Imagine a computer with a rather large memory -- a combination of magnetic core and tape, say -- that contains, among others, the general problem-solving program augmented by programs for specific tasks like playing chess, detecting serial patterns, solving differential equations, inverting matrices, calculating correlation coefficients, and so on. Each job that is input to the computer is examined and assigned a priority (on some basis that need not concern us) that gradually changes with the length of time the job has been waiting to be processed. When the priority of a job reaches a sufficiently high level, the job that is currently being processed in interrupted and replaced by the high priority job.

Such a computer (and, of course, systems organized in this general way are already in existence) would exhibit, in its behavior, motivation and a set of values. If we noticed that it gave high priorities to matrix inversions, we would say that this was an activity it preferred. Suppose we also noted that when certain brief new jobs were input, it immediately interrupted what it was doing to undertake one of the new tasks before returning to the original one. We might say that it was easily distracted, or even that it was exhibiting emotion.

I do not propose here to develop in detail the idea that the core of the behavior we call emotional derives from a mechanism for interrupting the ongoing stream of activity. However, this notion is consistent with a good deal of empirical evidence about the nature of emotion and provides an interesting avenue of exploration into the relation of emotion to cognitive activity. It suggests that we shall not be able to write programs for computers that allow them to respond flexibly to a variety of demands, some with real-time priorities, without thereby creating a system that, in a human, we would say exhibited emotion.

Herbert A. Simon
Thinking by Computers, 1966