Collaborative Research: A behavioral and computational investigation of the generality and transferability of category representations

Purdue University, West Lafayette IN

Investigators

Abstract

How people acquire and use different types of knowledge is a fundamental issue in cognitive science, applicable to problems in education, training, and the development of expertise. For example, learning to categorize types of materials (such as natural vs. synthetic, or polymers vs. ceramics) can be accomplished using textbooks, but it can also be accomplished with hands on experience in the field. Different types of training likely lead to different forms of knowledge, and the form of knowledge may constrain how, and in what situations, the knowledge can be used. Thus, the best way to train a person may differ depending not only on the type of information being learned, but also on the situations in which the knowledge will need to be used. The investigators will examine how to promote the learning of different forms of knowledge in different situations. They will also investigate the neural and computational bases of the differences in forms of knowledge in order to develop a unifying theory of how knowledge acquisition and application varies across situations in predictable ways. A larger goal is to determine how knowledge, once learned, can be transferred to new situations. This project will advance scientific understanding of human knowledge acquisition and its use. The project also has the potential to foster the development of new tools that may improve the training of students and professionals. Categorization researchers have made an enormous contribution to the understanding of the many ways in which knowledge can be represented and used. A current challenge facing the field of categorization, and cognitive science more generally, is how to best integrate these findings with the broader goal of understanding the extent to which different types of knowledge representations generalize to new situations. The proposed research utilizes a combination of behavioral, neuroimaging, and computational modeling techniques to address these challenges. Specifically, the investigators will explore (1) the factors influencing the development of different types of category representations, (2) the psychological functions and brain networks supporting category representations, and (3) the utility of different types of category representations for supporting performance in new tasks and/or with new stimuli. In addition, this research will highlight important relationships between machine learning techniques and methods used in cognitive science. As a result, the research should be of broad interest to psychologists, computer scientists, and the general cognitive science community.