Information-theoretic Sensorimotor Foundations of Fitts' Law
We propose a novel, biologically plausible cost/fitness function for sensorimotor control, formalized with the information-theoretic principle of empowerment, a task-independent universal utility. Empowerment captures uncertainty in the perception-action loop of different nature (e.g. noise, delays, etc.) in a single quantity. We present the formalism in a Fitts' law type goal-directed arm movement task and suggest that empowerment is one potential underlying determinant of movement trajectory planning in the presence of signal-dependent sensorimotor noise. Simulation results demonstrate the temporal relation of empowerment and various plausible control strategies for this specific task.