Architectural Explorations for Modeling Procedural Skill Decay

There is great interest in mission-critical domains to minimize the decay of expert skill: the decrement in performance observed after a period of disuse. To better understand skill decay and to test theories and systems aimed at minimizing its effects, it is useful to create computational models capable of exhibiting its main effects. No models of expert knowledge (represented as production rules) skill decay have been developed within cognitive architectures. This work explores how skill decay might be realized within an architecture. It also discusses implementation issues with a focus on psychologically-plausible approaches and explanations of skill decay effects. Although implemented in the EASE architecture, the issues discussed in this paper are relevant to all rule-learning, production system based cognitive architectures.