In attempting to capture the information search (learning) and decision-making processes in decisions from experience, researchers have proposed models that can be grouped into three classes. The first class—neo-Bernoullian models—rests on the premise that respondents form a mental representation of the relative frequency (probability) with which events occur in the process of sampling outcomes. Combined with outcome information, these probabilities then enter the evaluation of the two gambles’ desirability. But do decisions from experience inevitably give rise to an explicit representation of probabilities? The second and the third class of models—associative learning models and heuristics—reflect the assumption that decision-makers can and will do without probabilities. In this section we discuss the three classes of models.

Neo-Bernoullian Models

Expected utility theory postulates that one can, or should, model human choice by assuming that people behave as if they have multiplied some function of probability and value and then have maximized it. Applied to decisions from experience, expected utility theory and related models require explicit representation of probabilities. An example is the “two-stage model” (Tversky & Fox, 1995, p. 279) of decision under uncertainty, in which it is assumed that decision-makers first estimate the probability p of an uncertain event A and then make a choice. The psychological impact of the event A with its associated (estimated) probability p is then measured in terms of cumulative prospect theory’s probability weighting function n (Fox & Tversky, 1998).