This preview shows pages 1–2. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: Environment and Goals Jointly Direct Category Acquisition Bradley C. Love University of Texas at Austin ABSTRACT Developing categorization schemes involves discovering structures in the world that support a learners goals. Existing models of category learning, such as ex- emplar and prototype models, neglect the role of goals in shaping conceptual organization. Here, a clustering ap- proach is discussed that reflects the joint influences of the environment and goals in directing category acquisition. Clusters are a flexible representational medium that ex- hibits properties of exemplar, prototype, and rule-based models. Clusters reflect the natural bundles of correlated features present in our environment. The clustering model Supervised and Unsupervised Stratified Incremental Adaptive Network (SUSTAIN) operates by assuming the world has a simple structure and adding complexity (i.e., clusters) when existing clusters fail to satisfy the learners goals and thus elicit surprise. Although simple, this oper- ation is sufficient to address findings from numerous lab- oratory and cross-cultural categorization studies. KEYWORDS categories; clusters; goals; learning; stereo- types Judging a person as a friend or foe, a mushroom as edible or poisonous, or a sound as an l or r are examples of categorization problems. Because people never encounter the same exact stimulus twice, they must develop categorization schemes that capture the useful regularities in their environment. One challenge for psychological research is to determine how humans acquire and represent categories. Different models simulating theories of category learning have been proposed, but they have not been sufficient to resolve the theoretical debates. For example, the relative merits of exemplar models (in which information about a category is stored as independent episodes or experiences) and prototype models (in which information about a category is stored in a summary format) are still debated, with exemplar models appearing best suited to certain data sets and prototype models for others (see Nosofsky and Zaki, 2002; Smith, 2002). The difficulty in resolving the debate may indicate that both approaches are neglecting a critical variable that modulates human performance. In this article, I will suggest that the critical variable ne- glected by both prototype and exemplar approaches is the flex- ibility with which humans seek and identify structure in their environment and the extent to which this search for regularities is guided by the learners goals. Neither exemplar nor prototype approaches make room for this flexibility. Irrespective of the nature of the learning problem or the learners goals, a prototype model represents each category by a single prototype, whereas an exemplar model represents each category as the set of its members....
View Full Document
This note was uploaded on 11/15/2009 for the course ILRCB 3020 at Cornell University (Engineering School).