The Description–Experience Gap in Loss Aversion

Abstract

Humans make different decisions under risk when they acquire information about their options through experience than when they learn it from a summary description. Research on this description–experience gap has largely focused on differences in probability weighting—for instance, how low-probability events are reflected in people’s decisions. Here, we show that the gap manifests much more broadly and also occurs with regard to the influence of outcomes. Specifically, we investigated how description- and experience-based decisions differ in loss aversion—the tendency for loss outcomes to have a greater impact on choice than equivalent gain outcomes. We first conducted a meta-reanalysis synthesizing raw data of all available datasets that directly compared description- and experience-based decisions involving mixed-domain lotteries (88,409 choices in total across N_Desc = 439 and N_Exp = 468). The results indicate that loss aversion, measured with cumulative prospect theory (CPT), is more pronounced in experience-based than in description-based choices. We then replicated this description–experience gap in loss aversion in a large-scale experiment with preregistered methods (19,435 choices by N_Desc = 202 and N_Exp = 202) tailored to address potential limitations of the studies included in the meta-analysis, and also showed the gap in an analysis that does not depend on estimating a latent loss aversion parameter. An analysis of the experience-based choices with a reinforcement learning model suggests that a possible mechanism for the apparently higher loss aversion in decisions from experience could be asymmetric learning after positive versus negative prediction errors during experiential sampling.