Extended Bayesian inference incorporating symmetry bias.

Area of Science:

• Cognitive Science
• Computational Psychology
• Artificial Intelligence

Background:

• Causal induction is fundamental to human reasoning.
• Existing models often fail to capture the nuances of human judgment, particularly symmetry bias.
• The need for models that can adapt to changing inferential targets is critical.

Purpose of the Study:

• To propose a novel causal induction model incorporating symmetry bias.
• To develop a human-like Bayesian inference method for adaptive learning.
• To enhance the accuracy and adaptability of computational models of human reasoning.

Main Methods:

• Developed a causal induction model with parameters controlling symmetry bias.
• Validated the model against human judgments in eight causal induction experiments.
• Proposed a Bayesian inference method integrating causal induction for simultaneous inference and learning.

Main Results:

• The proposed model achieved a high mean coefficient of determination (0.93) in reproducing human causal induction.
• Demonstrated the model's ability to accurately replicate human judgment.
• Showcased the effectiveness of the human-like Bayesian inference method in unsteady situations.

Conclusions:

• The novel causal induction model with symmetry bias accurately reflects human judgment.
• The proposed human-like Bayesian inference method facilitates simultaneous inference and adaptive learning.
• This approach offers a more robust framework for understanding and modeling human reasoning in dynamic environments.