Meeting 21: Decisions and Utility

Reading: AIAMA 16.1-16.3

In order to make an action decision an agent needs to know more than the probability of an outcome, it also needs the cost associated with making a mistake. This can be expressed as a gain or loss in terms of a utility. In this meeting we will look at how to combine probabilities with utility to reach rational decisions under uncertainty.

Questions you should be able to answer after reading are:

  1. What is a utility function?
  2. How does one compute the expected utility?
  3. Why is it rational to maximize the expected utility?
  4. What are the constraints on rational preferences?
  5. How do rational preferences map onto utility functions?
  6. What are some semantic interpretations of utility in the real world?
  7. What is the difference between a normative and descriptive theory and how does it apply to human decisions?