Using priors to formalize theory: optimal attention and the generalized context model

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Psychological Modeling

Background:

• Formal models in psychology enhance theoretical precision and quantitative evaluation.
• Bayesian statistical approaches offer a powerful method for integrating theoretical assumptions into models.
• Informative prior distributions are key to encoding psychological theories within Bayesian models.

Purpose of the Study:

• To demonstrate the utility of informative prior distributions in formal psychological modeling.
• To enhance the psychological content and reduce the complexity of existing models.
• To validate the use of Bayesian model selection for testing formalized theoretical assumptions.

Main Methods:

• Utilizing Bayesian statistical approaches to formalize psychological theories.
• Implementing informative prior distributions to represent theoretical assumptions about psychological variables.
• Applying Bayesian model selection to evaluate theoretical assumptions encoded in priors using the generalized context model (GCM) of category learning.

Main Results:

• A generalized context model (GCM) with an informative prior demonstrated higher psychological content and lower complexity.
• Formalizing theory in informative priors allows for reliable Bayesian model selection, independent of prior sensitivity.
• Bayesian model selection successfully tested theoretical assumptions about optimal attention allocation.

Conclusions:

• Integrating psychological theory into informative prior distributions is a widely applicable and recommended practice in psychological modeling.
• This approach enhances the psychological interpretability and rigor of formal models.
• The Bayesian framework provides a robust method for both building and testing psychological theories.