Privileged (Default) Causal Cognition: A Mathematical Analysis.

Area of Science:

• Cognitive Science
• Artificial Intelligence
• Psychology

Background:

• Causal cognition is fundamental to human learning, reasoning, and decision-making.
• Existing research focuses on learning causal structures but lacks normative guidance on functional forms.
• Empirical studies show support for multiple functional forms in causal inference.

Purpose of the Study:

• To propose a privileged (default) functional form for causal relations.
• To theoretically unify existing causal models.
• To explain diverse empirical findings in causal inference.

Main Methods:

• Analysis of conceptual and mathematical constraints on causal functional forms.
• Theoretical unification of noisy-OR/AND and linear causal models.
• Demonstration of diverse functional forms as special cases of a general model.

Main Results:

• A combination of constraints identifies a privileged functional form for causal relations.
• This privileged function unifies noisy-OR/AND and linear models, showing their complementarity.
• The proposed general form explains empirical support for various specific functional forms.

Conclusions:

• The identified privileged functional form offers a unified perspective on causal inference.
• This unification reconciles competing empirical findings regarding causal functional forms.
• The research provides a normative framework for selecting functional forms in causal learning and reasoning.