Computational protocol for hierarchical Bayesian modeling of perception and generalization in fear conditioning

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Psychology

Background:

• Understanding human generalization is key to distinguishing cognitive and perceptual mechanisms.
• Individual differences in fear generalization are not fully understood.
• Existing models may not adequately separate perceptual and learning contributions.

Purpose of the Study:

• To present a computational protocol for analyzing individual differences in fear generalization.
• To integrate Bayesian state-space and hierarchical mixture models for this analysis.
• To differentiate between perceptual and learning-based generalization processes.

Main Methods:

• Applying a Bayesian state-space model to perceptual data.
• Calculating stimulus distance from probabilistic perceptual representations.
• Employing a hierarchical mixture generalization model to parse generalization components.

Main Results:

• The protocol enables quantitative analysis of individual differences in fear generalization.
• It provides a framework for disentangling perceptual and learning contributions.
• Stimulus distance can be derived from probabilistic perceptual representations.

Conclusions:

• The integrated computational protocol offers a novel approach to studying fear generalization.
• This method facilitates a deeper understanding of the cognitive and perceptual underpinnings of generalization.
• It allows for the distinction between perceptual and learning-based generalization mechanisms.