The construction and deconstruction of sub-optimal preferences through range-adapting reinforcement learning

Converging evidence suggests that economic values are rescaled as a function of the range of the available options. Critically, although locally adaptive, range adaptation has been shown to lead to suboptimal choices. This is particularly striking in reinforcement learning (RL) situations when options are extrapolated from their original context. Range adaptation can be seen as the result of an adaptive coding process aiming at increasing the signal-to-noise ratio. However, this hypothesis leads to a counterintuitive prediction: decreasing outcome uncertainty should increase range adaptation and, consequently, extrapolation errors. Here, we tested the paradoxical relation between range adaptation and performance in a large sample of subjects performing variants of a RL task, where we manipulated task difficulty. Results confirmed that range adaptation induces systematic extrapolation errors and is stronger when decreasing outcome uncertainty. Finally, we propose a range-adapting model and show that it is able to parsimoniously capture all the observed results.