Adaptive learning via selectionism and Bayesianism, Part I: connection between the two.

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Behavioral Economics

Background:

• Operant learning, or the Law of Effect, posits that action consequences shape future behavior probabilities.
• Bayesianism updates action probabilities based on evidence using Bayes' formula, distributing support across hypotheses.
• Selectionist and Bayesian frameworks differ in how they attribute evidence during hypothesis testing.

Purpose of the Study:

• To reveal an intimate connection between selectionist operant learning and Bayesian dynamics.
• To demonstrate how individual learning rules can lead to population-level Bayesian-like evolution.
• To bridge theoretical frameworks in learning and decision-making.

Main Methods:

• Analyzing the selection-by-consequence characterization of operant learning.
• Applying Bayesian principles of evidence evaluation and probability updating.
• Utilizing the linear operator model's learning equation under ensemble averaging.

Main Results:

• Proving that individual Law of Effect learning generates ensemble-level Bayesian-like dynamics.
• Demonstrating that the linear operator model's equation yields predictive reinforcement learning models.
• Establishing a formal link between selectionist and Bayesian learning theories.

Conclusions:

• Individual action selection based on consequences drives population-level Bayesian dynamics.
• This finding unifies selectionist and Bayesian perspectives on learning and decision-making.
• The study provides a theoretical foundation for understanding reinforcement learning in biological and artificial systems.