Accounting for the Relational Shift and Context Sensitivity in the Development of Generalization

Similarity-based generalization is fundamental to human cognition, and the ability to draw analogies based on relational similarities between superficially different domains is crucial for reasoning and inference. Learning to base generalization on shared relations rather than (or in the face of) shared perceptual features has been identified as an important developmental milestone. However, recent research has shown that children and adults can flexibly generalize based on perceptual or relational similarity, depending on what has been an effective strategy in the past in a given context. Here we demonstrate that this pattern of behavior naturally emerges over the course of development in a domain-general statistical learning model that employs distributed, sub-symbolic representations. We suggest that this model offers a parsimonious account of the development of context-sensitive, similarity-based generalization and may provide several key advantages over other popular structured or symbolic approaches to modeling analogical inference.