Applying Primitive Elements Theory for Procedural Transfer in Soar

Abstract

Detailed transfer of procedural knowledge has been modeled in Actransfer, an extension of ACT-R, by combining the primitive memory operations of productions (PRIMs) with the architecture’s procedural learning mechanism (Taatgen, 2013c). This work explores whether these same principles can be applied to the Soar cognitive architecture, which uses different models of working memory and procedural learning. We confirm that these principles can transfer to an unmodified version of Soar. Our research contributes a novel model of skill learning based upon a deeper level of primitive skill composition than described in the PRIM model that is suitable for unbounded working memory architectures, and which yields transfer profiles similar to those revealed in human studies.

9 Figures and Tables

Cite this paper

@inproceedings{Stearns2017ApplyingPE, title={Applying Primitive Elements Theory for Procedural Transfer in Soar}, author={Bryan Stearns and Mazin Assanie}, year={2017} }