Causal inference during closed-loop navigation: parsing of self- and object-motion

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Perception

Background:

• Bayesian Causal Inference (CI) is crucial for building internal models of the world by attributing sensory signals to their causes.
• While CI is understood in simple tasks, its role in complex action-perception loops, like navigation, is less clear.

Approach:

• Developed a normative model for intercepting moving targets during closed-loop navigation, considering beliefs about target and self-motion.
• Conducted experiments examining human perception and steering behavior when navigating and attempting to intercept targets.
• Analyzed eye-movement data to understand gaze dynamics during target pursuit under different self-motion conditions.

Key Points:

• Humans misattribute object motion to self-motion, perceiving targets as stationary and steering toward initial positions when self-moving.
• Misattribution of retinal motion increases during passive vs. active self-motion and with eccentric target presentation.
• Eye-tracking reveals accurate initial saccades but altered gaze pursuit based on motion conditions.

Conclusions:

• Demonstrates Bayesian Causal Inference (CI) within naturalistic action-perception loops during navigation.
• Suggests that CI computations unfold over extended time periods during dynamic interactions.
• Highlights the influence of self-motion on attributing causality to visual motion signals.