Defeasible Decisions: What the Proposal is and Isn't

Abstract

Publisher Summary This chapter discusses defeasible reasoning for decisions, which is the natural extension of philosophers' defeasible practical reasoning about action. The difference in it is that the arguments for actions are quantitative, often invoking expected utility calculations. In practical reasoning, reasoning about action is qualitative. If an act achieves a goal, that is a reason for performing that act. If an act achieves a goal but also invokes a penalty and that penalty is more undesirable than the goal itself, then that may be the reason of not performing the act. However, qualitative practical reasoning is a very weak way of analyzing tradeoffs. It does not take into account known risks of actions, that is, the known probabilities of acts achieving various effects. The chapter presents a quantitative version of defeasible reasoning about decisions. An act achieves an effect with known probability and there are independent reasons for the utilities of each of the resulting states. By weighing these independent utilities by their respective probabilities, an argument for the utility of the act is produced. With different independent reasons for the utilities of resulting states, one gets different arguments. With different accounting of the possible results of an act, different arguments are produced. Reasons can be written in an existing formalism for defeasible reasoning in such a way that those arguments that justify their conclusions are exactly those arguments that one would consider compelling among the multitude of potentially conflicting arguments.