Why Does Explaining “Why” Help Learning? A Subsumptive Constraints Account

• In Event: Different Perspectives on the Role of Explanation and Exploration in Learning

Abstract

Theoretical Perspective. The benefits of generating explanations or the self-explanation effect (Chi et al, 1994; Chi, 2000) are well established empirically, and are broadly consistent with many theoretical frameworks. However, this talk focuses on the challenge that there is much that is not understood about the precise mechanisms and processes that underlie the benefits of explanation. What are the conditions under which explaining will be most helpful, have no effect beyond typical learning strategies, or even be harmful?
The theoretical and empirical work in this talk is motivated by conceptual analyses from philosophy of science of what makes a good explanation (Salmon, 1990). In particular, the Subsumptive Constraints Account of explanation and learning (Williams & Lombrozo 2010) that we present proposes that when learners ask “Why?” a fact or observation is true, they seek to understand it as an instance of a general pattern or principle (Kitcher, 1989).

Objective. The talk presents the results of multiple experiments testing the Subsumptive Constraints Account’s prediction that explaining drives learners to discover and use underlying patterns and principles. The experiments examine a diversity of materials and learning contexts, which range from realistic statistical problems and concepts to carefully controlled artificial categories (see also Williams, Lombrozo & Rehder, 2013; and Williams et al, 2012). The experiments also examine whether there are conditions under which explaining can hurt learning (when there are exceptions to existing patterns).

Research Evidence. The talk reports evidence that the effect of explaining multiple observations differs depending on whether the observations conform closely to a reliable pattern, or the pattern is imperfect because some observations are exceptions to the pattern. We find that asking “why?” can impair learning (relative to alternative study strategies, like thinking aloud) by fostering overgeneralization, where learners repeatedly reason using a misleading pattern, at the expense of specific examples.
Additional experiments provide further insight into how to use explaining to help learning even when exceptions to patterns (or anomalies, Chinn & Brewer, 1993) are present. Explaining exceptions/anomalies can enhance learning and generalization when learners are asked to explain sufficiently many exceptions/anomalies, and when these anomalies are chosen to direct learners to reject imperfect patterns and recognize patterns that provide broader and more reliable generalizations.

Significance. The finding that explaining can hurt learning under certain conditions provides provides empirical support for the argument that understanding the processes that underlie explanation’s benefits is practically as well as theoretically important.
These findings support the Subsumptive Constraints Account’s prediction that explaining drives learners to seek underlying patterns, but underscore that the very feature of explaining “why?” that makes it powerful – discovering patterns that support generalization – can also produce negative effects on learning – when that feature leads to overgeneralization at the expense of specific exceptions.
Given that students receive a wide range of prompts to explain for a diversity of concepts and problems, these experiments and the conclusions they support may be helpful in providing principles to guide the practical use of explanation in classrooms, particularly for mathematical concepts.

Authors

• Joseph Jay Williams, Harvard University

• Tania Lombrozo, University of California - Berkeley