CRCNS: Computational Mechanisms for State-Driven Active Sensing

PROJECT SUMMARY (See instructions): Humans and other animals exhibit a critical behavioral phenomenon in the control of movement: they switch between two distinct modes-fast, exploratory "active sensing" and slower, goal-directed "task control." This proposal aims to uncover the computational, behavioral, and neural mechanisms that regulate these two modes. We hypothesize that animals use internal estimates of sensory uncertainty to decide when to switch between modes, with uncertainty thresholds triggering transitions between exploratory and goal-driven behavior. To test this, we will use a uniquely tractable animal model-weakly electric fish performing a refuge tracking task-in which behavior, sensory feedback, and neurophysiology can be measured and manipulated in real time. The project comprises three specific aims: (1) develop computational models and theoretical tools to identify how and why animals switch between active sensing and task control modes; (2) conduct high-throughput behavioral experiments to quantify how sensory salience and feedback influence mode switching and control strategies; and (3) perform neurophysiological recordings to identify neural correlates of locomotor control policies and mode switching in the brain. This multidisciplinary effort integrates control theory, machine learning, and neuroscience to reveal the computational strategies underlying the regulation of active sensing and task control. Our findings have the potential to transform our understanding of biological motor control, identify mechanisms for assessing sensory uncertainty, and advance strategies for movement control in complex environments.