<<

The Profession of IT Peter J. Denning

Is Science? Computer science meets every criterion for being a science, but it has a self-inflicted credibility problem.

What is your profession? criteria for science and see how include computational science, Computer science. computing stacks up. systems, engineering, and design. Oh? Is that a science? I’m listening. The 1989 report, Computing as a Sure, it is the science of infor- Discipline, defined the field as: mation processes and their inter- “The discipline of computing actions with the world. is the systematic study of - I’ll accept that what you do is tech- rithmic processes that describe nology; but not science. Science deals and transform information: with fundamental laws of their theory, analysis, nature. Computers are man- design, efficiency, imple- made. Their principles come mentation, and applica- from other fields such as physics tion. The fundamental and electronics engineering. question underlying all of Hold on. There are many computing is, ‘What can natural information processes. be (efficiently) auto- Computers are tools to imple- mated?’” [3, p. 12] ment, study, and predict them. Science, engineering, In the U.S. alone, nearly 200 and mathematics com- academic departments recognize bine into a unique and this; some have been granting CS potent blend in our field. degrees for 40 years. Some of our activities are They all partake of a mass delu- primarily science—for example, sion. The pioneers of your field gen- experimental , experi- uinely believed in the 1950s that mental computer science, and their new field was science. They computational science. Some are were mistaken. There is no computer Common Understandings primarily engineering—for exam- science. Computer art, yes. Com- of Science ple, design, development, soft- puter technology, yes. But no science. ur field was called com- ware engineering, and computer The modern term, Information puter science from its engineering. Some are primarily Technology, is closer to the truth. Obeginnings in the 1950s. mathematics—for example, com- I don’ accept your statements Over the next four decades, we putational complexity, mathe- about my field and my degree. accumulated a set of principles matical software, and numerical Do you mind if we take a closer that extended beyond its original analysis. But most are combina-

JASON SCHNEIDER look? Let’s examine the accepted mathematical foundations to tions. All three sets of activities

COMMUNICATIONS OF THE ACM April 2005/Vol. 48, No. 4 27 The Profession of IT

The objection that computing is not a science because it studies man-made objects (technologies) is a red herring. Computer science studies information processes both artificial and natural.

draw on the same fundamental science deals with prediction and day produce a more accurate the- principles. In 1989, we used the verification by observation, mea- ory than Big-O-Calculus and term “computing” instead of surement, and experiment. include a theory of locality. The “computer science, mathematics, Computing research is rife nascent Human-Computer Inter- and engineering.” Today, com- with examples of the scientific action (HCI) field is examining puting science, engineering, paradigm. Cognition researchers, the ways in which human infor- mathematics, art, and all their for example, hypothesize that mation processes interact with combinations are grouped under much intelligent behavior is the automated processes. the heading “computer science.” result of information processes in By these definitions, computing The scientific brains and ner- qualifies as an exact science. It paradigm, which Science Art vous systems; studies information processes, dates back to principles practice they build which occur naturally in the physi- Francis Bacon, is fundamental recurrences skilled performance systems that cal world; computer scientists work explanation action the process of discovery invention implement with an accepted, systematized forming hypothe- analysis synthesis hypothesized body of knowledge; much com- ses and testing dissection construction information puter science is applied; and com- them through processes and puter science is used for prediction experiments; successful hypotheses Table 1. Science compare them and verification. become models that explain and vs. art. with the real The objection that computing predict phenomena in the world. thing. The com- is not a science because it studies Computing science follows this puters in these studies are tools to man-made objects (technologies) paradigm in studying information test the hypothesis; successful sys- is a red herring. Computer sci- processes. The European synonym tems can be deployed immedi- ence studies information for computer science—informat- ately. Software engineering processes both artificial and nat- ics—more clearly suggests the researchers hypothesize models ural. It helps other fields study field is about information for how programming is done theirs too. Physicists explain par- processes, not computers. and how defects arise; through ticle behavior with quantum The lexicographers offer two testing they seek to understand information processes—some of additional distinctions. One is which models work well and how which, like entanglement, are between pure and applied science; to use them to create better pro- quite strange—and verify their pure science focuses on knowl- grams with fewer defects. Experi- theories with computer - edge for its own sake and applied mental algorithmicists study the tion experiments. Bioinformati- focuses on knowledge of demon- performance of real algorithms on cians explain DNA as encoded strable utility. The other is real data sets and formulate mod- biological information and study between inexact (qualitative) and els to predict their time and stor- how transcription enzymes read exact (quantitative) science; exact age requirements; they may one and act on it; computer models

28 April 2005/Vol. 48, No. 4 COMMUNICATIONS OF THE ACM of these processes help customize making as the processes by which although he also believed com- therapies to individual patients. scientific facts are proposed, puting is a mathematical science. Pharmaceutical and materials labs argued, and accepted. A new Walter Tichy, an experimentalist create man-made molecules proposition is argued and studied and accomplished software through computer simulations of in publications, conferences, let- builder, argues that computer sci- the information processes under- ters, email correspondence, discus- ence is science [12]. David Par- lying chemical compositions. sions, debates, practice, and nas, an engineer, argues that the To help define the boundaries repeated experiments. It becomes software part of computer science of science, lexicographers also a “fact” only after it wins many is really engineering [10]. I myself contrast science with art. Art allies among scientists and others have practiced in all three tradi- refers to the useful practices of a using it. To win allies, a proposi- tions of our field and do not see field, not to drawings or sculp- tion must be independently veri- sharp boundaries. tures. Table 1 lists some terms fied by multiple observations and Even the Computer Science that are often associated with sci- there must be no counterexam- and Technology Board of the ence and with art. Programming, ples. Latour sees science-in-the- National Research Council is not design, software and hardware making as a messy, political, consistent. In 1994, a panel engineering, building and validat- human process, fraught with emo- argued that experimental com- ing models, and building user tion and occasional polemics. The puter science is an essential aspect interfaces are all “computing scientific literature bears him out. of the field [9]. In 2004, another arts.” If aesthetics is added, the Everything Latour says is consis- panel discussed the accomplish- computing arts extend to graph- tent with the time-honored defini- ments of computer science ics, layout, drawings, photogra- tion of the science paradigm. After research; aside from comments phy, animation, music, games, sufficient time and validation, a about abstraction in models, they and entertainment. All this com- model becomes part of the scien- say hardly a word about the puting art complements and tific body of knowledge. experimental tradition [8]. enriches the science. , a prominent Internal Disagreement member of the generation who Science in Action Computer scientists do not all grew up with computers, In his remarkable book about the agree whether computer science is invented the Yahoo! store and workings of science, Science in science. Their judgment on this early techniques for spam filters; Action, the philosopher Bruno question seems to depend upon he identifies with computing art. Latour brings a note of caution to in which tradition they grew up. He says: “I never liked the term the distinction between science Hal Abelson and Gerry Sussman, ‘computer science’. … Computer and art [7]. Everything discussed who identify with the mathemati- science is a grab bag of tenuously in this column (a systematized cal and engineering traditions of related areas thrown together by body of knowledge, ability to computing, said, “Computer sci- an accident of history, like make predictions, validation of ence is not a science, and its ulti- Yugoslavia. … Perhaps one day models), is part of what he calls mate significance has little to do ‘computer science’ will, like ready-made-science, science that is with computers” [1]. They Yugoslavia, get broken up into its ready to be used and applied, sci- believe that the ultimate signifi- component parts. That might be ence that is ready to support art. cance is with notations for a good thing. Especially if it Much science-in-the-making expressing computations. Edsger means independence for my appears as art until it becomes set- Dijkstra, a mathematician who native land, hacking” [5, p. 18]. tled science. built exquisite software, fre- He is not arguing against com- Latour defines science-in-the- quently argued the same point, puter science, but for an appella-

COMMUNICATIONS OF THE ACM April 2005/Vol. 48, No. 4 29 The Profession of IT

Computer scientists do not all agree whether computer science is science. Their judgment on this question seems to depend upon in which tradition they grew up. tion like computer art Area Problem have saturated. They have that is more attractive to Computation • Unbounded error accumulation on finite machines discovered most of their • Non-computability of some important problems hackers (his term for • Intractability of thousands of common problems basic principles and new elite programmers). • Optimal algorithms for some common problems discoveries are less and less • Production quality Dana Gardner, of the frequent. Why is computer Communication • Lossless file compression Yankee Group, does not • Lossy but high-fidelity audio and video compression science different? Once the • Error correction codes for high, bursty noise channels like this notion. He • Secure cryptographic key exchange in open networks current round of com- compares the current Interaction • Arbitration problem puter-science-in-the-mak- state of software devel- • Timing-dependent (race-conditioned) bug problem ing settles out, and • Deadlock problem opment to the pre- • Fast algorithms for predicting throughput and response time assuming the hackers don’t • Internet protocols industrial Renaissance, • Cryptographic authentication protocols secede, will computer sci- when wealthy benefac- Recollection • Locality ence die out? • Thrashing tors commissioned • Search groups of highly trained • Two-level mapping for access to shared objects Computer Science artisans for single great Automation • Simulations of focused cognitive tasks Thrives on • Limits on expert systems works of art [4]. He • Reverse Turing tests Relationships says, “Business people Design • Objects and information hiding Horgan argued in 1996 are working much closer • Levels that new scientific discov- • Throughput and response time prediction networks of servers to the realm of Henry eries require mastering Ford, where they are Table 2. Some non-obvious problems ever-greater amounts of looking for reuse, interchangeable solved by computing principles. complexity. In 2004 he repeated parts, automated processes, his main conclusion: “Science will highly industrialized assembly Can Computer Science never again yield revelations as lines.” Surprise? monumental as the theory of evo- OK, so computing has much art Table 2 lists six major categories lution, general relativity, quantum and its own science, although some of computing principles along mechanics, the big bang theory, of your people are not sure about with examples of important dis- DNA-based genetics. ... Some far- the science. However, does computer coveries that are not obvious to fetched goals of applied science— science have depth? Are there fun- amateurs [2]. By exploiting these such as immortality, superluminal damental principles that are non- principles, professionals are able spaceships, and superintelligent obvious to those who do not to solve problems that amateurs machines—may forever elude us” understand the science? Who would would find truly baffling. [6, p. 42]. have thought that the speed of light OK. I’m finding this compelling. Has computer science already is the same for all observers until But I still have a concern. Is it made all the big discoveries it’s Einstein postulated relativity? Or worth investing either my time or going to? Is incremental progress that particles ride probability waves R&D dollars in computer science? all that remains? Has computer until Schroedinger postulated quan- In his 1996 book, The End of Sci- science bubbled up at the end of tum mechanics? Is there anything ence, journalist John Horgan the historical era of science? like this in computer science? argues that most scientific fields I think not. Horgan argues

30 April 2005/Vol. 48, No. 4 COMMUNICATIONS OF THE ACM that the number of scientific fields tributed to the Internet boom and nologies, is more open to critical is limited and each one is slowly then crashed with the dot-com bust. thinking. Computer science has being exhausted. But computer Now you’re making all sorts of always been part of their world; science is going a different way. It claims about secure systems, spam- they do not question its validity. is constantly forming relationships blocking, collaboration, enterprise In their research, they are increas- with other fields; each one opens systems, DNA design, bionics, nan- ingly following the science para- up a new field. Paul Rosenbloom otechnology, and more. Why should digm. Tichy told me that the has put this eloquently in his I believe you? recent research literature shows a recent analysis of computer sci- marked increase in testing. ence and engineering [11]. Validating Computer The science paradigm has not Rosenbloom charts the history Science Claims been part of the mainstream per- of computer science by its rela- There you have us. We have ception of computer science. But tionships with the physical, life, allowed the hype of advertising soon it will be. c and social sciences. With each departments to infiltrate our lab- one computer science has opened oratories. In a sample of 400 References new fields by implementing, computer science papers pub- 1. Abelson, H.G. and Sussman, G.J. Structure and Interpretation of Computer Programs, 2nd interacting, and embedding with lished before 1995, Walter Tichy ed. MIT Press, 1996. those fields. Examples include found that approximately 50% of 2. Denning, P. Great principles of computing. Commun. ACM 46, 10 (Nov. 2003), 15–20. autonomic systems, bioinformat- those proposing models or 3. Denning, P. et al. Computing as a discipline. ics, biometrics, biosensors, cogni- hypotheses did not test them Commun. ACM 32, 1 (Jan. 1989), 9–23. tive prostheses, cognitive science, [12]. In other fields of science the 4. Ericson, J. The psychology of service-ori- ented architecture. Portals Magazine (Aug. cyborgs, DNA computing, fraction of papers with untested 2004); www.portalsmag.com/articles/ immersive computing, neural hypotheses was about 10%. Tichy default.asp?ArticleID=5872. 5. Graham, P. Hackers and Painters: Big Ideas computing, and quantum com- concluded that our failure to test from the Computer Age. O’Reilly and Associ- puting. Rosenbloom believes that more allowed many unsound ates, 2004. the constant birth and richness of ideas to be tried in practice and 6. Horgan, J. The end of science revisited. IEEE Computer (Jan. 2004), 37–43. new relationships guarantees a lowered the credibility of our 7. Latour, B. Science in Action. Harvard Univer- bright future for the field. field as a science. The relative sity Press, 1987. All right, I’ll accept that. You youth of our field—barely 60 8. National Research Council. Computer Sci- ence: Reflections on the Field, Reflections from have science, you have art, you can years old—does not explain the the Field. National Academy Press, 2004. surprise, and you have a future. low rate of testing. Three genera- 9. National Research Council. Academic Careers for Experimental Computer Scientists and But you also have a credibility tions seems sufficient time for Engineers. National Academy Press, 1994. problem. In the 1960s your people computer scientists to establish 10. Parnas, D. Software engineering: An uncon- claimed they would soon build that their principles are solid. summated marriage. Commun. ACM 40, 9 (Sept. 1997), 128. artificially intelligent systems that The perception of our field 11. Rosenbloom, P. A new framework for com- would rival human experts and seems to be a generational issue. puter science and engineering. IEEE Com- puter (Nov. 2004), 31–36. make new scientific discoveries. In The older members tend to iden- 12. Tichy, W. Should computer scientists experi- the 1970s they claimed that they tify with one of the three roots of ment more. IEEE Computer (May 1998), 32–40. would soon be able to systematically the field—science, engineering, or produce reliable, dependable, safe, mathematics. The science para- Peter J. Denning ([email protected]) is the director of the Cebrowski Institute for and secure software systems. In the digm is largely invisible within information innovation and superiority at the 1980s it was the disappearance of the other two groups. Naval Postgraduate School in Monterey, CA, paper, universities, libraries, and The younger generation, much and is a past president of ACM. commuting. None of these things less awed than the older one once happened. In the 1990s you con- was with new computing tech- © 2005 ACM 0002-0782/05/0400 $5.00

COMMUNICATIONS OF THE ACM April 2005/Vol. 48, No. 4 31