Conditions for Maximizing Expected Value in Repeated Choices from Experience

It is largely expected that people can learn from past experiences and use this knowledge to make better decisions in the future. However, there are aspects inherent in experiential learning which may affect the extent to which people can extract and use information from experiential feedback to make advantageous decisions. Three aspects inherent in experiential learning were identified: (1) it is reliant on memory, (2) information is gathered exclusively through outcome feedback, and (3) outcome feedback is inherently dynamic. The current investigation explored how each of these aspects may help shape experiential decision making, and examined how the presence of competing types of information might hinder the ability of experiential information to guide people towards advantageous choices. A card-selection paradigm was used to examine learning about monetary outcomes from repeatedly sampling from two decks with different expected values (EVs, i.e., average payoffs). Effects on working memory were assessed by varying the number of outcomes within each deck and varying whether both decks had all-gain outcomes or one deck had some zero outcomes. Reliance on outcome feedback was manipulated by adding misleading (but technically correct) descriptive information which favored the less advantageous deck. To assess the impact of dynamic information, the dynamics of experience were contrasted with misleading dynamic descriptions. The primary dependent variable was the number of higher EV deck selections measured during the first and last 25 choices. The results of the investigation revealed little strain on working memory, but found a surprise zero effect in which identification of the more advantageous option was noticeably disrupted when the better option contained possible zero outcomes. Participants seemed drawn to options that were less advantageous but had only gain outcomes. Misleading descriptions provided at the outset only disrupted advantageous choice when zero outcomes were involved, but outcome feedback was found to help overcome the initial bias toward the lower EV all-gain deck. However, when no description was available, the zero effect grew more intense with experience. Finally, when misleading dynamic descriptions were presented, disruptions in experiential learning were seen throughout. The implications of these results contribute to our understanding of which conditions are likely to support versus disrupt our ability to use experiential feedback to guide us towards advantageous choices