Most of you have seen computer games. Even the simplest ones are quite difficult to program, because you have to use the very limited mechanisms the machine has for communication. When you instruct it to do something that it can do, your instruction has to be precisely organized in such a way that the information can be processed so that the computer can perform the task. Brains, like computers, are not "user–friendly." They do exactly what they're told to do, not what you want them to do. Then you get mad at them because they don't do what you meant to tell them to do!

One programming task is called modeling, which is what I do. The task of modeling is to get a computer to do something that a human can do. How do you get a machine to evaluate something, do a math problem, or turn a light on or off at the right time? Human beings can turn a light on and off, or do a math problem. Some do it well, others do it well sometimes, and some don't do it well at all. A modeler attempts to take the best representation for the way a person does a task, and make it available in a machine. I don't care if that representation really is how people do the task. Modelers don't have to have truth. All we have to have is something that works. We are the people who make cookbooks. We don't want to know why it is a chocolate cake, we want to know what to put in it to make it come out right. Knowing one recipe doesn't mean there aren't lots of other ways to do it. We want to know how to get from the ingredients to chocolate cake in a step–by–step fashion. We also want to know now to take chocolate cake and work backwards to the ingredients when someone doesn't want us to have the recipe.

Breaking down information in this way is the task of an information scientist.