A principle in economics, political economy, and/or game theory. Simply put it is a theory which attempts to predict behavior based on something more than cost/benefit analysis. The idea here is that "utility" is often a better indicator of rational behavior.

Utility can be loosely defined in this context as the pleasure or gratification one gets from a specific activity or the expectation of this pleasure (hence "expected"). This concept can be applied with probability to make very interesting arguments for and against gambling.

A decent bibliography for further study of utility theory can be found here (http://cepa.newschool.edu/het/essays/uncert/choiceref.htm)

An interesting application of this theory is something I cooked up when my wife and I were trying to figure out what home improvement projects to undertake this summer. I created a system of multivoting where the two of us would give weight to the possible projects based on criteria that would mean the greatest utility for us. Some of the things we looked at were energy efficiency, cost savings, safety, etc. Some of these concepts were more important to us than others, and were thus weighted higher. At the end of our voting session, I came up with a number (a factor of expected utility, if you will) that when the cost of the project is divided by this factor to come up with a figure that is, if you will, a weighted cost of the project per unit of utility. In this way we were able to prioritize.

Utility theory normally is used, as novasoy writes, in the context of expected utility. To illustrate:

Imagine that I'm playing the lottery. The odds that I will win are one in a million, say, and it costs me five dollars to play. However, if I win, I take home, say, one hundred thousand dollars. The expected utility calculation that utility theory posits is that you can determine whether or not to play by solving the following problem.

(1/1,000,000) X \$100,000 >? (999,999/1,000,000) X \$5

We see that the inequality above is not satisfied. Therefore, you have a higher expected utility from not playing than from playing. So you should not play the lottery.

It is evident from the above example that expected utility theory is a powerful tool in the analysis of behavior in politics, economics, and other fields of human interaction. It should also be evident, though, that the problem with applying such an analysis to a problem is the possibility of tautology.

Imagine that I wish to figure out why someone voted for Candidate A. I assume that we should use expected utility theory to analyze this problem. Accordingly, I say, "Voting for Candidate A must have given this person her highest expected utility." I then look at her other choices (Candidates B through N), and discern that all other Candidates espoused platforms that give the voter a lower expected utility. Why? Because if not, she would have voted for someone else. So I've solved the problem. Or have I?

The idea of a positivist social science epistemology is that, conceivably, I should be able to be proven wrong. How, if I adopt the analysis above, could I have been proven wrong? I have argued that utility theory gives me a viable explanation for my problem. I have achieved this by assuming utility theory to be correct. If I assume that utility theory is correct, than I can have no result that contradicts the viability of expected utility theory. I cannot be proven wrong. So I can never know if my assumption of expected utility theory is driving my conclusions. We have a tautology.

Such are the strengths and weaknesses of expected utility theory.