Rational causal induction from events in time

Area of Science:

• Cognitive Science
• Psychology
• Machine Learning

Background:

• Traditional causal learning research often ignores the continuous flow of real-world events.
• Understanding causal inference in naturalistic, continuous-time environments remains a challenge.

Purpose of the Study:

• To develop a rational framework for causal learning in continuous time.
• To model how temporal patterns influence causal inference.

Main Methods:

• Utilized Bayesian rational analysis and stochastic processes (Poisson-Gamma family).
• Derived computational principles for inferring causal structure from temporal data.

Main Results:

• The framework explains human preference for simpler, more reliable causal influences.
• Successfully reanalyzed seven experimental datasets, unifying diverse findings.

Conclusions:

• The proposed framework offers a unified explanation for continuous-time causal learning.
• It has implications for understanding various cognitive tasks, from explicit induction to implicit learning.