Comprehensive Working Memory Activation in Soar

Memory activation has been modeled in symbolic architectures in the past, but usually at the level of individual chunks or productions in long term memory. Recent research (Chong 2003) has demonstrated activation at the level of individual elements of working memory. In this paper, we present a comprehensive implementation of working memory activation in Soar that takes advantage of the unique characteristic of Soar's working memory structure, namely persistence. We also explore modifications to activation so that the activation of new working memory elements is not a fixed level, but is based on the activation of the working memory elements tested in its creation. We demonstrate our model in terms of how it aids the selection of features relevant to learning.