A Process Model of Causal Reasoning.

Area of Science:

• Cognitive Science
• Psychology
• Computational Modeling

Background:

• Humans excel at causal reasoning across various tasks.
• The precise mental processes underlying causal judgments remain largely unknown.
• Existing models do not fully capture human error patterns in causal inference.

Purpose of the Study:

• To propose and validate a new process model for causal judgments: the mutation sampler.
• To investigate the computational mechanisms behind human causal reasoning.
• To explain predictable errors in causal judgments.

Main Methods:

• Developed the mutation sampler model using the Metropolis-Hastings sampling algorithm.
• Tested the model across diverse tasks with over 1,700 participants.
• Employed a novel experimental methodology to examine representative samples.

Main Results:

• The mutation sampler model provided a closer fit to human judgments than standard causal graphical models.
• Biases from mutation sampling explained predictable human errors not accounted for by normative models.
• These biases were evident in samples participants deemed representative.

Conclusions:

• Sampling methods, like the mutation sampler, are plausible process-level accounts of causal inference.
• The findings highlight the importance of computational sampling in understanding human causal reasoning.
• Future research should focus on identifying the 'how' of causal computation in the human mind.