Adaptive non-interventional heuristics for covariation detection in causal induction: model comparison and rational...

Area of Science:

• Cognitive Psychology
• Decision Science

Background:

• Covariation detection is crucial for learning and decision-making.
• Existing models often struggle with rare events and data selection biases.

Purpose of the Study:

• To evaluate 41 existing models of covariation detection.
• To introduce and validate a new model incorporating extreme rarity assumptions.
• To assess the adaptive rationality of the new model.

Main Methods:

• Evaluation of 41 models against literature and experimental data.
• Introduction of a novel model based on the phi-coefficient under extreme rarity.
• Rational analysis using two computer simulations.

Main Results:

• The new model demonstrated superior performance compared to existing models.
• Simulations identified optimal environmental conditions and memory constraints for the new model.
• The new model approximates the normative model effectively under specific task conditions.

Conclusions:

• The proposed model offers a more accurate account of human covariation detection, especially with rare events.
• The findings highlight the importance of rarity and adaptive rationality in cognitive models.
• The study provides a framework for understanding decision-making under uncertainty.