Grid-like representations of decision vectors in subjective value space for human risky choice

Abstract

Individuals are widely thought to make choices based on subjective valuations of options that combine multiple attributes into a unified subjective value signal in the brain’s value coding system. Yet how a choice options’ attributes are transformed into a subjective valuation remains poorly understood. One candidate is the cognitive mapping system in the entorhinal cortex (EC) and medial prefrontal cortex (mPFC), which efficiently represents relational, multi-dimensional information and is thought to afford inferences. Here, we develop a risky decision-making task and use fMRI to show that a two dimensional (2D) subjective value space of reward probability and amount is represented in the cognitive mapping system as both a grid-like representation and a 2D ‘positional’ code in the EC and mPFC. These neural representations are better explained by a subjective value space that is distorted based on individuals’ risk preferences than an objectively defined space, linking the cognitive map to subjective preference. Further, we find that the strength of grid-like representations is associated with the strength of subjective value representations, suggesting the cognitive mapping and value coding systems work in tandem. These findings bridge the cognitive mapping and value coding systems and provide one possible mechanism by which individuals convert multiple attributes into a unified subjective value signal, together supporting a new framework for understanding how the brain constructs and compares values.