BUCKLE: A Model of Causal Learning

Abstract

Dealing with alternative causes is necessary to avoid making inaccurate causal inferences from covariation data. However, information about alternative causes is frequently unavailable, rendering them unobserved. Some theories of causal learning make simplifying assumptions to ease the difficulty associated with unobserved alternative causes. Here I present a new model of causal learning, BUCKLE (Bidirectional Unobserved Cause LEarning), which extends existing models of causal learning by dynamically inferring information about unobserved, alternative causes. During the course of causal learning, BUCKLE continually computes the likelihood that an unobserved cause is present during a given observation and then uses the results of these inferences to learn the causal strengths of the unobserved as well as observed causes. I will also present empirical evidence demonstrating that BUCKLE provides a better explanation of people’s causal learning than existing models.