Agents and causes: dispositional intuitions as a guide to causal structure

Area of Science:

• Cognitive Science
• Philosophy of Science
• Artificial Intelligence

Background:

• Two competing frameworks exist for causal reasoning: dependency theories and dispositional theories.
• Dependency theories focus on cause-effect relationships, while dispositional theories emphasize agent-patient interactions and intrinsic properties.
• A key finding bridging these frameworks is the differential attribution of blame to agents and patients when causal failures occur.

Purpose of the Study:

• To develop and test a computational model that explains differential error attribution in causal reasoning.
• To augment a causal Bayes net model with separate error sources for causes and effects.
• To investigate how the location of agents and the causal structure influence error attribution.

Main Methods:

• Augmented a causal Bayes net model with distinct error sources for causes and effects.
• Conducted experiments to test the model's predictions.
• Used the magnitude of Markov violations as an empirical measure of differential error attribution assumptions.

Main Results:

• The augmented causal Bayes net model successfully predicted differential error attribution.
• The size of Markov violations was influenced by the location of agents.
• The influence of agent location on error attribution was moderated by the causal structure and variable types.

Conclusions:

• The proposed model provides a unified framework for understanding error attribution in causal reasoning.
• Agent location and causal structure are critical factors in how errors are attributed.
• This research bridges dependency and dispositional theories of causation through a computational approach.