Learning absence of relation is challenging

An analysis from the perspective of inductive bias

Abstract

People learn relations between two variables (e.g., height and weight or date and temperature) from everyday experiences or information transfer. However, people have a variety of cognitive biases and this is true for learning of relations. Kalish et al. (2007) showed that when learning relation between two variables, people have a strong inductive bias such as implicit assumption of a positive linear relation. This finding indicates that differences exist in difficulty of learning depending on the relation between two variables.In the present study, we conducted the knowledge transmission task used in Kalish et al. (2007) to analyze theoretically and experimentally the effect of inductive bias on learning absence of relation　between two variables. We found that participants did not learn the absence of relation between two variables well because of the inductive bias. In particular, participants people falsely found positive linear relation between two variables wherein no relation actually existed and this false cognition facilitate learning of relation between two variables.