ACTIVE INFERENCE IN MOTION PERCEPTION: INVESTIGATION PARADIGMS UNDER NON-INVASIVE BRAIN STIMULATION

Recognizing biological motion from Point Light Displays (PLDs) may rely on the brain’s ability to minimize prediction errors under uncertainty. Although transcranial direct current stimulation (tDCS) can modulate social perception, its computational mechanisms remain unclear. Using an Active Inference framework, we examined whether tDCS affects perceptual learning rates or decision precision.In addition, we are extending this paradigm by using PLD stimuli specifically designed to display non-biological motion, with the aim of testing whether Point Light Displays can provide an acceptable and reliable representation of non-biological movement, analogous to what has been demonstrated for biological motion.
Eleven healthy participants performed a Signal Detection Task involving PLD stimuli (Biological/Non-Biological, Goal/Non-Goal oriented) with varying levels of noise. Participants underwent four randomized stimulation conditions: Baseline, Sham, Anodal, and Cathodal tDCS, with stimulation targeting the ventral premotor cortex. Behavioral responses were analyzed trial by trial using a two-level Bayesian model. The perceptual component, inspired by the Hierarchical Gaussian Filter, described belief updating as a function of precision-weighted sensory uncertainty, whereas the response component mapped beliefs onto behavior through two parameters: Decision Threshold (τ), indexing response bias, and Sensitivity (ζ), indexing decision precision.
The model indicated that stimulation did not substantially modify perceptual belief trajectories or response bias, with τ remaining close to 0.5 across conditions. Compared to Sham, Anodal stimulation increased
Decision Sensitivity (ζ), suggesting a higher signal-to-noise ratio in the transformation of beliefs into actions.