Towards a theory of decision-making with paradoxes.

Human subjects often violate the rational decision-making theory, which is based on the notion of expected utility and axioms of choice (Neuman & Morgenstern, 1944; Savage, 1954). The counterexamples, suggested by Allais (1953) and Ellsberg (1961), deserve special attention because they point at our lack of understanding of how humans make decisions. The paradoxes of decision-making are particularly important for the ACT-R theory which currently relies on expected utility. The paper presents two alternative methods: a random prediction method that uses subsymbolic computations and a method that uses symbolic reasoning for qualitative decision-making. Both methods are tested on ACT-R models of the paradoxes, and the advantages of each method are discussed