Biased samples not mode of presentation: Re-examining the apparent underweighting of rare events in experience-based choice

Most experimental investigations of risky choice provide participants with a description of the probabilities and outcomes for each option, and observe that small probabilities are overweighted. However, when payoffs are learned from repeated experience of outcomes (as in many real-world decisions), different patterns of choice are observed--consistent with underweighting rare events. We re-examined this phenomenon to determine whether biased sampling and recency effects in experience-based choice could account for this description-experience gap. Two hundred and forty paid participants made choices for 12 pairs of simple gambles. In the objective description condition, probabilities and outcomes were specified. In the free sampling condition, participants observed repeated plays of each gamble before choosing. Participants in four yoked conditions received the same information as the free sampling participants--either described or experienced. Differences between objective description and free sampling were consistent with underweighting rare events in experience-based choice. However, consistent with a biased sampling account, patterns of choice in the yoked conditions barely differed from the free sampling condition: given identical information, presentation mode has no effect. Recency effects in choice occurred only when outcomes were actively sampled, and were unaffected by working memory capacity. The absence of recency for passive observation implies actor-observer differences in forming expectations or testing hypotheses. The results provide no support for the claim that decisions from description and decisions from experience require separate descriptive theories.