We propose a generalization of expected utility that we call generalized EU (GEU), where a decision maker's beliefs are represented by plausibility measures, and the decision maker's tastes are represented by general (i.e.,not necessarily real-valued) utility functions. We show that every agent, “rational” or not, can be modeled as a GEU maximizer. We then show that we can customize GEU by selectively imposing just the constraints we want. In particular, we show how each of Savage's postulates corresponds to constraints on GEU.

Many different rules for decision making have been introduced in the literature. We show that a notion of generalized expected utility proposed in [F. Chu, J.Y. Halpern, Great expectation. Part I: On the customizability of generalized expected utility, in: Proc. IJCAI-03, Acapulco, Mexico, 2003] is a universal decision rule, in the sense that it can essentially all other decision rules. This approach gives us a general technique for designing new decision rules as well as providing a framework for comparing decision rules to each other.

A standard assumption in the database query optimization literature is that it suffices to optimize for the "typical" case---that is, the case in which various parameters (e.g., the amount of available memory, the selectivities of predicates, etc.) take on their "typical" values. It was claimed in [CHS99] that we could do better by choosing plans based on their expected cost. Here we investigate this issue more thoroughly. We show that in many circumstances of interest, a "typical" value of the parameter often does give acceptable answers, provided that it is chosen carefully and we are interested only in minimizing expected running time. However, by minimizing the expected running time, we are effectively assuming that if plan p1 runs three times as long as plan p2, then p1 is exactly three times as bad as p2. An assumption like this is not always appropriate. We show that focusing on least expected cost can lead to significant improvement for a number of cost functions of interest.

We argue that the tools of decision theory need to be taken more seriously in the specification and analysis of systems. We illustrate this by considering a simple problem involving reliable communication, showing how considerations of utility and probability can be used to decide when it is worth sending heartbeat messages and, if they are sent, how often they should be sent.

We identify two unreasonable, though standard, assumptions made by database query optimizers that can adversely affect the quality of the chosen evaluation plans. One assumption is that it is enough to optimize for the expected case—that is, the case where various parameters (like available memory) take on their expected value. The other assumption is that the parameters are constant throughout the execution of the query. We present an algorithm based on the “System R”-style query optimization algorithm that does not rely on these assumptions. The algorithm we present chooses the plan of the least expected cost instead of the plan of least cost given some fixed value of the parameters. In execution environments that exhibit a high degree of variability, our techniques should result in better performance.