Active inference under visuo-proprioceptive conflict: Simulation and empirical results

Area of Science:

• Neuroscience
• Cognitive Science
• Robotics

Background:

• The brain uses internal models integrating visual and proprioceptive information for hand movements.
• Active inference models suggest sensory precision (reliability) dictates modality influence on action and perception.
• Expected precision is linked to attentional allocation in sensory processing.

Purpose of the Study:

• To investigate if increasing attention to vision or proprioception improves performance in a hand-target matching task with conflicting sensory cues.
• To explore the role of attention in balancing conflicting visual and proprioceptive inputs for action control.

Main Methods:

• A simulated agent based on predictive coding and active inference principles was used.
• Human participants performed a hand-target phase matching task in a virtual reality environment.
• Task performance and self-reported attentional allocation were measured under visuo-proprioceptive conflict.

Main Results:

• Increasing expected precision of vision or proprioception in the simulated agent improved task performance under sensory conflict.
• The simulation successfully replicated human participants' behavior and attentional allocation.
• Selective attention was shown to effectively balance conflicting visual and proprioceptive cues.

Conclusions:

• Attention serves as a crucial mechanism for flexible body representation and action control.
• Selective attention dynamically adjusts the weighting of sensory information based on expected precision.
• Findings support the role of attention in resolving visuo-proprioceptive conflicts for effective motor control.