Proceedings of the Annual Meeting of the Cognitive Science Society

We present a framework to assess the relative cognitive cost of alternative representational systems for problem solving.The framework consists of 19 cognitive properties of representational systems, which are distributed across 4 dimensions(registration, semantic encoding, inference, and solution) and three scales of granularity (symbol, expression, and sub-representations). It examines components and processes spanning the internal mental representation and external physicaldisplay, and also addresses heterogenous representations of problems. We provide functions to evaluate the cost of eachcognitive property by examining, for example, types of matches between display symbols and concepts, the arity ofrelations, or the depth of solution trees. The cognitive costs for each property are combined to estimate the overallcognitive cost, and hence the relative effectiveness, of a representation. The frameworks development is motivated byour goal of engineering an automated system that will select representations suited to specific classes of problems forindividual users.