Two Soar Studies: Towards Chunking as a General Learning Mechanism

John E. Laird, Paul S. Rosenbloom, Allen Newell

R-1-Soar: An Experiment in Knowledge-Intensive Programming in a Problem-Solving Architecture

Paul S. Rosenbloom oot Now at the Standord University Departments of Compuer Science and Psychology. , John E. Laird oot Now at the Xerox Palo Alto Research Center. , John McDermott, Allen Newell, Edmund Orciuch

CMU-CS-85-110

January 1985

The Soar project is attempting to build a system capable of general intelligent behavior. We seek to understand what mechanisms are necessary for intelligent behavior and how they work together to form a general cognitive architecture. Our knowledge about this enterprise is reflected in the evolving assumptions embedded in the Soar architecture Laird, 1985 oot Soar User's Manual , J.E. Laird . The body of this report consists of a pair of papers lightly edited from their original published form reporting on investigations with Soar into two components of general intelligence: learning, and performance in knowledge-intensive tasks.

The first paper is titled Towards Chunking as a General Learning Mechanism . Chunks have long been proposed as a basic organizational unit for human memory. More recently chunks have been used to model human learning on simple perceptual-motor skills. In this paper we describe recent progress in extending chunking to be a general learning mechanism by implementing it within Soar . By implementing chunking within a general-problem solving architecture we take significant steps toward a general problem solver that can learn about all aspects of its behavior. We demonstrate chunking in Soar on three tasks: the Eight puzzle, Tic-Tac-Toe, and a part of the R1 computer-configuration task. Not only is there improvement with practice, but chunking also produces significant transfer of learned behavior, and strategy acquisition.

The second paper, titled R-1 Soar: An Experiment in Knowledge-Intensive Programming in a Problem-Solving Architecture , presents an experiment in knowledge-intensive programming in Soar . In Soar , knowledge is encoded within a set of problem spaces, yielding a system capable of reasoning from first principles. Expertise consists of additional rules that guide complex problem-space searches and substitute for expensive problem-space operators. The resulting system uses both knowledge and search when relevant. Expertise knowledge is acquired either by having it programmed, or by a chunking mechanism that automatically learns new rules reflecting the results implicit in the knowledge of the problem spaces. The approach is demonstrated on the computer-system configuration task, the task performed by the expert system, R1 .

30 pages

Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by [email protected]