Causal imprinting in causal structure learning.

Area of Science:

• Cognitive Science
• Psychology
• Causal Inference

Background:

• Humans often infer causation from observed correlations.
• Prior causal beliefs can be resistant to revision, even with contradictory evidence.

Purpose of the Study:

• To investigate the phenomenon of "causal imprinting," where initial causal beliefs are difficult to revise.
• To determine the conditions under which causal imprinting occurs and its normative status.

Main Methods:

• Four experiments were conducted using multiple dependent measures and control conditions.
• Participants observed correlations between events and were later presented with evidence that explained away the correlation.
• Bayesian analysis was used to assess the normativity of causal judgments.

Main Results:

• Participants persisted in believing an initial causal relationship (B causes C) even after learning that a third event (A) explained it away.
• Causal imprinting occurred even when revising the belief was clearly non-normative.
• The order of evidence presentation influenced causal belief revision, suggesting prior knowledge impacts interpretation.

Conclusions:

• Causal imprinting represents a significant difficulty in revising learned causal structures.
• Prior knowledge and the interpretation of later evidence play a crucial role in causal imprinting.
• Understanding causal imprinting is key to understanding human causal reasoning and belief updating.