Neural Correlates of Causal Confounding.

Area of Science:

• Cognitive Neuroscience
• Causal Inference
• Computational Psychiatry

Background:

• Causal confounding, where multiple causes covary, poses a challenge to discerning independent influences.
• Behavioral studies indicate that humans, including children, possess an intuitive sensitivity to causal confounding.

Purpose of the Study:

• To investigate the neural substrates underlying human sensitivity to causal confounding.
• To differentiate neural mechanisms supporting complex causal inference from simpler learning processes.

Main Methods:

• Functional magnetic resonance imaging (fMRI) combined with computational cognitive modeling.
• Participants judged the influences of confounded and nonconfounded, deterministic and stochastic causes.
• A Bayesian causal model and an error-driven algorithm were used to account for neural activity.

Main Results:

• Neural activity in the dorsomedial prefrontal cortex (dmPFC) during causal judgments was better explained by a Bayesian model.
• The Bayesian model, sensitive to both confounding and stochasticity, outperformed an error-driven algorithm that only accounted for stochasticity.
• This suggests dmPFC activity supports sophisticated causal reasoning that considers confounding.

Conclusions:

• The dorsomedial prefrontal cortex (dmPFC) plays a crucial role in mediating sensitivity to causal confounding.
• Findings support the role of domain-general Bayesian causal inference mechanisms in human cognition.
• Implications for understanding uncertainty estimation and causal induction are discussed.