Independence and dependence in human causal reasoning.

Area of Science:

• Cognitive Science
• Psychology
• Artificial Intelligence

Background:

• Causal graphical models (CGMs) are widely used for modeling human causal reasoning.
• The causal Markov condition is a key property of CGMs, defining independence relationships between variables.

Purpose of the Study:

• To investigate whether human causal inferences adhere to the principles of CGMs, specifically the causal Markov condition.
• To explore common errors in causal reasoning, such as associative reasoning and lack of discounting.

Main Methods:

• Five experiments were conducted to assess adult causal inferences.
• Participants' reasoning patterns were analyzed, including tendencies to violate the Markov condition and exhibit associative reasoning.

Main Results:

• Adults demonstrated a tendency to violate the causal Markov condition, exhibiting associative reasoning.
• Participants often failed to discount alternative causes and incorrectly inferred conditional independence.
• These associative tendencies persisted even under time pressure, suggesting deliberate thought, and varied significantly across individuals.

Conclusions:

• Human causal reasoning deviates from the strict independence assumptions of CGMs.
• Models of human causal reasoning must account for associative thinking, individual differences, and the influence of disabling factors and conjunctive strategies.
• Future theories need to relax CGM independence constraints for higher fidelity to human cognition.