Normative problems in the psychology of reasoning: deductive constraints on Bayesian rationality and the problem of arbitration

This paper addresses the question of which normative standards apply in the psychology of reasoning. In a first strand, the present paper contributes to a debate concerning whether the New Paradigm in the psychology of reasoning really constitutes a paradigm shift by presenting an argument concerning the deductive competence presupposed by ideal Bayesian agents. On the construal of the New Paradigm as the replacement of deductive logic by probability theory, no such shift was possible: the standard probabilistic frameworks used in the psychology of reasoning were never independent of logical principles to begin with. It is possible, however, to replace the treatment of a fragment of natural language (e.g. conditionals, quantifiers) in classical logic with probabilistic alternatives, as the paper discusses. Thus construed, the probabilistic turn was a major step forward in the psychology of reasoning but one that is better pruned of some of the misleading characterizations that surround it. When multiple norms apply to experimental tasks, a further practical problem presents itself: how should researchers arbitrate between them? In a second strand, an experimental research program is outlined which confronts this problem of arbitrating between competing rationality norms by presenting a way of making rationality assessments of participants based on verbal arguments. In this part, a recent attempt to dissolve problems of arbitration in psychology is critically discussed.