said that tools such as think a handful of steps into the future, RoboScout aren’t used much yet, if that. He also documented the fact although they are starting to catch on that people often misapply or under-

Santa Fe Institute Bulletin SPRING 2005 33 The Logic of Diversity The Complexity of a Controversial Concept

By Cosma Shalizi

34 Santa Fe Institute Bulletin SPRING 2005 Over the last fifty years or wanted a diverse student body, Scalia that are strongly interdependent. so, diversity has joined motherhood said, it simply had to “lower the stan- Because of that interdependence, it’s and apple pie as something everyone dards” and admit a certain number of hard to modify just one aspect of a

embraces, at least in public. But many incompetents. solution at a time—if you try to COURTESY CORBIS : who would never say publicly that Such views understandably infuriate improve one thing, you’ll often end up

there is only one correct way to do many people, who nonetheless can’t breaking several others. The idea of a PHOTO things and one admirable type of per- say just what’s wrong with them. This “search landscape” is a convenient son, express such feelings in private, at is where the work of SFI External way of visualizing this. Picture potential least about what’s important to them. Faculty member Scott Page and his col- solutions as points on a relief map; the For them, diversity may be okay, even laborators comes in, by showing that height of the map at each point indi- good, for the recreational, shopping- diversity can actually help find the One cates the quality of the solution. The mall side of life—food, clothes, music. Right Answer. Page, a political scientist fact that the problem is complex, with But when it comes to the serious work and economist at (coincidentally) the many interdependent aspects, corre- of the world, there’s One Right University of Michigan, has drawn on sponds to the landscape having many Answer, and what counts is getting as ideas from the study of complexity to local peaks. (Search landscapes come close to it as possible. Diversity seems outline what he calls “the logic of diver- from evolutionary biology, where ones beside the point, if not a liability. To sity.” Just as classic work in political like this are called “rugged.”) Even if such people, the usual platitudes about economy established the “logic of col- there’s a unique highest point on the diversity sound like obvious, if perhaps lective action” (Mancur Olson) and the landscape—a single optimal solution or well-intentioned, nonsense. Such atti- logic of social choice (SFI External right answer—it can be hard to find. tudes do sometimes surface in public. Faculty member and Nobel laureate For large, industrial-strength problems, In the recent Supreme Court case Kenneth Arrow), Page thinks he has the only way to find the optimum may about affirmative action at the found the basic rules explaining how be to enumerate all the possible solu- University of Michigan’s law school, diversity works in society, and complex- tions, which is prohibitively time-con- Justice Antonin Scalia opined, during ity science plays an essential part in his suming. So, in practice, any agent try- the oral arguments, that the only rea- explanation. He shows that not only can ing to solve such a problem will have to son there was an issue was that diversity be helpful in finding good solu- employ heuristics, tricks or short-cuts Michigan insisted on having a “super- tions, but it can even be more beneficial that may work on particular problems, duper law school” (as he put it) that than individual competence. but can’t be guaranteed to always find was also diverse, assuming that these Start with the idea of a complex the right answer. were incompatible goals. If Michigan problem—one that has many aspects Closely associated with this, agents

Santa Fe Institute Bulletin SPRING 2005 35 Cartogram of the results in the 2004 U.S. presidential election. Each county is shown with an area proportional to the number of major-party voters, and with a color that reflects the proportion of those voters going to the two parties—bluer counties voted more Democratic, redder counties more Republican. Figure by Michael Gastner, Cosma Shalizi, and Mark Newman, using a cartogram algorithm developed by Gastner and Newman, and a color scheme proposed by Robert Vanderbei. will have particular “perspectives” on bines age and size in guessing mainte- tions, a diverse population of agents, the problem, paying attention to some nance costs. Conversely, every heuris- each with a different weak heuristic, aspects of it and filtering out all the tic has an implicit perspective, because will outperform a single agent with a others. If an agent faces two problems it responds to some aspects of the very strong heuristic—as Page and that, from its perspective, look the problem but not to others. Hong say, “Diversity trumps ability.” same, it will attempt the same solution A weak heuristic is one which One way to grasp Page and Hong’s to both, even if it notices differences comes up with solutions to the prob- result is to imagine the diverse but between them. From its perspective, lem which are, on average, only a little inept agents taking turns at the prob- everything important is identical. If the better than one would expect from lem, each one starting from where the problem is to estimate the value of a chance. These tend to be heuristics last one got stuck. Each of them tends building, one agent might look at just which get stuck near local peaks in the to get trapped at local peaks, but, its location and size. It will then guess search landscape, and can’t get out of because they’re diverse agents, they similar values for similarly situated those traps, never reaching the opti- get trapped at different peaks. By using buildings, regardless of, say, age—it mum. A strong heuristic, on the other each other’s work, the group avoids doesn’t “see” age. Because perspec- hand, is one that gives a solution that is these local traps, and gets arbitrarily tives filter out some aspects of the generally nearly as good as the actual close to the optimal solution—closer problem, they limit the kinds of heuris- optimum. Counterintuitively, Page, with than any given individual agent with a tics agents can use. For example, if an long-time collaborator Lu Hong, profes- strong heuristic. agent doesn’t “see” a building’s age, it sor of finance at Loyola University, has Remarkably enough, one doesn’t can’t use a valuation heuristic that com- shown that, under very general condi- get the same improvement from using

IN THE END THE LOGIC OF DIVERSITY EXPLAINS WHY DEMOCRACY IS SO HARD, AND SO NECESSARY.

36 Santa Fe Institute Bulletin SPRING 2005 a diverse population of agents with ing independent guesses, you will be The division of labor is, in part, an strong heuristics. The reason is that the puzzled about situations where it works adaptation for handling complex prob- strong heuristics all tend to be similar when people’s guesses are dependent. lems, but only those which are com- to one another—they know the same In particular, you’d expect that if each plex in the straightforward sense of tricks, as it were—and so tend to get person’s guess is strongly dependent being very large. It relies on finding a stuck on the same local peaks. on the last person’s guess, then the way of decomposing the large problem Because they’re all good in the same group should fail miserably. But in Page into nearly-separate parts, so that it can way, they have no ability to compen- and Hong’s model, remember, each be attacked through a strategy of sate for each others’ weaknesses. “If agent starts from where the last one divide-and-conquer, with different peo- the best problem-solvers tend to think got stuck, so its guess does depend ple specializing in conquering the vari- about a problem similarly, then it strongly on the previous agent’s. Yet ous divisions. (This topic, and its rela- stands to reason that as a group, they these groups not only don’t do badly, tion to hierarchical structure, was may not be very effective,” Page says. they do better than they would if each explored by Herbert Simon in his clas- It’s worth noting that this isn’t just a agent acted independently of the oth- sic Sciences of the Artificial.) Diversity, trick with averaging. A new book, The ers. So while Surowiecki’s idea is not in the sense Page is talking about, is Wisdom of Crowds: Why the Many Are wrong, it’s incomplete. another way of adapting to complexity, Smarter than the Few and How It’s also important to distinguish this and specifically to complex problems Collective Wisdom Shapes Business, idea from the division of labor, even the which are not decomposable into neat Economies, Societies and Nations, writ- division of cognitive labor. Big engineer- hierarchies. ten by The New Yorker columnist ing projects (say, designing a new jet), Put strategically, the idea is like this: James Surowiecki, has recently popu- are often broken down into modules Agents have only a limited capacity to larized the idea that groups can, in which are nearly self-contained, with represent, learn about, and predict their some ways, be smarter than their well-defined interfaces connecting world, and so solve their problems. members, which is superficially similar them. This means that someone can When the problem or environment is to Page’s results. While Surowiecki work on the autopilot, without knowing too complex for any one agent, then gives many examples of what one all the details of the engines, just some you should have many weak agents might call collective cognition, where of its interface properties (for example, make partial, incomplete, overlapping groups out-perform isolated individuals, the maximum thrust it can deliver), and representations. You’ll be better off by he really has only one explanation for someone else can work on the engines doing this and then learning a way to this phenomenon, based on one of his without knowing the details of the combine them, than by trying to find a examples: jelly beans. In a long series autopilot. This in turn means that engi- single, globally accurate representation, of experiments, students in psychology neers can specialize in control or such as a single super-genius agent that classes are asked to guess, for exam- power, solve their individual problems, can handle the problem all by itself. ple, the number of jelly beans in a jar. and then put their individual partial solu- Collectively, the combined representa- Quite reliably, the average guess of all tions together, with some hope of the tions of the group of agents are equiva- the students in the class is closer to result working. Often some tweaking is lent to a single high-capacity represen- the real number than the individual needed, because the interfaces don’t tation. But nobody, individually, has any- guesses, sometimes astonishingly perfectly encapsulate the different thing like the complete picture; in fact, close. The reason this works, modules, but much less than if the everybody’s individual picture is pretty Surowiecki says, is that averaging project wasn’t broken up this way to much wrong, or at best drastically together many independent, unbiased start with. Specialization is a way of incomplete. guesses gives a result that is probably using diversity, because control and Powerful, high-level capacities closer to the truth than any one guess. power engineers do learn to think dif- which emerge from the interplay of While true—it’s the central limit theo- ferently (as anyone who’s had to coor- low-level components are a common rem of statistics—it’s far from being dinate both can tell you). But this isn’t feature of complex systems, but here the only way in which diversity can be what’s going on in Page and Hong’s as elsewhere, just having the compo- beneficial in problem solving. set-up, where the team is dealing with nents and letting them interact is not If you think the only way that collec- a single, undivided, perhaps indivisible, enough. The organization of the interac- tive cognition can work is through pool- problem. tions is crucial. In the brain, for “DIVERSITY TRUMPS ABILITY.”

Santa Fe Institute Bulletin SPRING 2005 37 instance, this is the difference between grasp, it’s disconcerting to think this tain conditions of rationality and equali- coherent thought and delirium, or even way about something like a scientific ty, it is impossible to guarantee that epilepsy. In distributed problem solving, discipline, or about more formal institu- societal preferences will correspond to social organization is the key to realiz- tions, such as governments and busi- individual preferences when more than ing the potential benefits of diversity, nesses. Or, for that matter, law two individuals and alternative choices and avoiding mutual incomprehension schools. In a provocative mood, Page are involved. or socially-amplified folly. Improving suggests that Scalia “got things com- The way to avoid the impossibility organization raises performance in pletely backwards.” Given the complex theorem is for people in the group to diverse groups by making it easier for interdependencies of the problems we agree in their preferences (or not dis- the agents to utilize each others’ abili- ask the legal system to resolve, it agree too much). One way to achieve ties and efforts, which can be more might well be that diversity is the only this is to limit the membership to those important, as we’ve seen, than improv- way to excellence, that “ability is diver- with the “right” values, but this will in ing those individual abilities. Page and sity, and to say there’s a trade-off is in turn reduce the diversity of heuristics Hong’s model shows, in a sense, how some sense to misunderstand the and perspectives. Such groups may well the group could do with the right nature of ability.” find it easy to decide what to do, but organization, but not how to find that Which is not to say that diversity they’re ineffective at doing it. The way structure. has no drawbacks. Diversity of heuris- to preserve diversity is to reach agree- When political scientists, say, come tics and perspectives tends to be linked ment on preferences. This could be up with dozens of different models for to diversity of values and interests. either by reaching new, shared values, predicting elections, each backed up by This, as Page says, is where things get or by crafting compromises which sat- their own data set, the thing to do tricky. We’ve been assuming everyone isfy divergent values. These are the might not be to try to find the One agrees on what makes a solution good cornerstones of democratic delibera- Right Model, but instead to find good or bad, that everyone shares a com- tion. In the end, perhaps, the logic of ways to combine these partial, overlap- mon set of interests. But, unless every- diversity explains why democracy is so ping models. The collective picture body values exactly the same thing, hard, and so necessary.

WHEN POLITICAL SCIENTISTS SAY, COME UP WITH DOZENS OF DIFFERENT MODELS FOR PREDICTING ELECTIONS, EACH BACKED UP BY THEIR OWN DATA SET, THE THING TO DO MIGHT NOT BE TO TRY TO FIND THE ONE RIGHT MODEL, BUT INSTEAD TO FIND GOOD WAYS TO COMBINE THESE PARTIAL, OVERLAPPING MODELS. could turn out to be highly accurate, and receives exactly the same share of Cosma Shalizi is a postdoctoral fellow at the even if the component models are bad, what the group gets, this won’t likely University of Michigan’s Center for the and their combination is too complicat- be the case. Diverse groups, good at Study of Complex Systems. ed for individual social scientists to solving problems, will tend to be ones grasp. whose members have diverse ideas It’s already widely appreciated that about which problems they ought to markets perform this kind of distributed solve. Should a police department problem-solving. No individual in the catch criminals, deter potential crimi- market can grasp all the information nals, reduce the harm done by crime, about goods and services in every or get the police chief reappointed? economy, much less search over allo- The study of how to aggregate differing cations to ensure that supply and agents’ preferences into a common demand balance. But the market as a collective choice brings us back to whole not only finds such allocations, it Kenneth Arrow’s “logic of social adjusts them as conditions change, and choice” mentioned earlier. does so by using the diverse local This is bad news, because the main knowledge of the participants. Even thrust of the logic of social choice is though it’s appreciated that markets Arrow’s “impossibility theorem.” The can solve problems individuals can’t gist of the theorem is that under cer-

38 Santa Fe Institute Bulletin SPRING 2005 TRANSITIONS

Crutchfield and Fontana plan to continue this productive association Accept UC Davis and Harvard and expect that both will contribute substan- SFI Community Mourns Death Appointments of Lee Segel tively to SFI’s future development as SFI Research Professor Jim External Professor Lee A. Segel, of External Faculty professors. Crutchfield moved in late 2004 from the Weizmann Institute of Science, passed SFI to California as a professor of physics at SFI Names International away on January 31, 2005. His brief intense the University of California, Davis, and to Fellows from Argentina, battle with illness came unexpectedly, dur- Columbia, and Mexico help create that campus’s new Center for ing what had been the prime of an early The International Program has named Computational Science and Engineering retirement. three new International Fellows to (CSE). Crutchfield has been associated with Segel did his undergraduate work at 2004–2006 terms. Fellows are invited for SFI since 1991 working in the areas of non- Harvard, and received a Ph.D. in applied short-term visits to the Santa Fe Institute linear dynamics, solid-state physics, critical mathematics from MIT in 1959. After two where they have the opportunity to partici- phenomena and phase transitions, chaos, years in the Aerodynamics Division of the pate in the Institute’s many educational pro- and pattern formation. His research at SFI National Physical Laboratory in England, he grams, workshops, and symposia. focused on computational mechanics, the joined the Mathematics Department at Miguel Fuentes is currently a physics of complexity, evolutionary theory, Rensselaer Polytechnic Institute. Until 1968, researcher at CONICET statistical physics machine learning, and distributed intelli- Segel worked mainly on problems in nonlin- group at Centro Atomico Bariloche in gence ear stability theory, but his major focus Argentina. He will initially be working on a CSE’s founding director is SFI alum John switched to theoretical biology starting with project with David Krakauer on mathemati- Rundle. The new CSE at Davis will develop a sabbatical at the Sloan-Kettering Institute cal models in biological systems. Fuentes into a graduate program in complex systems and Cornell Medical School in 1968–9. He was a visiting fellow with the Consortium of over the next three to four years and then a joined the Department of Applied Americas for Interdisciplinary Sciences at full academic department over the next six Mathematics and Computer Science of the the University of New Mexico and attended to seven years with graduate and undergrad- Weizmann Institute in 1973. lectures at SFI. uate curricula in scientific computation and He was one of the earliest promoters of Francisco Sanin is a political sci- complex systems; we look forward to see- the need for close contact between theoreti- entist from the Universidad Nacional de ing CSE students here at SFI in the future! cal and experimental biology and was a fore- Columbia specializing in the study of political In September 2004 Research Professor father of the field now known as theoretical violence and, in particular, the microfounda- Walter Fontana moved from Santa Fe immunology. At his death, he was director tions of civil wars. Francisco is currently col- to join Harvard University as a professor of of the Institute’s annual Summer Workshop laborating with SFI Resident Faculty mem- systems biology at the newly created on Mathematics and Biology, a NIH-support- ber Elisabeth Wood and will be combining Department of Systems Biology at the ed program that explores concrete exam- visits to SFI with research visits to Yale Harvard Medical School. Fontana’s research ples of how mathematical modeling can University. focuses on novelty in evolution including improve biological understanding. Jorge Velasco-Hernandez is a RNA folding and evolutionary dynamics, Segel wrote many papers in mathemati- mathematician currently resident at the Abstract Chemistry, and self-rewiring signal- cal biology and was the author of a handful of Instituto Mexicano del Petroleo. He has ing networks. He notes, “I first set foot on books based on his teaching at the Weizmann broad interests in mathematical models of SFI’s premises in 1989, 15 years ago. Since Institute. His work led to the creation of new population dynamics, ecology, and epidemi- then I have had a virtually uninterrupted mathematics and computational tools for ology but will be focusing his time at SFI sequence of associations with this place or investigating riches of biological behavior and working with SFI External Faculty member state of mind, as Harold Morowitz aptly put his research on pattern formation and mor-

Simon Levin on epidemiological issues such ANNE LEVY it—first as a “hang-out while at LANL, then phogenesis has been seminal in launching a : as mixing patterns among individuals, inter- as a postdoc, as External Faculty and as burgeoning field at the intersection of biology, actions among strains, and spatial aspects Resident Faculty.” mathematics, and computation. of disease dynamics. Both Walter and Jim have made seminal ILLUSTRATIONS contributions to the Institute’s research. We

Santa Fe Institute Bulletin SPRING 2005 39 TRANSITIONS TRANSITIONS TRANSITIONS

TRUSTEES clients including Warburg Pincus, SFI’s Board of Trustees is drawn Andersen Consulting, EDS, TRW from leaders in business and finance, Corporation, and numerous high- the academic world, and the public technology companies and private sector. Here are the newest addi- equity investment firms. tions to an accomplished roster: Murdock currently serves as a Christopher Davis, Davis director of Quest Software, Peace Selected Advisors, has more than 15 Software, CallWave, Inc., Dorado years experience in investment man- Software, Inc., KWI, Inc., and Digital agement and securities research. He Harbor, Inc. Previously he served as a joined Davis Advisors in 1989 after director of Click Commerce, working as a securities analyst, and Convergent Group (acquired by now leads the portfolio management Schlumberger), McKinley (acquired team for the Davis Funds. Davis by Excite) and SeeBeyond received his M.A. from the Technology Corp. Mr. Murdock also University of St. Andrews. led Insight’s investment in Illuminet William Enloe has served (acquired by Verisign). as the chairman and chief executive Murdock graduated with a degree in officer of Los Alamos National Bank Political Science from San Diego State since 1994. Mr. Enloe has been University and subsequently worked at employed by Los Alamos National the Georgetown Center for Strategic & Bank since 1971 and served as the International Studies (now known as president and chief executive officer CSIS) where he was a contributor to the from 1978–1994; vice president from export competitiveness project. 1975–1978; cashier from 1973–1975; J. Leighton Read, M.D., and as a loan officer from became a general partner at Alloy 1971–1973. Additionally, he has Ventures in 2001, after 14 years as a served as chief executive officer and biotechnology entrepreneur and chairman of the board of Title investor. He co-founded Affymax NV, Guaranty since May 2000. under the direction of Dr. Alejandro Enloe serves as a member of Zaffaroni. He founded Aviron, where the boards of directors of the State Private Online. Additionally, Mr. Melton was the initial he served as chairman and CEO until 1999 Equity Investment Committee, the funder of Transaction Network Systems and director until its sale to MedImmune in Association of Commerce and Industry, Los (TNS), founder and CEO of CyberCash, and early 2002. He received a B.S. from Rice Alamos Economic Development Committee, an initial funder of other high-tech companies University in Psychology and Biology (1973), Los Alamos Technical Associates, Inc., including Prio (sold to Infospace, Inc.) and an M.D. from the University of Texas Health MIOX Corporation, and as a member of the Maxager Technology. Melton is currently an Science Center at San Antonio (1976), and Los Alamos National Laboratory Foundation, active investor and board member of several completed internal medicine training at the the Industrial Business Development early stage venture capital companies. He Peter Bent Brigham Hospital in Boston, Advisory Board, the American Banker’s holds a master’s degree in Asian Studies and where he held appointments at the Harvard Association Government Relations Council, Chinese Philosophy. Medical School and School of Public Health. and Quality New Mexico Judges Panel. Jerry Murdock is a managing He has served as director for a number of William Melton, president of director and the co-founder of Insight other biotechnology companies, on the exec- Melton Investments, has been an entrepre- Venture Partners. Since Insight’s inception in utive committee of the Biotechnology neur and investor in the high-tech industry for 1995, he has played a leading role in defin- Industry Association, and has won several over 30 years. In 1981, he founded VeriFone, ing the company’s investment strategy and awards as co-inventor of technology underly- Inc., the transaction automation company has been primarily responsible for the devel- ing the Affymetrix GeneChip™. that made credit authorization terminals ubiq- opment of many of the firm’s portfolio Peter Schwartz is a co-founder and uitous on retail merchant counters. From investments. Prior to Insight, Murdock chairman of Global Business Network (GBN), 1991 to 1999, Mr. Melton was an early founded the Aspen Technology Group to a Monitor Group Company. GBN is a mem- investor in and board member of America provide strategic consultancy services to bership organization specializing in scenario

40 Santa Fe Institute Bulletin SPRING 2005 TRANSITIONS TRANSITIONS TRANSITIONS

thinking, strategic conversation, and futures ties and research institutions throughout the on dynamic models of social and economic research. Schwartz is also a venture partner world. systems, environmental economics and reg- of Alta Partners in San Francisco, and serves Seven individuals—Jim ulation, global change science and policy, on the advisory boards of numerous organiza- Crutchfield, Walter Fontana, and industrial organization and economic tions and companies ranging from The Alfred Hubler, George geography. Current projects include industri- Highlands Group to USC’s Institute for Gumerman, Timothy Kohler, al ecology and agent-based modeling, social Creative Technologies. Before founding GBN Peter Schuster, and Martin influences and smoking behavior, firm in 1987, Schwartz headed scenario planning Shubik —well known to the SFI commu- dynamics, market volatility, and discount for the Royal Dutch/Shell Group of Companies nity (either as Resident Faculty or former rate policy. in London and directed the Strategic External Faculty members) recently were Timothy Buchman is professor of Environment Center at SRI International. added to this roster. surgery at Washington University School of

Schwartz is the author of Inevitable Surprises Other recent additions: Medicine, with specific expertise in trauma and The Art of the Long View. He is also the Robert Axtell is a senior fellow at and intensive care, practices that present co-author of The Long Boom, When Good Brookings Institution whose work focuses some of the most dramatic situations of com- Companies Do Bad Things and China’s plex systems failure and recovery. His Futures. He received a B.S. in research agenda is focused on the role Aeronautical Engineering and of noise or “biological variability” in inter- Astronautics from Rensselaer acting organ systems. Polytechnic Institute. Arizona State University William J. Spencer was Biomathematics Professor Carlos named chairman emeritus of the Castillo-Chavez’ research focus- International SEMATECH board in es on the role of dynamic social land- November 2000 after serving as chair- scapes in disease evolution. He has man of SEMATECH and International also worked on the role of dispersal SEMATECH boards since July 1996. and disease as enhancing mechanisms He came to SEMATECH in October of ecological diversity. He is currently 1990 as president and chief executive working problems at the interface of officer. He continued to serve as pres- homeland security and disease inva- ident until January 1997 and CEO sions (natural or deliberate) and on until November 1997. During this models for the spread of social “dis- time, SEMATECH became completely eases” such as alcoholism and privately funded and expanded to Ecstasy use. Carlos Castillo-Chavez include non-U.S. members. directs the Mathematical and Spencer has held key research Theoretical Biology Institute, which positions at Xerox Corporation, Bell offers research opportunities at the Laboratories and Sandia National interface of the biological, computa- Laboratories. Before joining SEMATECH tional and mathematical sciences. in October 1990, he was group vice The current research of Eric president and senior technical officer at Heller, professor of physics at Xerox Corporation in Stamford, Harvard University, is in the area of Connecticut from 1986 to 1990. chaotic phenomena in quantum Spencer received an A.B. degree mechanics, on mesoscopic physics, from William Jewell College in and on areas relevant to quantum com- Liberty, Missouri, an M.S. degree in puting. His publications include works Mathematics, and a Ph.D. in Physics on scattering, cold atom collisions, and from Kansas State University. quantum dots (a proposed technology for quantum computing). Heller leads a EXTERNAL FACULTY large research group and collaborates An important driving force of SFI’s closely with chemists and experimen- scientific life is its network of external talists. researchers, affiliated with universi- Kunihiko Kaneko, professor

Santa Fe Institute Bulletin SPRING 2005 41 TRANSITIONS TRANSITIONS TRANSITIONS of pure and applied sciences at the (acting) team leader for the Self-Organizing phyla at great and shallow time depths University of Tokyo, is one of the leading Systems team at Los Alamos National include a reconstruction of Proto-Kiranti and complex systems researchers in Japan. As Laboratory. He also leads the Los Alamos Tibeto-Burman for the Himalayan Languages an Ulam Fellow at the Center for Nonlinear Astrobiology program working on experi- Project of Leiden University. Starostin is the Studies at Los Alamos National Laboratory mental and computational protocells. In co-author of the recent Etymological in the early 1990s, Kaneko pioneered a class addition he is one of the principal investiga- Dictionary of Altaic Languages and is a of models known as “coupled map lattices” tors on the new European Union sponsored founding member (along with Murray Gell- for pursuing the structure and mechanisms Programmable Artificial Cell Evolution Mann and Merritt Ruhlen) of SFI’s Evolution of complex phenomena. He has developed (PACE) program and he was one of the of Human Languages Project, a program globally coupled maps (GCM) and in recent founders of the Artificial Life movement. devoted to tracing the genealogical tree or years has considered the origin of life with a David Sherrington is Wykeham phylogeny of human language. computational model developed from GCM. Professor of Physics and Head of Bette Korber, staff scientist at Los Theoretical Physics at University of Oxford. Alamos National Laboratory, is an interna- Sherrington is a condensed matter theoreti- tionally recognized AIDS researcher. She cal physicist best known for his seminal applies computational methods to study contributions to the theory of spin glasses. human retrovirus evolution and diversity, Sherrington is one of the world’s experts in focusing on international variation and vac- the application of statistical mechanics to cine design. Her work brings together complex problems in physics and biology; immunology, virology, genetics, mathemat- he and his co-authors have worked on modi- ics, and computing technology. The Los fications of the minority game including Alamos HIV sequence database that she has noise, and have demonstrated how statisti- overseen for close to a decade provides a cal mechanics may be applied in this con- basic resource for HIV researchers, facilitat- text. He is also currently interested in emer- ing the investigations of thousands of scien- gent behavior in neural networks and in cog- tists in the field. nitive processes. Michel Morvan is a professor at James Sidanius received his Ecole Normale Supérieure in Lyon, France. Ph.D. in political-psychology at the Morvan’s interdisciplinary activities began University of Stockholm, Sweden in 1977 with a study of the lattice structure of the where he also taught for 10 years. He joined states of sandpile models, creating a formal the Psychology Department at UCLA in mathematical structure for self-organized 1988 and is currently a fellow at UCLA’s criticality. Recently his work has focused on Center for the Study of Society and Politics. how one can do complex systems research Sidanius is author of some eighty scientific in a mathematically principled way, and papers in the general field of political-psy- especially understanding when simulations chology. This work includes study of the are robust under different conditions. interface between political ideology and cog- Morgan is playing a pivotal role in building nitive functioning; the political psychology of complex systems activities in Europe includ- gender, group conflict, and institutional dis- ing recent EU-sponsored Thematic Institutes crimination; and evolutionary psychology. and invitational workshops involving young Sidanius is co-author of several books complex systems researchers. including Social Dominance: An Intergroup Steen Rasmussen has pioneered Theory of Social Hierarchy and Oppression several methods and applications for self- and Racialized Politics: Values, Ideology, and Steven Strogatz is a professor in organizing processes in natural and artificial Prejudice in American Public Opinion. the Department of Theoretical and Applied systems including abstract self-programma- Sergei Starostin is a Russian Mechanics and the Center for Applied ble matter, molecular dynamics (MD) lattice scholar at the Moscow University for the Mathematics at Cornell University. He is the gas simulations for molecular self-assembly, Humanities, who is also affiliated with author of the textbook Nonlinear Dynamics rational and evolutionary protocell design, as Leiden University in the Netherlands. A and Chaos: With Applications to Physics, well as novel simulations for large-scale member of the Russian Academy of Biology, Chemistry, and Engineering and the socio-technical systems. He is currently the Sciences, his reconstructions of language trade book Sync: The Emerging Science of

42 Santa Fe Institute Bulletin SPRING 2005 TRANSITIONS TRANSITIONS TRANSITIONS

Spontaneous Order. Strogatz’ seminal 120 papers and nine books including SFI Postdoctoral Fellows Bieda research on human sleep and circadian The work of Colleen Webb, assis- and Choi Move to New Institutions rhythms, scroll waves, coupled oscillators, tant professor of biology at Colorado State SFI Postdoctoral Fellow Mark Bieda synchronous fireflies, Josephson junctions, University, focuses on theoretical evolution- has taken a position in the Farnham and small-world networks has been featured ary ecology, coevolution, species interac- Laboratory at the UC Davis Genome Center. in Nature, Science, and Scientific American, tions, resiliency, quantitative genetics, This new organization conducts research on among other publications. genomics, nonlinear dynamics, and spatial the phenotypic consequences of genetic Joseph Traub is the Edwin Howard models. Currently, she is studying how variation through an approach that combines Armstrong Professor of Computer Science at ecosystem level characteristics, such as comparative, functional, and structural Columbia University. He was founding chair- modularity, enhance the robustness of genomics. While at SFI, Mark’s work man of the Computer Science Department at ecosystems to disturbance (resiliency), and focused on the evolution and development Columbia from 1979 to 1989 and founding whether these higher level characteristics of single neuronal properties, genomic can evolve from selection at the individual analysis of neural genes, and analysis of level. From 2001 to 2003 Webb held a joint poxvirus interactions with the immune sys- Santa Fe Institute/Yale University postdoc- tem. toral fellowship, working on the robustness Jung-Kyoo Choi is now at the of ecosystems. School of Economics and Trade at Kenneth Weiss is Evan Pugh Kyungpook National University in South Professor of Biological Anthropology and Korea. Choi’s work at the Institute focused Genetics at Pennsylvania State University. on the evolutionary dynamics of human Weiss has been a leader in the use of behavior and institutions. He studied the for- genetic variation to ask and answer anthro- mation, evolution, and spread of cooperative pological and evolutionary questions about norms, and various institutions within com- humans. His work spans paleodemography munities that support these norms. as well as the genetic basis of diseases. At Penn State, Weiss has developed his lab Borenstein is 2005 Steinmetz with specializations in mammalian models of Fellow complex traits and in the genetics of human Elhanan Borenstein, a graduate disease. He is the author of numerous publi- student at Tel-Aviv University, has been cations including a widely-read column in selected as the 2005 Steinmetz Fellow. The the journal Evolutionary Anthropology. Steinmetz Fellowship, awarded to Complex Douglas White is a professor of Systems Summer School (CSSS) alumni, anthropology and graduate director of social supports a one-month research residency at networks at the University of California at the Institute. Its aim is to provide the oppor- Irvine. Trained at the University of tunity to CSSS students to further pursue Minnesota, White has done field research of research projects in complex systems and Ojibwa Indians, and societies in Veracruz to participate in SFI scientific activities. and other parts of Mexico. White’s research Borenstein’s research interests focus on interests focus on social structure/networks, artificial life and evolutionary computation, modeling social systems and dynamics, and the interaction between imitation and evolu- mathematical anthropology. His co-edited tion, cultural evolution and social behavior, books include Kinship, Networks and and models of gene-meme co-evolution. Exchange (1998) and Research Methods in During his stint at SFI, Elhanan will work on chair of the Computer Science and Social Network Analysis (1991). White’s cur- developing a conceptual Telecommunications Board of the National rent modeling work is on large-scale longitu- computational/mathematical model for geno- Academies from 1986 to 1992. From 1971 dinal studies of human populations and the type-phenotype mapping. to 1979 he was head of the Computer network dynamics of changing institutional Science Department at Carnegie Mellon configurations. University. Traub is the founding editor of the Journal of Complexity and an associate edi- tor of Complexity. He is the author of some

Santa Fe Institute Bulletin SPRING 2005 43 PATRICK MCFARLIN : PHOTO

Understanding Communication— From Interstellar to Intercellular and Beyond By Rebecca E. McIntosh

While considering how best to approach the subject throughout the network. The meeting aimed to empha- of communication for a recent Santa Fe Institute size the need for a new, non-equilibrium network infor- Annual Business Network Meeting and Symposium, mation theory applicable in many research areas. SFI Research Professor David Krakauer contemplated Krakauer explained that because the problem of orders of magnitude. What would it be like to integrating across multiple information sources is com- approach the subject of communication networks from mon in many types of communication networks, be the most vast expanses, such as space, to the most they natural or engineered, the conference should minute, such as quantum mechanics? Thus, the struc- explore networks by common structure and function, ture of the Symposium was set, taking members on a rather than by discipline. To further emphasize this, journey of discovery through many fields of research, the presentations were arranged according to network and revealing that although the networks may differ scale, illustrating the important property of self-similar- greatly, some principles remain generally applicable: ity, such that no matter how large or small the net- whether examining communication between planets, works, many of the fundamental properties persist. between primates, or within cells in a human body, This way, he explained, representatives from different understanding the flow of information is key. areas of expertise could converge on the subject, mak- Krakauer set the stage by defining many models of ing contributions and applying new insights to their communication networks, both traditional and innova- respective fields. tive, including the canonical information theoretic prin- ciple of a sender (or node) giving specific information Somewhere Out There to a receiver (another node) through a channel or link. The first presenter, Seth Shostak, from the SETI What complicates the situation, though, is that in most (Search for Extraterrestrial Intelligence) Institute began cases, multiple nodes are in fact sending and receiving by addressing the topic of interstellar communication. information simultaneously, which necessitates integra- He explained that SETI has worked to develop ways to tion over multiple channels and filtering out different detect sentient life on other planets by looking for sources of noise. This problem becomes even more signs of radio or optical communication. In order to do complicated when communicating through multiple so, SETI has used a large radio telescope in Puerto networks, as is common in biology and social systems. Rico, an optical telescope at the Lick Observatory, and Furthermore, in some networks, most nodes are equally is currently developing an array of 350 radio telescopes effective, while certain “key” nodes, if they were dis- in California. rupted, would drastically change communication Discussing the likelihood of life existing on other

44 Santa Fe Institute Bulletin SPRING 2005 planets, Shostak explained that most scientists in his Power Structures in Communication— field agree that statistically life probably does exist out “Knocking Out” a Primate there somewhere. In fact, he noted, there are more Another type of communication discussed was sta- stars in our galaxy alone than grains of sand in North tus-signaling interactions among nonhuman primates. America, that many are not unlike our sun, and that Jessica Flack, a postdoctoral fellow at SFI, explained more than 130 planets have already been detected. the role of communication in the “organizational However, even though planets with biology—even mechanics” and robustness of animal social organiza- very old ones—could be common, the matter of tions. Flack is interested in how power structures, whether or not there is intelligent life on them is, of which emerge from status-signaling interactions, influ- course, a separate question. “It’s not entirely clear that ence intra- and inter-organizational heterogeneity in if I give you a bit of time, you get intelligence,” joked conflict management performance and style, and how Shostak. conflict management mechanisms, in turn, influence Shostak explained that our ideas about what extra- social network structure. terrestrial intelligent beings might be like are probably rather naive. Given the fact that the Sun is a relative Using ethological techniques that include video and newcomer to the galaxy, other intelligent life forms, if voice-recording of primate social interactions in large, they exist, could very well have mastered artificial captive groups of macaques (monkeys found mainly in intelligence and produced thinking machines. Any Asia) and chimpanzees (apes found in west, central, aliens that we might get a signal from, could, in fact, and east Africa), Flack collects data on status signaling be machinery, not protoplasm. interactions, conflicts, interventions by third parties, Shostak emphasized that sending information reconciliation, and pro-social behaviors including between stars is relatively easy, and therefore by eaves- grooming and play. It turns out that the macaque dropping on the communication networks throughout social group studied by Flack and her colleagues has a the galaxy, scientists might detect advanced life if it’s log-normal power distribution, in which a few individ- really out there. “We’re not trying to figure out what uals receive many status signals from many individu- E.T. is saying,” he said. “We’re just trying to see if the als. These individuals are responsible for a large major- transmitter is on.” ity of effective conflict management. This conflict Also discussing large-scale networks, Chris Wallace, management, called policing, occurs when third-parties of Northrop Grumman Mission Systems, presented the impartially intervene into ongoing disputes among audience with a glimpse of the satellite communication group members, thereby terminating the conflict. systems around the planet and how they are used. This Although policing is relatively rare, occurring in only network of communication has layers in space, within about 15 percent of disputes, Flack hypothesized that Earth’s atmosphere, on the ground, and in the water. it plays an important role in organizational robustness. Each layer has many capabilities. For instance, with To test this hypothesis, Flack and colleagues performed data from some of the space satellites (in the upper lay- a “knockout” study. Knockout studies are common in ers of the system), it is possible to view the Earth clear- the field of developmental genetics, where researchers ly enough to identify different species of trees. infer the function of genes by disabling them and Wallace showed how networks of satellites commu- assessing changes to organismal phenotype following nicate with networks of aircraft, which then pass infor- knockout. mation to networks on the ground—which could Using this same logic, Flack temporarily and repeat- include individuals in the military. Wallace described edly removed powerful individuals responsible for the use of coordinated systems in military operations effective policing and asked how social network struc- and even showed the Symposium participants exam- ture and social system variables, like levels of aggres- ples of how this information could be of use to soldiers sion and reconciliation, changed when the policing on the ground in Iraq. He explained that soldiers in one mechanism is disabled. In the absence of policing, area are able to use satellite information to locate sol- aggression increased, pro-social behavior decreased, diers in other areas (not visible to them) and to monitor and social networks fragmented – indicating that the and communicate about each group’s position. presence of powerful policers plays an important role Naturally, one of the key components in this network is in organizational plasticity and robustness by promot- that not all nodes are privy to the information, and that ing positive interactions among relatively unfamiliar it must be kept secret, especially from the enemy. individuals and maintaining network structures that

Santa Fe Institute Bulletin SPRING 2005 45 promote information flow through the system. The ways can be illustrated through examining certain dis- upshot of all of this is that organizational robustness, eases. For instance, the protein interferon is produced as measured through changes to social network struc- by HIV-infected cells to warn other cells (by binding to ture, is made possible by effective conflict manage- a receptor upstream of a signaling pathway) to enter ment, which in turn is made possible by a power into a virus-activated state, thus making them less sus- structure that emerges from a status communication ceptible to infection. However, there can be substantial network. variation in this response. For example, some people have large amounts of interferon and are thus less like- ly to get HIV, and males and females vary in their How Cells Communicate interferon pathways. Viruses can attack these signaling Further down the spatial scale of the networks, pathways, disrupting the flow of information and crip- David Krakauer discussed his work with External pling the host response. Faculty member Walter Fontana, of Harvard’s Systems Another example of the importance of cell commu- Biology Institute, on networks of communication nication has been found in cancer research. It is under- within cells. These networks, which typically consist stood that certain signal pathways are disrupted in can- of modified proteins, are not only essential to the cells’ cerous cells with mutations to proto-oncogenes. These ability to communicate with each other, but they also pathways are normally used to induce a cell to prolifer- make up the primary mechanism by which a message ate, but in cancerous cells they are prevented from turn- that reaches a cell can then be transmitted to the nucle- ing off, causing the cell to proliferate uncontrollably. us within the cell. These signals frequently induce Conversely, it is normal for cells to be regularly sig- DNA to express copies of genes that will be translated naled to commit suicide and die, termed “apoptosis” into proteins, which modify communication networks (especially when defective), but in many cancer cells, or perform more general functions for the cells. The the pathways that instruct the cell to enter apoptosis two types of regulatory networks Krakauer discussed are disrupted and the cell immortalized. The disruption were signal transduction networks (protein-protein of these networks can be likened to memory loss in the interactions) and gene regulatory networks (protein- nervous system, whereby adaptive states are erased gene interactions). through the action of mutations, and hence disease In describing how signal transduction pathways becomes a form of molecular network amnesia. function, Krakauer explained that in response to a stimulus, proteins floating in a cell are able to sponta- neously come together to create a network or pathway. Quantum Communication This is in contrast to engineered networks that typical- Carlton Caves, of the University of New Mexico, ly exist as a fixed topology. The self-assembling prop- captivated the audience with a discussion of quantum erty of biological regulatory networks present new the- information theory and quantum key distribution. As oretical challenges in biological information theory. mankind continually advances its methods of commu- One example is the cascade network, in which a set of nicating, through satellite and computer networks, one proteins work as a relay of connections in order to of the biggest challenges continues to be security of pass a message from the exterior of a cell to the nucle- the communicated information. Secret information, us. The cascade circuit has been observed to have such as issues of national security or even bank many computational properties, including amplifica- account details, is constantly being transmitted. To tion, filtering, and integration. In contrast, ionic signal- protect this information from the prying eyes of hack- ing pathways occur when a protein (or ligand) attaches ers, communication is encrypted; the sender and to the outside of the cell causing the release of charged receiver share a security code or key used for encryp- particles that then affect many different proteins within tion, without which decryption is practically impossi- the cell. The ionic interactions are more diffuse, which ble. However, this merely changes the emphasis of the Krakauer likened to the systemic modulation of nerve security. Caves addressed the question: how are the cells by circulating hormones, as opposed to the pre- security keys, packets of information themselves, kept cise connections between nerve cells in the brain. The secure when transmitted over open networks? ionic signals are like a cloud, and the cascade is a net- Caves used the example of secret communication work. between two parties, Alice and Bob, to illustrate con- The importance of these cell communication path- cepts of quantum key distribution. Essentially, Alice

46 Santa Fe Institute Bulletin SPRING 2005 wants to send information to Bob so that they can someone is trying to eavesdrop. Furthermore, once share a secret code. The scheme proposed uses lasers they have a secret key, Alice and Bob can send data to create individual photons, which Alice can send back and forth using the key to encrypt it. through two channels. The photons represent binary Caves explained that although this technique is still information (0 or 1)—characterized by their polariza- very much a theory, there are some products already tion. This is done in two ways, vertical-horizontal available to help individuals create secret keys. polarization or diagonal left-to-right and right-to-left Traditionally, research efforts have been based upon polarization. When Bob receives the photons in the quantum information being transmitted with lasers, first channel, he measures the polarization of each, computers, and fiber optic cables; however, research using a randomly selected coordinate set, sometimes (including some at the Los Alamos National vertical-horizontal, and sometimes diagonal. Alice will Laboratory) is now being conducted to try sending then send, through the second channel, the correct ori- these keys through the atmosphere. entation of the coordinate set used to generate the Other conference speakers included Peter Monge, polarization of the photons, but not their digital values of the University of Southern California, who explored (0 or 1). Comparing results, Bob can discard measure- the theory of communication networks. Some celebri- ments where the incorrect coordinate set was used, ties also visited the proceedings. James Surowiecki, therefore leaving the correct digital values. This digital writer for The New Yorker and author of The Wisdom of sequence becomes their secret key, and can be used as Crowds – Why the Many Are Smarter Than the Few and a “one-time pad” to send secure information. How Collective Wisdom Shapes Business, Economies, Societies The crux of this encryption scheme is the use of and Nations, spoke on collective reasoning. Also on individual quantum objects, photons, to transmit the hand was New Mexico Governor Bill Richardson, who information. Quantum mechanics dictates that an presented a few points about communication in poli- observation cannot be made without perturbing the tics and declared November 5th as “Santa Fe Institute state observed. In the case of photons, they cannot be Day” in New Mexico. seen without being absorbed (or destroyed), thus only Alice and Bob share the key and can determine if Rebecca E. McIntosh is a science writer living in Santa Fe. COPYRIGHT CORBIS : PHOTO

Santa Fe Institute Bulletin SPRING 2005 47 Seeking Security in a Hostile Environment By Rebecca E. McIntosh COPYRIGHT CORBIS : PHOTO A major theme of the early part of the 21st century seems to be security—how to create it and maintain it. Though the primary focus has been on national security, a quieter foe relentlessly threatens our daily life. In November of 2004, members of the Santa Fe Institute Business Network held a topical meeting to discuss this very issue: security of electronic information. The Adaptive and Resilient Computing Security (ARCS) meeting illuminated the scope of this growing problem, but also showed that although solutions are daunting, many can be found in examining the natural world around us.

48 Santa Fe Institute Bulletin SPRING 2005 The ARCS meeting, which was held Vespignani said that the “patching” dreds of countries converge into the on November 3 – 4, 2004, at the Santa methods that are often used are not close quarters of crowded airports, the Fe Institute, was the third of its kind, the most effective. He emphasized that flow of disease is accelerated, as the and included experts and Business even if vaccinations are developed individuals are in close contact and Network members from around the early on in an outbreak, the virus can therefore more likely to share germs. world in the field of computer security. still linger for years. In the world of Similarly, when the distance between It was chaired and organized by Robert computers, this is a significant prob- computers is shortened by connecting Ghanea-Hercock of British Telecom and lem. New viruses are constantly many of them to a few centralized Matthew Williamson of Sana Security appearing, and patching software is machines, infection problems may be Inc. By focusing on adaptive and immediately developed to protect com- exacerbated. The epidemic is accelerat- resilient technologies, the conference puters from these threats. However, ed when “there is a strong correlation emphasized approaches inspired by customers rarely realize that although between the degree of the nodes and studying biological systems, including they may be updating their computer the flow of information,” Vespignani contributions on topics such as self- systems with the latest patches, they said. The most interesting direction of healing networks, machine learning, are still susceptible to an attack from a research, said Vespignani, is to “find and immunological models. virus that circulated years ago and the interplay between the flow of infor- Alessandro Vespignani, of Indiana everyone has since forgotten. mation on networks…and the physical University’s School of Information, One reason the patching strategy is pathway of the exchange.” opened with a keynote address about ineffective is because it is random. the epidemiology of computer viruses Certain individuals are diligent in updat- Taking Aim and attacks, making clear the parallels ing their virus software and installing Other speakers fur- between a study of biological viruses patches when needed. However, there ther emphasized the and computer ones. Epidemiology, a are plenty of people who don’t bother, importance of targeted term usually reserved for public health and this allows the viruses to spread. immunization. Hong Li of circles, is the study of how disease epi- Instead, he proposed that targeted Intel Corporation spoke of the demics propagate around the world. immunization of a few key computers Internet as a network of fully connect- Vespignani drew analogies from the would be much more effective in keep- ed networks. Her network of networks basic models epidemiologists use to ing the viruses from infecting too many model showed not only that there are map a virus throughout its course of machines. He emphasized that knowl- roughly six degrees of separation infection. For example, a virus first edge of the network can be helpful in between networks, which coincides spreads freely, infecting each individual identifying these “key” players. Every with “small world” characteristics, but it contacts, but it can eventually reach a computer has a different susceptibility it also showed the value of recognizing threshold and slow down either because of its location. Because some isolated networks such as a network of because a vaccine has been developed computers may serve as gateways computers in a company that are all or because too many individuals have between a small network (all the com- protected by a firewall. died, thus stopping further spreading. puters in a company) and the rest of the “The maximum number of hops The virus eventually reaches an endem- Internet, they are key when it comes to between two nodes is five,” said Li, ic or steady state. protection. again demonstrating a “small world Vespignani pointed out that this “You take advantage of this hetero- effect.” The research she presented research has not generally been able to geneity, and just by protecting a few showed a virus spreading among the predict epidemics, rather it helps to individuals, one can protect the entire network of networks through direct understand how they spread, and that system,” said Vespignani. Vespignani paths (not e-mails) when the comput- understanding can be applied to com- discussed the need to understand the ers are interacting and sending informa- puter security. Although they are simi- flow of information in computer net- tion packets to other machines on the lar, computer viruses don’t behave works in order to later identify where network. For instance, sending a exactly like biological ones. For vaccination is most effective. For request to Google requires one’s com- instance, they may not always reach an instance, biological epidemiologists puter to generate packets of informa- epidemic threshold. However, these study the populations of the world and tion to transmit to Google’s computer models do offer a way to visualize the their susceptibility, as well as the and vice versa. Because every action spread of disease and the effective- modes through which they interact. the computer performs involves this ness of vaccination. When thousands of people from hun- type of exchange of packets, malicious

Santa Fe Institute Bulletin SPRING 2005 49 packets of virus information can Queuing Up sneak into the system. Li examined SPAM Taking a slightly different the computers, or nodes, in her net- approach, Miranda Mowbray of works and their susceptibility to Hewlett-Packard Bristol (a research Slamming the Spam viruses. The nodes were described as facility of Hewlett-Packard in Bristol, Although many attacks on comput- being infected, infectable (susceptible) England) explained that her research er networks are direct, and many users or immune (patched). Li demonstrated, was prompted by many complaints that are effective in protecting their sys- as Vespignani had discussed, that if filtering junk mail, before delivering it to tems with the latest technology—it certain gateway nodes are protected, employees, was simply causing too seems almost everyone is still suscep- the virus didn’t propagate the same long a delay. Mowbray and Matthew tible to junk e-mail or spam, which can way. Williamson of Sana Security Inc. (who certainly transmit viruses, or can just Kihong Park, of Purdue University, previously worked at HP Bristol) be annoying. Christian Renaud of Cisco talked about proactive and reactive worked on mitigating this problem by Systems, a worldwide leader in net- defenses against distributed denial-of- creating separate queues for mail to working for the Internet through the service (DDoS) and worm attacks in the enter the content scanner: one for junk development of software, hardware, global Internet, which is known to mail and one for regular mail. This way, and service systems, talked about how exhibit power-law connectivity. His the mail in the regular queue would spam not only can present a physical research showed the effects of filtering take a shorter amount of time to reach problem, but can make people lose data packets based on knowledge the recipient, and junk mail would have faith in doing business on the Internet. about addresses for DDoS and content the long queue. In order to determine Renaud explained Cisco’s policy of signatures for worm protection. Park what mail belonged in which queue, both content inspection and explained that the packets flow through they studied the IP addresses and path/sender-based filtering, which is the Internet from computer to comput- found some interesting trends. similar to Park’s approach of filtering er, based on addresses stating what By monitoring e-mail for a period of packets. By placing digital signatures or computer they are from and where two months, they found that junk e- keys on Cisco employees’ e-mails that they are going. Typically, these address- mail does not all come from the same identify the sender, Cisco’s servers can es, or IP addresses showing the com- server, but a server that sends junk e- identify whether or not the sender is puters’ locations, are looked at by mail will likely send more junk. “authorized” to send e-mail from Cisco routers to identify where the packet is Essentially, they had expected to find a computers, or authorized to send e- going—and the “from” address is usu- few servers that would spew lots of mail to Cisco computers. In order to ally ignored. Park stressed the value of junk e-mail, but in reality, the junk was maximize the effectiveness of such a strategically placing filters at transit coming from all over the place. system, Renaud explained that other sites, as opposed to leaf sites, to However, they did find that servers that companies would need to adopt a sig- specifically look at “from” addresses to sent junk, were likely to send more nature/key system. determine whether or not the sender junk, and servers that sent good e-mail, Renaud also discussed the impor- of the packet is spooling for DDoS were likely to send more good e-mail. tance of preserving certain characteris- attack prevention and content-based This enabled them to track down junk tics about e-mail such as the ability to signatures for worm containment. In servers by monitoring IP addresses send it to anyone and to have it be rela- the case of worms, Park and his col- that identified the server that sent the tively anonymous. He explained the leagues show that very small (1.5%) fil- mail. The system then scans the IP necessity of developing this system in ter deployment suffices to effect con- addresses in order to determine conjunction with systems that recog- tainment, facilitated by the power-law whether the mail came from a good nize behavior. For instance, if an individ- connectivity of the Internet. Feasibility server and should enter the short ual’s e-mail system is taken over by a of this approach is demonstrated by queue, or whether it came from a junk virus, and made to send millions of prototype systems built using Intel IXP server and should enter the long malicious e-mails (becoming a so-called network processors that perform worm queue. Since the Hewlett-Packard serv- zombie), the system would need to rec- filtering at gigabit (billion bits per sec- er only needs to remember five IP ognize the abnormal sudden increase in ond) wire speeds. addresses from each server in order to sending and stop it—even if the e-mails accurately assess the server’s tenden- had the appropriate signature keys. cies, the whole system could then adapt: if a server sent one junk e-mail,

50 Santa Fe Institute Bulletin SPRING 2005 but then ten good e-mails, it would no nize that any deviation from this behav- head of the Israeli Labor Party, Shimon longer be categorized as a junk server. ior is worrisome. Peres: “If a problem has no solution, it “Once you’ve sent me three good Of course, Hofmeyr explained the may not be a problem, but a fact—not e-mails in a row, then you’re a good importance of building this normal to be solved, but to be coped with over guy again and you get fast access to behavior in a production environment. time.” Therefore, Patterson explained, the content scanner,” said Mowbray. He also pointed out that if the comput- attacks are always going to happen, This way you don’t have to keep er is infected with a virus while one is and finding ways to recover quickly and records of all the IP addresses. Also, observing its “normal” behavior, this without too much disruption is the best it’s somewhat foolproof. If good e-mail approach would be problematic. And way to cope. happens to fall into the junk queue, it is he stressed the goal of trying to make Patterson said that most delays in still scanned and delivered—just the system automated in its response recovery are caused by human error delayed longer than the mail in the reg- to attack. and that most errors are immediately ular queue. Justin Balthrop, a graduate student self-detected. He explained that at the University of New Mexico, humans have patience thresholds Friend or Foe? described another system that uses when dealing with their computers. For One of the most innovative tech- learning of normal behavior to detect instance, humans are willing to wait niques in adaptive and resilient comput- attacks. RIOT (Responsive Input Output quite a few seconds, perhaps even a er security involves machine learning. Throttling) is a system he helped create minute, for a website to load. Thus, Researchers studying the immunology (also with Matthew Williamson) that Patterson proposed taking advantage of of biological systems have noticed the includes detectors that are designed to small windows of time such as those value of self and non-self recognition. learn the user’s behavior and to during which the computer could con- Immune systems often depend on this observe all network connections made duct repairs without really disrupting principle: if something does not appear by the computer. The detectors have the user (a minute or less). These quick to be “self” or normal, it must be an an immature state during which they repairs, which sometimes include attacker. are learning normal behavior. Once they “micro-reboots” when sections of the Steve Hofmeyr, of Sana Security have matured, they are capable of rec- computer need to restart, can help Inc., discussed anomaly detection and ognizing when the computer is con- solve problems within the system building an adaptive system, specifical- necting to a network in a way that is before they get too big. Rather than ly through looking at the order of not “normal,” and therefore blocking waiting and having long system out- instructions, or system calls, sent from such connections. ages, repairs could be broken down the computer’s operating system to However, the user has the ability to into smaller chunks of time, thus caus- various programs or disk drives. These override the detector’s decision to ing less anguish for the user. system calls are the computer’s pri- block a certain activity. By viewing a Recovery-oriented computing, as mary way of communicating with its scrollbar that shows each connection Patterson put it, requires building tools different parts in order to perform or program that is being blocked or to help operators design rapid-recover- tasks. Understanding these instructions slowed, the user can decide to allow a ing-based systems on benchmarks cre- can help illuminate what behavior is connection to take place, thus sending ated from real failure data. Along with considered normal for each individual the detector back to an immature state the “micro-reboots” discussed above, computer and user. so that it does not block the same he also encouraged processes to cre- “You’ve got to be able to collect action again. This makes the entire sys- ate “undo” mechanisms within com- and define the normal very easily, and tem adaptive. puter systems. Patterson ended with a you’ve got to be able to track changes very realistic analogy: he reminded or adapt to changes effectively,” said Learning to Cope audience members that locks are not Hofmeyr. There is also a difference in David Patterson’s keynote address considered either breakable or unbreak- monitoring behavior verses data. provided a good summary for the con- able, but rather rated by how long it Because data, such as packets and e- ference as he reflected on the idea of takes to break them. mails, can be spoofed or misinterpret- computer security as a whole. ed, it is not a reliable source to use. By Predominantly, Patterson, from U.C. Rebecca E. McIntosh is a science writer liv- focusing on system calls, the security Berkeley, emphasized the overall value ing in Santa Fe. software can learn sets of tasks and of recovery of data and information pathways that are normal for the spe- after a virus attack. Patterson’s philoso- cific computer and therefore can recog- phy boiled down to a quote by the

Santa Fe Institute Bulletin SPRING 2005 51 2005 SFI Community WEDNESDAY, JUNE 15 WEDNESDAY, AUGUST 10 Mark Newman Robert P. Kirshner Associate Professor of Physics and Complex Systems, University of Clowes Professor of Science, Harvard University Michigan, Ann Arbor; External Faculty, Santa Fe Institute The Internet, Epidemics, and The Accelerating Universe: A Blunder Undone Kevin Bacon: The Emerging Supernova explosions halfway across the universe show Science of Networks that cosmic expansion is not slowing down due to grav- There are networks in almost every part of our lives. ity. Instead, to the surprise of all astronomers, it is Some of them are familiar and obvious: the Internet, the speeding up. We attribute this astonishing cosmic accel- power grid, the road network. Others are less obvious eration to a “dark energy” that pervades all of space. but just as important. The patterns of friendships or Albert Einstein anticipated this discovery in 1917, when acquaintances between people form a social network, he stuck a “cosmological constant” into his equations of for instance, and boards of directors join together in general relativity. Some called this Einstein’s greatest networks of corporations. The workings of the body’s blunder and it has been theoretical poison ivy since the cells are dictated by a metabolic network of chemical 1930s. Now we find that something just like Einstein’s reactions. In recent years, sociologists, physicists, biolo- cosmological constant is needed to explain modern gists, and others have learned how to probe these net- observations. This discovery points to something miss- works and uncover their structures, shedding light on ing in our understanding of gravity, right at the heart of the inner workings of systems ranging from bacteria to theoretical physics. This talk will show the evidence for the whole of human society. This lecture looks at some cosmic acceleration, describe what we know about the new discoveries regarding networks, how these discov- dark energy, and outline future observations that will eries were made, and what they tell us about the way reveal more about this strange new component which the world works. makes up two-thirds of the universe.

The lectures are made possible through support from For more current information about a particular talk, visit community supporters, and they are underwritten by Los our website at http://www.santafe.edu/sfi/events/publi- Alamos National Bank. For information on how you can clectures.html or call 505-984-8800. help support the Public Lecture Series, please contact Ann Stagg at 505-946-2724, or [email protected]. Please contact the Santa Fe Institute to arrange for sign language interpretation if necessary. There is no admission charge, but seating is limited. The talks are generally held at 7:30 p.m. at the James A. Little Theater on the campus of the New Mexico School of the Deaf, 1060 Cerrillos Road, Santa Fe.

52 Santa Fe Institute Bulletin SPRING 2005 Lectures TUESDAY – THURSDAY, SEPTEMBER 20 – 22 The Bulletin of the Santa Fe Institute is published by SFI to The Twelfth Annual Stanislaw Ulam keep its friends and supporters informed about its work. Memorial Lecture Series Marcus W. Feldman The Santa Fe Institute is a private, independent, multidiscipli- nary research and education center founded in 1984. Since Burnet C. and Mildred Finley Wohlford Professor of Biological Sciences, and Director, Morrison Institute for Population and its founding, SFI has devoted itself to creating a new kind of Resource Studies, Stanford University; scientific research community, pursuing emerging synthesis Science Board, Santa Fe Institute in science. Operating as a visiting institution, SFI seeks to Molecules, Machines and catalyze new collaborative, multidisciplinary research; to break down the barriers between the traditional disciplines; Mathematics: Issues in to spread its ideas and methodologies to other institutions; Biological and Social and to encourage the practical application of its results. Evolutionary Theory Published by the Santa Fe Institute 1399 Hyde Park Road The lectures begin with a description of what the various Santa Fe, New Mexico 87501, USA genome projects have shown about the patterns of Phone (505) 984-8800 genomic variation in modern humans. Advances in com- fax (505) 982-0565 putational and mathematical analyses have helped to home page: http://www.santafe.edu formulate what will be described as the standard theory Note: The SFI Bulletin may be read at the website: for the evolution of modern humans. For some popula- www.santafe.edu/sfi/publications/Bulletin/. tions, recent history and genetic patterns have an If you would prefer to read the Bulletin on your computer extremely interesting concordance. rather than receive a printed version, contact Patrisia Brunello at 505/984-8800, Ext. 2700 or [email protected].

The lectures continue with a discussion of how, for bio- EDITORIAL STAFF: logical evolution, in many cases it is incorrect to view Ginger Richardson organisms as evolving “to solve a problem.” Following a Lesley S. King suggestion by Richard Lewontin, a new framework has Kathleen Raphael Andi Sutherland been developed for biological evolution that replaces the notion of adaptation by one of niche construction, CONTRIBUTORS: according to which there is symmetry between organ- Matthew Blakeslee ism and environment, with feedbacks between these Barbara Ferry forming the driving force of evolution. For cultural evo- David Krakauer Shannon Larsen lution, a similar framework can be developed with some Rebecca E. McIntosh culturally transmitted traits forming a “cultural” environ- James W. Minett ment that permits or prevents other social or, indeed, Daniel Rockmore biological changes from occurring. Feldman will present Cosma Shalizi an important example from the epidemiology of infec- DESIGN & PRODUCTION: tious disease in the presence of antibiotic therapies. Patrick McFarlin

Santa Fe Institute Bulletin SPRING 2005 3 J STEPHEN LANSING : PHOTO

SANTA FE INSTITUTE NON-PROFIT ORGANIZATION 1399 HYDE PARK ROAD U.S. POSTAGE

SANTA FE, NEW MEXICO 87501 PAID PERMIT NO. 49 U.S.A. LUBBOCK, TEXAS