Effects of Explanation and Comparison on Category Learning

Effects of Explanation and Comparison on Category Learning Brian J. Edwards (Brian.Edwards@U.Northwestern.Edu) Department of Psychology, Northwestern University, 2029 Sheridan Road-102 Swift Hall Evanston, IL 60208 USA Joseph J. Williams (Joseph_Williams@Berkeley.Edu), Tania Lombrozo (Lombrozo@Berkeley.edu) Department of Psychology, University of California, Berkeley, 3210 Tolman Hall Berkeley, CA 94720 USA Abstract increase in metacognitive awareness (Chi, 2010) and an increase in attention and engagement (e.g., Siegler, 2002), among others. In the context of category learning, generating explanations also enables learners to generalize beyond a specific set of observed data. In particular, Williams and Lombrozo (2010, 2013) proposed a subsumptive constraints account of how explanation impacts learning, whereby explaining leads people to interpret individual cases as part of a general pattern. As a result, explanation can help people unify multiple observations and focus on patterns with broader scope, increasing the discovery of rules that account for 100% of the data versus only 75% (Williams & Lombrozo, 2010). One mechanism by which comparison has been hypothesized to support learning is by promoting explicit structural alignment, leading people to focus on alignable differences between two entities (i.e., differences that are embedded in a common relational structure) (Gentner, 1983; Gentner & Markman, 1997). Since comparison causes people to analyze these differences in the context of the common structure, comparison can illuminate deeper similarities and support the formation of an abstract relational schema, even (and especially) when the items being compared have surface differences (Gentner et al., 2009). For example, the analogy “an atom is like a solar system” highlights the fact that an atom consists of electrons orbiting around a nucleus, whereas a solar system consists of planets orbiting around the sun. Across a number of domains, comparing two examples that are superficially dissimilar but share a common relational structure supports transfer more effectively than studying the same examples separately (e.g., Kurtz, Miao, & Gentner, 2001; Loewenstein, Thompson, & Gentner, 2003). Despite the abundance of research showing that explanation and comparison can (individually) enhance learning, few studies have investigated the effects of both explanation and comparison on the same experimental task. Kurtz, Miao, and Gentner (2001) found that comparing two analogous examples of heat flow helped participants discover similarities between the two examples more effectively than describing and explaining the same examples sequentially. Additionally, comparison was most effective when participants performed a task that involved listing which elements of the second scenario corresponded to specific elements of the first scenario. In another study, Nokes-Malach et al. (2012) found that introductory physics Generating explanations and making comparisons have both been shown to improve learning. While each process has been studied individually, the relationship between explanation and comparison is not well understood. Three experiments evaluated the effectiveness of explanation and comparison prompts in learning novel categories. In Experiment 1, participants explained items’ category membership, performed pairwise comparisons between items (listed similarities and differences), did both, or did a control task. The explanation task increased the discovery of rules underlying category membership; however, the comparison task decreased rule discovery. Experiments 2 and 3 showed that (1) comparing all four category exemplars was more effective than either within-category or between-category pairwise comparisons, and that (2) “explain” participants reported higher levels of both spontaneous explanation and comparison than “compare” participants. This work provides insights into when explanation and comparison are most effective, and how these processes can work together to maximize learning. Keywords: learning. Explanation; comparison; categorization; Introduction Explanation (i.e., answering “why” questions) and comparison (i.e., describing the similarities and differences between entities) are both powerful learning processes. Although they have typically been studied independently, they are often interconnected. Asking people to generate explanations can invite implicit comparison, and the patterns that people discover by comparing can motivate a search for explanations. For example, explaining why someone prefers coffee versus tea might lead one to identify similarities and differences between the two beverages, and comparing coffee and tea might provide insights into why a person would prefer one over the other. Explanation and comparison can also support similar ends: both promote abstraction and generalization, and both facilitate the discovery of patterns that are deep in a system’s underlying structure (for reviews, see Gentner, 2010, on analogy and comparison; Lombrozo, 2012, on explanation). Although explanation and comparison can generate similar effects, these two processes might rely on different cognitive mechanisms and exert different constraints on learning. Explanation has been hypothesized to improve learning through a variety of mechanisms, including an