Causal competition based on generic priors.

Area of Science:

• Cognitive Science
• Psychology
• Machine Learning

Background:

• Human learning often involves complex, multi-causal environments.
• Cognitive limitations can hinder the acquisition of accurate causal models.
• A preference for simple, efficient causal representations is hypothesized.

Purpose of the Study:

• To investigate cue competition in human causal learning.
• To determine if independently occurring causes compete.
• To test predictions of sparse and strong causal models.

Main Methods:

• Three experiments were conducted.
• Participants made judgments of causal strength.
• Independent generative and preventive causes were manipulated.

Main Results:

• Cue competition was observed for both generative and preventive causes.
• The strength of a cause was underestimated when a stronger, uncorrelated cause was present.
• Findings support the existence of competition between independent causes.

Conclusions:

• Human causal learning exhibits cue competition, even for independent causes.
• Results align with theories favoring sparse and strong causal models.
• The study provides data for computational models of complex causal inference.