Modelling Representations of Device Knowledge in Soar

Abstract

This paper presents two simulation models which address the effect of alternative instruction types on the development of skilled performance when using a device. Two instruction types are considered: (i) “how-to-do-the-task” or “operational” knowledge, which provides step-by-step action sequences specifying how to perform typical tasks, and (ii) “conceptual device model”, “how-the-device-works” or “figurative” knowledge, which specifies the effects of users’ actions on the device. Performance differences between groups presented with these alternative instruction types suggest that presenting users with a conceptual model facilitates the development of robust skilled performance. Using Soar, a theoretically committed cognitive architecture, the impact of instruction type on the development of skilled performance is addressed through simulation modelling. Skilled performance is assumed to result from the execution of automatised solution methods which specify procedures for accomplishing goals. It is demonstrated with these simulation models that even where observed performance may be identical, the knowledge that underlies skilled performance may differ considerably given alternative instruction types.