Sensorimotor priors in nonstationary environments.

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Modeling

Background:

• The human nervous system estimates environmental variables using probabilistic prior information.
• Previous research focused on stable environments, with limited understanding of nonstationary statistics.
• The development of priors in environments with changing statistical properties remains largely unexplored.

Purpose of the Study:

• To investigate if humans alter their reliance on prior information based on recent environmental variance changes.
• To compare human sensorimotor prediction adaptation with computational models in dynamic environments.

Main Methods:

• Experimental estimation of human sensorimotor prior (prediction) in real-time.
• Comparison of human predictions against adaptive and nonadaptive computational models.
• Simulations to evaluate model performance in predicting stimuli under nonstationary conditions.

Main Results:

• Adaptive models demonstrated superior prediction accuracy in environments with nonstationary statistics compared to nonadaptive models.
• Adaptive models generally provided the best predictions of human participants' responses.
• No evidence was found for increased reliance on recent experience immediately following systematic environmental changes.

Conclusions:

• Human sensorimotor prediction adapts to environmental changes, aligning with adaptive computational models.
• Despite adaptive prediction, humans do not show a heightened immediate reliance on recent data post-environmental shifts.
• Further research is needed to understand the precise mechanisms of prior adaptation in dynamic environments.