When do we think that X caused Y?

Area of Science:

• Cognitive Science
• Psychology
• Artificial Intelligence

Background:

• Human causal judgment often ignores background factors (e.g., oxygen in forest fires).
• Existing models struggle to fully explain nuanced causal selection in humans.

Purpose of the Study:

• To develop a computational model of causal judgment based on "portable" causes.
• To formalize the idea that causes generalize across different circumstances.

Main Methods:

• Developed a computational model based on counterfactual correlation.
• Tested the model against a fine-grained dataset of human causal judgments.

Main Results:

• The model successfully predicts human causal judgments.
• It explains the influence of cause normality on judgment.
• Outperforms existing computational models on human graded causal judgments.

Conclusions:

• Causal judgment prioritizes factors that are robust across counterfactual scenarios.
• The model provides a parsimonious explanation for human causal selection.
• This work advances computational approaches to understanding causality.