Boosting Retrieval Via Target Elaborations (the “Late Abstraction Principle”)

Abstract

A series of interventions aimed at eliciting an abstract encoding of learning contents succeeded in rendering them more retrievable during the processing of analogous situations. However, this approach is not applicable to learning contents that had not been originally encoded in ways that highlighted their abstract features. Drawing on computational accounts of the retrieval advantage of abstracting source analogs, the late abstraction principle posits that a comparable retrieval advantage should be obtained by abstracting target situations. We begin by reviewing the initial experimental evidence for the psychological reality of this computational insight: the fact that comparing two analogous situations increases access to distant analogs, as compared to their independent presentation. After discussing the limitations of the target-comparison strategy for being autonomously implemented by students, we review very recent interventions designed to help students capitalize on the late abstraction principle in truly autonomous ways. We end by describing how this finding was simulated within extant computational models of similarity-based retrieval, as well as by discussing theoretical objections to the claim that late abstraction can boost the retrieval of learning contents whose structural features had not been originally highlighted.