Computer science isn't a science and it isn't about computers

By Jonah Kagan

August, 2012

As a computer scientist, there’s nothing that annoys me more than when my friends ask me for help setting up their wireless internet, or when my mom calls and asks why her laptop keeps freezing. I try to tell them that I’m not studying computer repairs or computer usage, I’m studying computer science.

But that doesn’t help, because nobody seems to know exactly what the term “computer science” means. When I urge my friends to take a computer science course, they shrug me off with comments like “I’m no good with computers” or “I don’t do science.” Assuming my friends aren’t just unadventurous, there must be some big misconceptions outside of the computer science community about what computer science is all about.

Computer scientists are concerned with questions like: How do you find the shortest route between two points on a map? How do you translate Spanish into English without a dictionary? How do you identify the genes that make up the human genome using fragments of a DNA sequence?

There’s a difference between the question, “How do you identify the genes that make up the human genome?” and the question, “What are the genes that make up the human genome?” The latter, a question posed by biologists, asks for a specific fact, while the former asks for a procedure which can produce that fact.

Consider any science: chemistry, biology, physics, or even one of the “soft” sciences like psychology. All are concerned with answering factual questions about the world around us. In computer science, the goal is not to figure out the answers to factual questions, but rather to figure out how to get answers. The procedure is the solution. While scientists want to figure out what is, computer scientists want to know how to.

This is not to say that scientists don’t ever need to know how to figure out the answers to their questions. The key distinction is that computer scientists care only about how to figure out the answer, and not what the answer is. Scientists, in some sense, either rely on computer science to help with their process (for instance, if they make use of data-analysis software) or are in part computer scientists themselves.

The distinction between questions of fact and questions of procedure leads naturally to a difference in methodology between scientists and computer scientists. When scientists come up with a possible answer to a question–a hypothesis–they try to prove or disprove it using experiments. Experiments are in essence tests to see whether a hypothesis matches the behavior of the natural world. If a hypothesis accounts for how the world behaves (or at least the behavior that the scientists can see), then it’s a useful theory.

We’re all familiar with this process from elementary school. It’s called the scientific method: you observe some occurence, come up with a hypothesis about it, test your hypothesis with experiments, and then analyze the results. This is how scientists justify the answers to their factual questions, and it’s how our society generates knowledge.1

Knowledge in computer science, however, doesn’t work the same way. Procedures don’t exist in the natural world–they’re devised by humans. When we come up with a procedure, we can’t just run experiments to see if it works. Although the procedure might be applied to data gathered from the real world, the procedure itself is not a part of nature. Think back to all the sciences I mentioned before. All of them seek knowledge about that which already exists. Procedures, however, are completely constructed–they only exist in the abstract.

For instance, consider the procedure used in a spell checker that recommends possible correct spellings when you make a typo. This procedure takes a sequence of letters and tries to find the closest match in a giant list of valid sequences, or as we normally call them, words. What separates this procedure from the real world problem of correcting spelling is that the sequences don’t have to represent words–that’s just one possible application for the procedure. The procedure itself can be reused with other kinds of sequences. In fact, this very same procedure is used for the DNA sequencing problem I mentioned before.2

Since the problems solved by computer scientists are defined separate from the real world, we can’t use the scientific method to analyze their validity. We can only analyze procedures within the realm of abstraction in which we have created them. Luckily, this type of reasoning is exactly why we have mathematical logic. Mathematicians, too, are concerned with the idea of truth in the abstract. Instead of running experiments, computer scientists define problems and procedures mathematically, and then analyze them using logic. This is the fundamental reason why computer science is not a science.

Given that the correctness of procedures is proved using mathematical logic, it might seem like computer science is really just a branch of mathematics, which it is, in some sense. In fact, much of the “math” we learn in school is actually computation.

Consider, for example, the problem of dividing two numbers. When presented with this problem, a mathematician might derive the properties of division, such as when there will be a remainder. A computer scientist, in contrast, would focus on figuring out how to perform the division.3

The computer scientist might eventually come up with the long division algorithm. Just like any 4th grader, however, he wouldn’t want to perform the division by hand. Instead, he would write a series of instructions, or program, describing how to perform the calculation, and tell a computer to execute it.

Notice that this is the first time I’ve mentioned computers at all. That’s because there’s nothing fundamental about procedures that requires the use of computers. Computers aren’t the only tools that can be used to execute programs. For instance, elementary school students are perfectly capable of executing the long division algorithm. We use computers instead of small children because computers are fast and reliable (after all, that’s why we built them), while small children are adorably uncoordinated and prone to unexpected naps.

The great computer scientist Edsger Dijkstra summed it up best: “Computer science is no more about computers than astronomy is about telescopes.” Even though a complex ecosystem of programs has developed, allowing computers to serve a variety of purposes, computers are still nothing more than a tool for executing procedures. Computer science is about the procedures themselves, not so much the tools used to execute them.

At this point, though, I should say that I haven’t painted an entirely accurate picture of the field–or rather, I left out some parts. There are probably some computer scientists reading this who are thinking, “This doesn’t describe my work at all.”

While at its core, computer science really is the pure study of procedures in the abstract as I described, in reality, the field has grown to encompass a wide variety of pursuits. Some computer scientists are concerned mostly with designing intricate systems that rely heavily on the specifics of computer architecture. Others study human-computer interaction, which actually does use the scientific method to determine what types of interfaces work the best for computer users.

It would be easy to dismiss the outliers and say they are not true computer scientists, that their work falls under the umbrella of some related but fundamentally different field. But I think the breadth of study within computer science is not necessarily a bad thing. It doesn’t need to be strictly defined.

Within the computer science department at my university, there’s a huge variety of interests among the students and professors. The multitude of perspectives complement each other, and help the field grow.

In the end, its the rate of growth of the field that makes all this definition business so tricky. Computer science is still young, and always undergoing new growth spurts. It’s that awkward teenage boy at the school dance whose limbs are growing so fast that he can’t make them all move together harmoniously just yet.

For now, I’m content to just think back to when I was an awkward teenager trying to figure out exactly who I was and how to express that identity to the rest of the world. I grew up and figured it out, and we computer scientists will eventually figure out our communal identity too.

  1. Of course, this sort of justification relies on the assumption that the natural world will continue to behave the same way as it has before, but that is a topic for another essay.

  2. It’s called the Damerau-Levenshtein distance algorithm.

  3. Schools, for some reason, don’t seem to care about deriving properties or developing procedures, but rather teaching students how to memorize and execute procedures (something computers happen to do quite well).

Thoughts? Am I crazy?
Let me know: