Simultaneous Learning of Perceptual Ambiguity and Expected Value Estimates Alter Preferences, Discrimination Performance, and Choice

Abstract

The ability to simultaneously learn how to distinguish between visually similar unknown items and their value was tested in three studies conducted across multiple days. In each study, participants took the role of a merchant in a market for obscure fruits. Merchants traded unknown fruits and earned money by identifying ripe ones. The ripe fruits had a fixed positive value, while the unripe ones had varying values at certain unknown probabilities. Participants had to learn that one fruit category (ripe or unripe) was always associated with a fixed monetary gain, whereas another category was associated with a lottery that had either monetary gains or losses. The fruit categories were always perceptually similar. We show that learning the value of items typically leads to (1) improvement in the ability to visually discriminate between high/low-value ones, and (2) a change in subjective preference towards the high-value items. When item values are ambiguous and determined by lotteries, participants do not show preference change. In those ambiguous cases, as the values of items become known, individuals’ risk-tolerance is predictive of item preferences such that a reinforcement learning model can be used to estimate the preferences. Despite the consistent behavior by the participants, their reflections on the decision heuristics typically fail to recognize key drivers of the rational choices. This suggests a discrepancy between the explicit interpretation of the choice heuristics and the intuitive decision practices. We propose a rationale for the behavioral outcomes and ways to use the novel experimental paradigm for further decision theory explorations.