Rationality in the new paradigm: Strict versus soft Bayesian approaches

Abstract

The new paradigm in psychology of reasoning is a family of theories sharing common assumptions and research goals. However, there are still some issues under debate. Although authors within the paradigm broadly agree on Bayesianism as a suitable theoretical framework, approaches vary considerably in adopting Bayesian principles to rationality, creating a continuum between “strict” and “soft” Bayesianism. Stricter Bayesian approaches adopt an epistemological stance which postulates strong coherence between the normative, computational, and algorithmic levels of analysis, all of which conform with Bayesian strictures; the closer an approach is to the softer end of the scale, the less it tends to postulate such strong coherence. In stricter Bayesian models instrumental rationality, which depends on achieving one's goals, is inexorably intertwined with normative rationality, which depends on conforming to normative criteria. In contrast, soft Bayesians adopt Bayesianism for its psychological advantages, such as subjectivity, uncertainty, and degrees of belief. This drives a more descriptivist approach to rationality, in which instrumental rationality is neither defined nor justified by normative Bayesian models. The challenge for soft Bayesian approaches is to link more explicitly instrumental rationality to these psychological Bayesian features, while avoiding drawing on normative considerations such as Dutch book arguments.