Fernand Gobet & David J. Wood

Expertise, models of learning and computer-based tutoring. Computers and Education, 33, 189-207.



In a wide and diverse range of contexts, from academic disciplines through to games and sports, comparisons of the performance of novices and experts have established consistent relations between knowledge, task performance and level of expertise. In this paper, we identify and discuss the theoretical significance of this research in relation to a formal, computational theory of expertise, CHREST, developed out of Feigenbaum and Simon's (1984) EPAM theory. The main thrust of our paper is the argument that the theory both helps to identify and explain theoretical limitations on current approaches to computer based tutoring, and offers a means of overcoming some of these. We argue that, without such knowledge-based models of the learning process, attempts to develop effective, computer-based tutoring systems have achieved limited progress towards the goal of helping learners to construct links between their procedural knowledge and conceptual understanding. Further, current knowledge-base approaches to learner modelling need to be developed in two main directions to reach this goal. First, they will have to integrate a theoretically sound account of the relation between perception and memory (such as that developed within the EPAM approach) in order to build upon what has already been achieved to date in relating memory, learning and problem solving. Second they need an extended theory of declarative (or conceptual) knowledge and its relation to procedural skills in order to explain and model learning with multiple, external representations. We illustrate how the EPAM model of expertise can be exploited towards these ends. We then draw out from the theory a number of implications for the achievement of contingent computer-based tutors to support conceptual understanding, and for curriculum planning and evaluation.