A neural mechanism for conserved value computations integrating information and rewards

Abstract Behavioral and economic theory dictates that we decide between options based on their values. However, humans and animals eagerly seek information about uncertain future rewards, even when this information does not provide any objective value. This implies that decisions can be made by endowing information with subjective value and integrating it with the value of extrinsic rewards, but the mechanism is unknown. Using a novel multi-attribute decision making task we found that human and monkey value judgements are regulated by strikingly conserved computational principles, including how they compute the value of information and scale it with information’s timing and ability to resolve a specific form of uncertainty. We then identified a neural substrate in a highly conserved and ancient structure, the lateral habenula (LHb). LHb neurons signal the subjective value of choice options integrating the value of information with extrinsic rewards, and LHb activity both predicts and causally influences online decisions. Key input regions to LHb provide the necessary ingredients for these computations, but do not themselves signal an integrated value signal to guide multi attribute decisions. Our data thus identifies neural mechanisms of the conserved computations underlying multi-attribute, value-based decisions to seek information about the future.