I've been making a bunch of comments on Computer Science education lately. The New York Times has an excellent article about "Making Computer Science More Enticing" which focuses on Stanford's new curriculum. The Stanford curriculum is very similar to IU's new specialization-based curriculum and seems to be an excellent approach to "teaching the discipline".
Also, I found the "definitive" document on CS education - The ACM/IEEE Computing Curriculum 2008 Update [PDF].
Why so much focus on education? Computer Science is a (relatively) new discipline with a multitude of high-impact applications, giving us an imperative to train students quickly. Unfortunately, the speed at which our field is moving can cause us to lose sight of the philosophy behind the science.
If someone wants to learn Biology, you would point them to Campbell & Reece. If someone wants to learn computation, where do you point them? A list of books. There are books focused on introducing algorithms and functional programming (SICP); there are tomes focused on general computation (Knuth); there are books focused on application (the entire O'Reilly library); there are definitive texts on specific languages (The C Programming Language, The Scheme Programming Language); there does not seem to be a widely-accepted, integrative introduction that emphasizes computation -- algorithms and models. From what I'm observing in CS curricula across the country, the coursework is moving in this direction, but we still need this cohesive "Introduction to Computing" book.
As a final message, this video linked in the NYT article captures the beauty, richness and excitement of our discipline right now -- "It's sort of like you're geometers and you're living in the time of Euclid":
Sunday, February 21, 2010
Tuesday, February 16, 2010
Computer Studies
The latest issue of Communications of the ACM, the premier computer science journal contains an interesting article by IU Professor Dennis Groth -- Why an Informatics Degree? The article has much to say about the necessity of application and applied computing as a measure of computer science success.
However, there are some questions left unanswered and one answered incorrectly. First, I address two questions in philosophy of science: "What is Computer Science?" and "Why Informatics?" I then address the pedagogical implications of these questions in a section on "Computer Studies".
This conflation of implementation and quantum is extremely pervasive. For example, Biology is "the study of life", but in a (typical) biology class one never addresses the basic question: "What is life?" The phenomena of life can be studied independently of the specific carbon-based implementation we have encountered. This doesn't deny the practical utility of modern biology, but it does raise the question of how useful our study of the applied life is to our understanding of life itself. (If you're interested in this line of questioning, I highly recommend Larry Yaeger's course INFO-I486 Artificial Life.)
Similarly, Computer Science can study procedures independently of the hardware and software implementations. Consider the sorting problem. (If you are unfamiliar with sorting, see the Appendix: Sorting Example.) One would not start by looking at processor architecture or software deisgn, but would instead focus on the algorithm. Pure Computer Science has nothing to do with hardware or software, they are just an extremelypractical medium on which we experiment.
The first definition defines a methodology but does not define a subject. It should be obvious that we live in a collaborative, interdisciplinary world. Fields should inform one another but there is still a distinction between fields: Biology studies life; Computer Science studies computation; Cognitive Science studies cognition; Chemistry studies chemicals; etc. One can approach any problem with any number of techniques - computing is one part of this problem-solving toolkit, along with algebra, calculus, logic and rhetoric. However, each of the particular sciences should answer some natural question - whether that be a better explanation of life, computation, mathematics or cognition. Positing a discipline as the use of one field to address problems in another field is not a new field. It's applied [field] or [field] engineering.
The other definition, that informatics studies the interaction of people and technology, hints at a new discipline studying a quantum of "interaction". This area has tons of exciting research, especially in human-computer interaction (HCI) and epidemiology. Further emphasizing this would go a long ways toward creating a new discipline and set a clear distinction between the informaticist and the computer scientist. Computer scientists study computation; informaticists study interaction; both should be encouraged. As it stands, both study "computers" and both step on each other's toes.
However, as currently implemented at IU, the Informatics curriculum is a liberal arts degree in technology. Formal education should teach either a vocation, a discipline or (ideally) both. Informatics seems to answer to neither claim by emphasizing how informaticists "solve problems with computers" without diving into programming or modeling. If it aims to teach such a vocation, then more application is necessary to give expertise; if it aims to teach a discipline, it is fine to do that through application, but we must recognize that application is only useful insofar as it benefits theory (and vice versa). Additionally, if the field does indeed have a quantum of interaction, then interaction should be the forefront of the curriculum.
IU's Computer Science ex-department is a valiant effort to teach a discipline - in the span of 4 years we cover at least 3 distinct programming paradigms (functional, object-oriented and logic) spread over 4 distinct languages, bristling with an exploration of algorithms. That being said, I would be surprised if more than 25% of the graduating class could explain a Turing Machine.
Not everyone is into theory - most people really just want to "solve problems with computers" and have a good job. Where do these programmers go? Informatics does not address this challenge, and shouldn't attempt to. The answer is software engineering - just as applied physics finds a home in classical engineering. By establishing a third program for those clearly interested in application, IU would have a very solid "computer studies" program (as distinguished from computation or technology). [A friend has pointed out that IU cannot legally offer an engineering degree, so we'd have to get creative on the name or tell people to go to Purdue. This works as a general model of Computer Studies pedagogy.]
As another example of how to split "computer studies", Georgia Tech recently moved to a three-prong approach with the School of Computer Science (CS), School of Interactive Computing (IC), and Computational Science and Engineering Division (CSE). My view of Informatics roughly correlates to that of IC; the Computer Science programs are equivalent but include software engineering. The CSE division is a novel concept, presently captured by IU's School of Informatics, and it seems this is another working group, but I feel it is best captured by adjunct faculty and interdisciplinary programs, rather than a whole new field.
However, there are some questions left unanswered and one answered incorrectly. First, I address two questions in philosophy of science: "What is Computer Science?" and "Why Informatics?" I then address the pedagogical implications of these questions in a section on "Computer Studies".
What is Computer Science?
Any new discipline needs to consider its philosophy in order to establish a methodology and range of study. Prof. Groth's definitions of Computer Science and Informatics do not quite capture these considerations:Computer science is focused on the design of hardware and software technology that provides computation. Informatics, in general, studies the intersection of people, information, and technology systems.In explicitly linking the science to its implementation, this definition of Computer Science fumbles away its essence. Yes, the technology is important and provides a crucial instrument on which to study computation, but at its core computer science studies computation -- information processing. Computer science empirically examines this question by studying algorithms (or procedures) in the context of a well-defined model (or system).
This conflation of implementation and quantum is extremely pervasive. For example, Biology is "the study of life", but in a (typical) biology class one never addresses the basic question: "What is life?" The phenomena of life can be studied independently of the specific carbon-based implementation we have encountered. This doesn't deny the practical utility of modern biology, but it does raise the question of how useful our study of the applied life is to our understanding of life itself. (If you're interested in this line of questioning, I highly recommend Larry Yaeger's course INFO-I486 Artificial Life.)
Similarly, Computer Science can study procedures independently of the hardware and software implementations. Consider the sorting problem. (If you are unfamiliar with sorting, see the Appendix: Sorting Example.) One would not start by looking at processor architecture or software deisgn, but would instead focus on the algorithm. Pure Computer Science has nothing to do with hardware or software, they are just an extremelypractical medium on which we experiment.
Why Informatics?
Informatics seems to be ahead of itself here in asking "Why an Informatics degree?" before asking the more fundamental "Why Informatics?" There are two primary definitions implied in the article. The more popular answer is that "Informatics solves interdisciplinary problems through computation". The second, emerging answer is that "Informatics studies the interaction of people and technology".The first definition defines a methodology but does not define a subject. It should be obvious that we live in a collaborative, interdisciplinary world. Fields should inform one another but there is still a distinction between fields: Biology studies life; Computer Science studies computation; Cognitive Science studies cognition; Chemistry studies chemicals; etc. One can approach any problem with any number of techniques - computing is one part of this problem-solving toolkit, along with algebra, calculus, logic and rhetoric. However, each of the particular sciences should answer some natural question - whether that be a better explanation of life, computation, mathematics or cognition. Positing a discipline as the use of one field to address problems in another field is not a new field. It's applied [field] or [field] engineering.
The other definition, that informatics studies the interaction of people and technology, hints at a new discipline studying a quantum of "interaction". This area has tons of exciting research, especially in human-computer interaction (HCI) and epidemiology. Further emphasizing this would go a long ways toward creating a new discipline and set a clear distinction between the informaticist and the computer scientist. Computer scientists study computation; informaticists study interaction; both should be encouraged. As it stands, both study "computers" and both step on each other's toes.
Computer Studies
This discussion of philosophies has important implications for how we structure computer-related education (formalized as Computer Studies). Despite major differences in our approaches, it does seem clear that Computer Science and Informatics should work together, especially in applications.However, as currently implemented at IU, the Informatics curriculum is a liberal arts degree in technology. Formal education should teach either a vocation, a discipline or (ideally) both. Informatics seems to answer to neither claim by emphasizing how informaticists "solve problems with computers" without diving into programming or modeling. If it aims to teach such a vocation, then more application is necessary to give expertise; if it aims to teach a discipline, it is fine to do that through application, but we must recognize that application is only useful insofar as it benefits theory (and vice versa). Additionally, if the field does indeed have a quantum of interaction, then interaction should be the forefront of the curriculum.
IU's Computer Science ex-department is a valiant effort to teach a discipline - in the span of 4 years we cover at least 3 distinct programming paradigms (functional, object-oriented and logic) spread over 4 distinct languages, bristling with an exploration of algorithms. That being said, I would be surprised if more than 25% of the graduating class could explain a Turing Machine.
Not everyone is into theory - most people really just want to "solve problems with computers" and have a good job. Where do these programmers go? Informatics does not address this challenge, and shouldn't attempt to. The answer is software engineering - just as applied physics finds a home in classical engineering. By establishing a third program for those clearly interested in application, IU would have a very solid "computer studies" program (as distinguished from computation or technology). [A friend has pointed out that IU cannot legally offer an engineering degree, so we'd have to get creative on the name or tell people to go to Purdue. This works as a general model of Computer Studies pedagogy.]
As another example of how to split "computer studies", Georgia Tech recently moved to a three-prong approach with the School of Computer Science (CS), School of Interactive Computing (IC), and Computational Science and Engineering Division (CSE). My view of Informatics roughly correlates to that of IC; the Computer Science programs are equivalent but include software engineering. The CSE division is a novel concept, presently captured by IU's School of Informatics, and it seems this is another working group, but I feel it is best captured by adjunct faculty and interdisciplinary programs, rather than a whole new field.
Appendix: Sorting Example
Let's say we have a list of numbers and want to sort them from smallest to largest. One naive way is to compare each term to the next one, and swap them if they are in the wrong order and restart until you can make it to the end without swapping:1: *4 3* 2 1 -> 3 *4 2* 1 -> 3 2 *4 1* -> 3 2 1 4 2: *3 2* 1 4 -> 2 *3 1* 4 -> 2 1 *3 4* -> 2 1 3 4 3: *2 1* 3 4 -> 1 *2 3* 4 -> 1 2 *3 4* -> 1 2 3 4 4: *1 2* 3 4 -> 1 *2 3* 4 -> 1 2 *3 4* -> 1 2 3 4
This is called bubble sort, and solves the problem of sorting. However, consider what you'd have to do to sort a bigger list: each time you make a swap you have to rescan the whole list! A smarter way to sort this list would be to divide the list into two smaller lists, sort the smaller lists, and then merge them together:
1a: *4 3* -> 3 4 1b: *2 1* -> 1 2 Now merge by comparing the items in each list: 2a: *3* 4 -> *3* 4 -> 1 2 3 4 2b: *1* 2 -> 1 *2* -^
This only takes 4 comparisons, compared to 12! We just did a classic problem in Computer Science without even once mentioning computer hardware or writing a single line of code!
Subscribe to:
Posts (Atom)
Posts
