Risk-Sensitivity in Bayesian Sensorimotor Integration

Information processing in the nervous system during sensorimotor tasks with inherent uncertainty has been shown to be consistent with Bayesian integration. Bayes optimal decision-makers are, however, risk-neutral in the sense that they weigh all possibilities based on prior expectation and sensory evidence when they choose the action with highest expected value. In contrast, risk-sensitive decision-makers are sensitive to model uncertainty and bias their decision-making processes when they do inference over unobserved variables. In particular, they allow deviations from their probabilistic model in cases where this model makes imprecise predictions. Here we test for risk-sensitivity in a sensorimotor integration task where subjects exhibit Bayesian information integration when they infer the position of a target from noisy sensory feedback. When introducing a cost associated with subjects' response, we found that subjects exhibited a characteristic bias towards low cost responses when their uncertainty was high. This result is in accordance with risk-sensitive decision-making processes that allow for deviations from Bayes optimal decision-making in the face of uncertainty. Our results suggest that both Bayesian integration and risk-sensitivity are important factors to understand sensorimotor integration in a quantitative fashion. Author Summary: Statistically optimal decision-makers use probabilistic predictive models of their environment to achieve their goals. However, in real life such probabilistic models can be wrong or only approximately true, in which case basing decisions exclusively on the statistics of such models can constitute a problematic decision criterion. In contrast, risk-sensitive decision-makers can take model uncertainty into account. They allow deviations from their probabilistic model depending on the quality of the predictions of the model. In particular, they trust their model less if it makes imprecise predictions and bias their decisions towards worst-case or best-case outcomes. Here we designed a sensorimotor task where subjects exhibit Bayesian information integration when they infer the hidden location of a target and they had to decide to make a more or less costly movement. We found that subjects exhibited a bias with respect to the statistically optimal movement towards less costly outcomes, the higher the uncertainty about the target location was. This interplay between estimation uncertainty and movement cost is consistent with a risk-sensitive decision criterion that takes model uncertainty into account.