Choosers Adapt Value Coding to the Environment, But Do Not Attain Efficiency

Abstract

We investigate how human choosers adapt their value encoding strategy to the statistics of the choice environment. Specifically, we ask whether the human value encoding mechanism exhibits divisive normalization only in the Pareto-distributed environments in which it is information-maximizing. To test this theory, we conduct a risky choice experiment in which subjects are presented with blocks of choice stimuli drawn from either a Pareto-distributed environment or a uniform-distributed environment. Our results show that subjects exhibit some degree of normalization regardless of whether it is efficient or not, but do adapt the curvature of their encoding function to the environment. These findings suggest that human value coding mechanisms are flexible but biologically constrained to be perfectly efficient only in specific environments. This study provides new insights into the neural mechanism of human decision-making and the role of environmental statistics in shaping it.