Flexible Execution of Cognitive Procedures.

Abstract

Several current theories of procedural knowledge hypothesize that procedures are organized as hierarchies of goals, wherein accomplishing a goal requires accomplishing all or some of its subgoals. This form of knowledge is most naturally executed with the aid of a temporary last-in-first-out stack of goals. This article presents evidence that a stack regime is not flexible enough to account for the procedural problem solving exhibited by a sample of 26 third-graders solving subtraction problems. Two alternative control regimes are investigated. One stores goals on an agenda (an unordered set) and the other stores goals in tree. Both the agenda regime and the tree regime employ a rule-based scheduler that picks the next goal for execution. Both regimes succeed at modelling our subjects' problem solving strategies. The tree regime is able to account for data from another study as well. However, a closer examination of the fit between models and data shows that some students change their execution strategies in the midst of problem solving. This finding challenges fundamental assumptions underlying research on cognitive architectures.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jun 30, 1987
Accession Number
ADA186645

Entities

People

  • Kurt VanLehn
  • William Ball

Organizations

  • Carnegie Mellon University

Tags

Communities of Interest

  • C4I
  • Energy and Power Technologies
  • Human Systems

DTIC Thesaurus Topics

  • Acquisition
  • Artificial Intelligence
  • Classification
  • Cognition
  • Cognitive Science
  • Computer Languages
  • Computer Programming
  • Computer Programs
  • Computer Science
  • Computers
  • Language
  • New York
  • Plastic Explosives
  • Programming Languages
  • Psychology
  • Quantum Cascade Lasers
  • Students

Readers

  • Parallel and Distributed Computing.
  • Strategic Security Studies
  • Team-Based Human-Centered Cognitive Task Decision Making and Information Performance.