Generating Inductive Inferences
Categorical inductive inference is the process by which we project features believed to be true of one class to another related class. Traditional approaches to studying inductive inference have focused on the evaluation of inductive arguments. In this chapter, we introduce a new approach by examining the way people generate inductive inferences. We focus on how relations among premise categories, and the nature of the property being projected, impact the kind of inferences generated. Participants were taught that two animal species shared a novel substance, disease, or gene, and were asked what other species might also have the property, and why. Results show that people attend to salient relations between premise categories, determine their relevance based on the property they are asked to project, and then generate inferences consistent with those relations. Participants drew a broad range of inferences based on taxonomic similarity, contextual similarity, and causal relations. Inference generation was constrained both by salient premise relations and the nature of the projected property. We discuss how these findings expand the list of challenges for the models of induction, question the primacy of taxonomic relations in guiding inductive inference, encourage further investigation into the process by which inductive inferences are generated, and emphasize the knowledge-driven and flexible nature of human inductive reasoning.