A Magic Act in Causal Reasoning: Making Markov Violations Disappear

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Causal Inference

Background:

• Causal reasoning theories should explain both the occurrence and timing of errors.
• The mutation sampler is a rational process model predicting errors under limited cognitive resources.
• Markov violations occur when reasoners incorrectly assume statistical dependence between causally independent variables.

Purpose of the Study:

• To investigate conditions under which Markov violations disappear.
• To test the mutation sampler's predictions regarding the elimination of causal reasoning errors.
• To explore the impact of inhibitory versus generative causal relations on reasoning.

Main Methods:

• Utilized a novel causal structure with generative and inhibitory relations.
• Presented subjects with reasoning tasks involving these structures.
• Employed theoretical model fitting to validate predictions.

Main Results:

• Reasoning with purely generative causal relations produced standard positive Markov violations.
• Introducing even a single inhibitory causal relation eliminated these Markov violations.
• Model fitting confirmed the mutation sampler's ability to predict this elimination.

Conclusions:

• The type of causal relation (generative vs. inhibitory) significantly impacts the presence of Markov violations.
• The mutation sampler accurately predicts the disappearance of Markov violations under specific conditions.
• This finding offers new insights into the mechanisms of human causal reasoning errors.