Abstract

The role of comparison in structure learning: Developmental, learning science, and computational perspectives Stella Christie ([email protected]) & Dedre Gentner ([email protected]) University of British Columbia, Dept. Psychology; Northwestern University, Dept. of Psychology Mutsumi Imai ([email protected]), Etsuko Haryu ([email protected]) & Hiroyuki Okada ([email protected]) Keio University, Dept. Psychology; University of Tokyo, Dept. of Education; Tamagawa University, Brain Science Institute Ji Y. Son ([email protected]) & James W. Stigler ([email protected]) California State University, Los Angeles, Dept. Psychology; UCLA, Dept. of Psychology Leonidas A.A. Doumas ([email protected]), Robert G. Morrison ([email protected]), & Lindsey E. Richland ([email protected]) University of Hawaii, Dept. Psychology; Loyola University, Dept. of Psychology; University of Chicago, Dept. of Human Development than when they were given the less systematic set of terms on, in, under. To discover the generality of these effects, in this series of studies we asked whether children given nonspatial (but systematic) language would still show an advantage in the spatial mapping task. We presented children with a spatial mapping task as in Loewenstein & Gentner (2005). There were three groups: one heard a systematic set of spatial terms (top/middle/bottom); one heard a systematic set of nonspatial terms (one/two/three); and a third heard a nonsystematic set of nonspatial terms (dog/pig/cat). In addition to the standard three-tiered mapping task, we also conducted a vertical-to-horizontal mapping task. The results suggest that (1) children benefit from systematic language; (2) domain-specificity benefits early learning; and (3) at older ages, abstract language can have a larger advantage in a difficult transfer task. Keywords: relational thinking, analogy, comparison, cognitive development, computational modeling The ability to perceive, comprehend and reason about relations (i.e., relational thinking) is central in human cognition. Relational thinking is powerful because it is structured. Specifically, relational thought allows inferences and generalizations that are constrained by the roles that elements play, rather than strictly the properties of the elements themselves. The role of relational comparisons in learning is emerging as an important area of developmental and learning science research. Relational comparisons allow learners to derive symbolic, abstract, and conceptual knowledge representations that are generative, in that children and adults can then use them broadly in new contexts to reason about new elements. Indeed, comparison seems to underlie the very development of the structured relational representations that underlie relational cognition. This symposium aims to bring to together research on the role of comparison in developmental and adult learning. Specifically, we present research on the role of comparison in the development of spatial reasoning, language learning, adult mathematics learning, and computational approaches to learning structured (i.e., symbolic) representations. Imai, Haryu, & Okada: Progressive alignment in verb learning Verbs should be extended by the sameness of action, whereas nouns should be extended attending to similarity of objects. Children under four years of age easily generalize a novel noun to other objects of like kinds, whereas even 4-year-olds tend to fail extending a novel verb to the same action performed by a different agent or with a different object (Imai et al., 2005, 2008). Children fail to segregate the action from the objects constituting it. In other words, children fail to structurally align action events. Previous research suggests that object similarity between objects in corresponding relational roles can promote structural alignment and help children notice higher-order relational similarity (e.g., Gentner & Toupin, 1986). Borrowing this idea, two experiments examined whether young children’s verb generalization would be fostered by similarities between corresponding objects in the two events. In the first experiment 4 year-old children were shown a video in which a woman was doing a novel action with a novel object, and heard a novel verb. Children were then asked to extend the verb to either a Christie & Gentner: Domain Specific vs Abstract Language in Spatial Learning Many studies have suggested that language provides important tools for learning and thinking in cognitive development. In this work we test one specific claim concerning the cognitive effects of language learning: namely, that systematic semantic structure in language can invite correspondingly systematic conceptual structure (Gentner, 2010; Gentner & Christie, 2011). Evidence for this claim comes from prior studies by Loewenstein and Gentner (2005) in which children performed better on a difficult spatial mapping task involving three-tiered structures when they were given the monotonic set of spatial terms top, middle, bottom