ABSTRACT
A critical issue in computational modeling of human cognition is acquisition and encoding of the models and model parameters. Early work in Artificial Intelligence (AI) took the position that such models could be directly manually encoded (Luger, 2002; Russell & Norvig, 2002), whereas more recent exemplars of that tradition assume that cognitively plausible behavior will emerge from sufficiently thorough sets of hand-coded axiomatic world knowledge or heuristics (Lenat, 1995; Rosenbloom, Laird, & Newell, 1993). Although it is clear that manually encoded knowledge will continue to be a significant part of cognitive models for the indefinite future, it currently appears that complete manual development is not scalable, either to developing a single model of general cognition or individualizing models to a variety of different human subjects. Thus, a key facet of current cognitive modeling efforts is an automated knowledge capture component. This component employs techniques drawn from the fields of User Modeling (UM), Machine Learning (ML), Human-Computer Interaction (HCI), and preference elicitation to derive knowledge representations and model parameters for individualized cognitive models. In this chapter, we give an overview of such techniques, focusing on their application to cognitive modeling.
