Portal de Periódicos da CAPES

Evaluating Two Mechanisms of Flexible Induction: Selective Memory Retrieval and Evidence Explanation

2013; Wiley; Volume: 35; Issue: 35 Linguagem: Inglês

1551-6709

Nadya Vasilyeva, John D. Coley,

Child and Animal Learning Development

Evaluating Two Mechanisms of Flexible Induction: Selective Memory Retrieval and Evidence Explanation Nadya Y. Vasilyeva (Nvasil@Brandeis.Edu) Department of Psychology, Brandeis University 415 South Street, Waltham MA 02454 USA John D. Coley (j.coley@neu.edu) Department of Psychology, Northeastern University 360 Huntington Avenue, Boston MA 02115 USA on induction that use argument evaluation task, we employed inference generation task: participants were given an inductive premise and asked to generate their own conclusions. Coley & Vasilyeva (2010) demonstrated that this task provides a particularly sensitive measure of participants’ spontaneous use of different kinds of relevant knowledge, in the context of an ecologically valid inductive problem. Abstract We report three studies examining mechanism of property- sensitive induction. First, we demonstrate that, contrary to a common assumption, property does not influence retrieval of knowledge about premise categories. Second, we introduce property-driven explanations as a possible source of property effects and provide first evidence for this proposal. Keywords: induction; property effects; retrieval; explanation. Generating hypotheses about uncertain outcomes from limited evidence – inductive inference - is a pervasive cognitive activity. In order to be successful, inductive inference must be flexible. For example, if you learn that a new influenza virus has been discovered in chickens, you may reasonably get concerned about your own health; but if chickens were announced to carry a certain defective gene, you are much less likely to worry about catching one during your next meal. Indeed, a vast body of empirical evidence demonstrates that people make systematically different inferences when they project different properties (see Coley & Vasilyeva, 2010, for a review). Heit and Rubinstein (1994) proposed that property affects induction by indicating different subsets of features as relevant for evaluating premise-conclusion similarity. Goodman (1972) provided a logical argument for constrained recruitment of features: since any category has a potentially infinite set of features and can be infinitely similar to any other category, it is a necessary logical requirement for inductive inference to impose some initial constraints to limit a subset of relevant features. Although it is generally agreed that induction requires constrained recruitment of prior knowledge, and there is evidence that projected property may provide one such constraint (Coley & Vasilyeva, 2010; Heit & Rubinstein, 1994), the mechanism of property-based constrained recruitment remains unclear. Existing models of induction either do not specify the psychological mechanism of property effects (McDonald, Samuels, & Rispoli, 1996; Medin, Coley, Storms, & Hayes, 2003; Rips, 1975; Osherson, Smith, Wilkie, Lopez, & Shafir, 1990; Sloman, 1993; Sloutsky & Fisher, 2004), or acknowledge the computational nature of their account that may not correspond to actual psychological processes involved in inductive inference (e.g. Tenenbaum, Kemp, & Shafto, 2007; Heit, 1998). We report three studies that examine two candidate psychological mechanisms of property effects in induction: property-moderated retrieval of relevant knowledge about premise categories from long-term memory, and generating explanations of why premise categories might have the property to begin with. In contrast to the majority of studies Property-Moderated Knowledge Retrieval as a Mechanism of Property Effects Generation of an inductive hypothesis is inherently knowledge driven; when one learns that A has a novel property X, one uses what they know about A and its relations to other things to form guesses about what else is likely to have X. One source of input to inductive inference is knowledge about premise categories. When such categorical knowledge is accessed, a probabilistically determined subset of features and relations that comprise the representation of that concept becomes available as a raw material for the inference. For example, if A turned out to be a duck, such features as “is a bird”, “flies”, “lives in ponds”, “quacks” and “eaten by foxes” may come to mind. Although there are many different types of knowledge, knowledge about living things is commonly divided into two broad classes: taxonomic knowledge is based on relations of intrinsic similarity between members, whereas contextual, or ecological knowledge, is based on extrinsic relations between members and other entities. For example, ducks belong to the taxonomic category of birds and ecological categories of aquatic animals and fox prey. Each of these types of knowledge can serve as a basis for an inductive projection from ducks – to other birds, or other aquatic animals, or things that eat ducks. In addition to the premise category, knowledge about the property can also serve as a source of information. If X in the example above is replaced with a more specific property, such as “carries a certain disease” or “has a certain gene”, new knowledge is brought to the table: independently of what we know about ducks, we also know something about diseases and genes: what they are, whether they can be transmitted via contact, etc. How can property influence what projections people end up making? One possibility is that property constrains what types of premise knowledge are used to produce an inductive hypothesis. A premise category label, as any word, is connected to a vast amount of conceptual knowledge; this knowledge is unlikely to be retrieved in its entirety on any given occasion (McElree, Murphy & Ochoa, 2006). Rather, retrieval of conceptual information from long term memory is selective and depends on context (e.g. Barsalou, 1982; Swinney,