Utility in case-based decision theory (Q697850)

The paper provides two axiomatic characterizations of a case-based decision rule. Given a decision problem \(p\) and memory \(M\) the decision maker chooses an act that maximizes the utility \(U(a)=U_{p,M}(a) = \sum_{(q,b,r) \in M} s((p,a),(q,r))u(r)\). The memory \(M\) consists of cases \((p,a,r)\) where \(p\) is a decision problem, \(a\) is an act and \(r\) an outcome, \(s\) is a similarity function. The paper considers two ways to construct hypothetical memories and derives axiomatic characterizations of the decision rule from both. In the first only actual outcomes can be replaced with other outcomes from \(R\), where \(R\) is, e.g., an interval of real numbers. In the second, the set of outcomes \(R\) is finite and combinations of outcomes and circumstances from actual memory are allowed. The first model leads to an axiomatic characterization in an outcome tradeoff approach, the second in a case repetition approach. Proofs of the two main theorems are given in an appendix.