Category-based induction (CBI) is the generalization of features from one category to another. For example, given that robins have a particular enzyme one would be more likely to think that a blue jay has the same enzyme than to think that an ostrich does. While CBI is a fundamental ability, current theories do not provide a uniform account of this important cognitive function. Some theories implicate featural similarity (number of shared attributes) between base and target categories as the central explanatory mechanism. However, recent studies with expert populations and laboratory experiments using familiar stimuli are beginning to suggest that in addition to similarity, causal reasoning (explanations for the emergence of a property) may play a part in guiding inference. There is a great need for a theory that is able to describe the circumstances under which people use similarity-based strategies and when they use causal strategies. One factor that might determine what strategy is used is causal knowledge availability - the idea that causal information will be used in inference if it is available. In cases where knowledge is restricted to featural information, similarity based strategies will be used. An alternative position, the causal primacy hypothesis, is the suggestion that inductive reasoning is just causal reasoning, and in certain cases the process of causal reasoning happens to demonstrate the same outcomes as similarity-based reasoning. These questions concern the way in which categories involved in reasoning are represented and how the structure of these representations interacts with inductive reasoning about these categories. Traditionally, researchers were not able to study this interaction because almost all of the research on induction had used existing categories as stimuli. While using rich well-established natural categories often has benefits over artificial categories (e.g., greater ecological validity), teaching categories in the laboratory may allow for a more control over category representations. Extending Yamauchi and Markman' s (1998, 2000) paradigm, my studies will involve two stages. The first is a standard category-learning task in which participants learn to classify exemplars into categories. The second stage of the experiment is the inference task, in which participants are told about a new feature of a familiar exemplar and are asked to generalize this feature to other exemplars in the same and different categories. This design will allow the manipulation of category structure of learned categories (e.g., prototype vs. causal model) and the types of inferences participants are drawing (e.g., projecting familiar or novel features).
| Blok, Sergey V; Medin, Douglas L; Osherson, Daniel N (2007) Induction as conditional probability judgment. Mem Cognit 35:1353-64 |
