TITLE: More on the Nature of Computer Science AUTHOR: Eugene Wallingford DATE: September 16, 2008 9:43 PM DESC: ----- BODY: Another entry generated from a thread on a mailing list... A recent thread on the SIGCSE list began as a discussion of how programming language constructs are abstractions of the underlying hardware, and what that means for how students understand the code they write. For example, this snippet of Java:
    int x = 1;
    while (x > 0)
does not result in an infinite, because Java ints are not integers. This is one of many examples that remind us how important it is to study computer organization and architecture, and more generally to learn that abstractions are never 100% faithful to the details they hide. If they were, they would not be abstractions! A few good abstractions make all the difference in how we work, but -- much like metaphor -- we have to pay attention to what happens at their edges. Eventually, the thread devolved toward a standard old discussion on this list, "What is Computer Science?" I conjecture that every mailing list, news group, and bulletin board has a topic that is its "fixed point", the topic toward which every conversation ultimately leads if left to proceed long enough, unfettered by an external force. Just about every Usenet newsgroup in which I participated during the late 1980s and early 1990s had one, and the SIGCSE list does, too. It is, "What is Computer Science?" This question matters deeply to many people, who believe that graduates of CS programs have a particular role to play in the world. Some think that the primary job of undergraduate CS programs is to produce software engineers. If CS is really engineering (or at least should be thought of that way for practical reasons), then the courses we teach and the curricula we design should have specific outcomes, teach specific content, and imbue in students the mindset and methodology of an engineer. If CS is some sort of liberal art, then our courses and curricula will look quite different. Much of this new thread was unremarkable if only because it all sounded so familiar to me. One group of people argued that CS is engineering, and another argued that it was more than engineering, perhaps even a science. I must have been in an ornery mood, because one poster's assertion provoked me to jump into the fray with a few remarks. He claimed that CS was not a science, because it is not a "natural science", and that it is not a natural science because the object of its study is not a natural phenomenon:
I don't believe that I have ever seen a general purpose, stored-program computing device that occurs in nature... unless we want to claim that humans are examples of such devices.
This seems like such a misguided view of computer science, but many people hold it. I'm not surprised that non-computer scientists believe this, but I am still surprised to learn that someone in our discipline does, too. Different people have different backgrounds and experiences, and I guess those differences can lead people to widely diverging viewpoints. Computer science does not study the digital computer. Dijkstra told us so a long time ago, and if we didn't believe him then, we should now, with the advent of ideas such as quantum computing and biological computing. Computer science is about processes that transform information. I see many naturally-occurring processes in the world. It appears now that life is the result of an information process, implement in the form of DNA. Chemical processes involve information as well as matter. And some physicists now believe that the universe as we experience it is a projection of two-dimensional information embodied in the interaction of matter and energy. When we speak of these disciplines, we are saying more than that computer scientists use their tool -- a general-purpose computation machine -- to help biologists, chemists, and physicists do science in their areas. We are talking about a more general view of processes and information, how they behave in theory and under resource constraints. Certainly, computer scientists use their tools to help practitioners of other disciplines do their jobs differently. But perhaps more important, computer scientists seek to unify our understanding of processes and information across the many disciplines in which they occur, in a way that sheds light on how information processing works in each discipline. We are still at the advent of the cycle feeding back what we learn from computing into the other disciplines, but many believe that this is where the greatest value of computer science ultimately lies. This means that computer science is wonderful not only because we help others by giving them tools but also because we are studying something important in its own right. If we broaden our definition of "naturally occurring" to include social phenomena in large. complex systems that were not designed by anyone in particular, then the social sciences give rise to a whole new class of information processes. Economic markets, political systems, and influence networks all manifest processes that manipulate and communicate information. How do these processes work? Are they bound by the same laws as physical information processing? These are insanely interesting questions, whose answers will help us to understand the world we live in so much better than we do now. Again, study of these processes from the perspective of computer science is only just beginning, but we have to start somewhere. Fortunately, some scientists are taking the first steps. I believe everything I've said here today, but that doesn't mean that I believe that CS is only science. Much of what we do in CS is engineering: of hardware systems, of software systems, of larger systems in which the manipulation of information is but one component. Much of what we do is mathematics: finding patterns, constructing abstractions, and following the implications of our constructions within a formal system. That doesn't mean computer science is not also science. Some people think we use the scientific method only as a tool to study engineered artifacts, but I think that they are missing the big picture of what CS is. The fact that people within our discipline still grapple with this sense of uncertainty about its fundamental nature does not disconcert me. We are a young discipline and unlike any of the disciplines that came before (which are themselves human constructs in trying to classify knowledge of the world). We do not need to hide from this unique character, but should embrace it. As Peter Denning has written over the years Is computer science science? Engineering? Mathematics? The answer need not be one of the above. From different perspectives, it can be all three. Of course, we are left with the question of what it is like for a discipline to comprise all three. Denning's Rebooting Computing summit will bring together people who have been thinking about this conundrum in an effort to make progress, or chart a course. On the CS education front, we need to think deeply about the implications of CS's split personality for the design of our curricula. Owen Astrachan is working on innovating the image of CS in the university by turning our view outward again to the role of computer science in understanding a world bigger than the insides of our computers or compilers. Both of these projects are funded by the NSF, which seems to appreciate the possibilities. I can't think about the relationship between computer science and natural science with thinking of Herb Simon's seminal Sciences of the Artificial. I don't know whether reading it would change enough minds, but it affected deeply how I think about complex systems, intentionality, and science. -----