Elamur Research⌕
☰
home / papers / 1987 / SOAR: An architecture for general intelligence

SOAR: An architecture for general intelligence

Paper · 1987 · John E. Laird, Allen Newell and Paul S. Rosenbloom

Summary

Soar addresses a central goal of AI: a single fixed architecture that can handle the full range of cognitive tasks, from routine to hard problems, and that improves with experience. Laird, Newell and Rosenbloom cast all goal-directed behaviour as search in problem spaces. Knowledge is stored as productions in long-term memory and matched against a shared working memory. Each decision cycle has two parts: an elaboration phase that fires every matching production in parallel, and a fixed decision procedure that weighs preferences to select problem spaces, states and operators. When knowledge is missing or conflicting, Soar reaches an impasse and automatically creates a subgoal to resolve it (universal subgoaling). Chunking stores the results of these subgoals as new productions, so learning comes from the same mechanism as problem solving. The authors show the system on classic puzzles, weak search methods and knowledge-intensive expert tasks. The work matters because it was one of the first unified cognitive architectures, and it shaped later research on cognitive modelling and integrated agents.

Key points

  • Architecture: problem spaces, a production-based long-term memory and a working memory. A decision cycle alternates elaboration with preference-based selection of spaces, states and operators.
  • Impasses: ties, conflicts, no-change and rejection automatically trigger subgoals. This lets the architecture generate its own goal hierarchy.
  • Weak methods: general methods such as hill climbing, means-ends analysis and lookahead arise from task knowledge rather than separate code (the "universal weak method").
  • Chunking: this is the only learning mechanism. It turns subgoal results into productions, which speeds up later performance and transfers within and across tasks.
  • Demonstrations: tasks include the Eight Puzzle and Tower of Hanoi, plus expert-system work such as R1-Soar for VAX configuration.
  • Source note: the extracted text supplied here contained only the title page (dated draft January 1986) and funding notes. This summary therefore reflects the paper's widely known published content, not the provided extract.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Stanford UniversityStanford · United States434 works · 314,419 citations🎓Carnegie Mellon UniversityPittsburgh · United States539 works · 179,533 citations🔬Palo Alto Research CenterPalo Alto · United States14 works · 8,123 citations

 Fields, topics and concepts

📚 Shelves 2The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 works

💡 Concepts 1Key ideas and methods explained on our concept pages.Artificial intelligence · 49 works

Cite

Cite as John E. Laird, Allen Newell and Paul S. Rosenbloom. 1987. SOAR: An architecture for general intelligence. Artificial Intelligence.

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{laird1987soar,
  title = {{SOAR: An architecture for general intelligence}},
  author = {John E. Laird, Allen Newell and Paul S. Rosenbloom},
  booktitle = {Artificial Intelligence},
  year = {1987},
  doi = {10.1016/0004-3702(87)90050-6},
  url = {https://doi.org/10.1016/0004-3702(87)90050-6}
}

References

18 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Rosenbloom, P. S. The Chunking of Goal Hierarchies: A model of practice and stimulus-response compatibility. Doctoral dissertation, Camegie-Mellon University, 1983, (available as Tech Rep \#83-148, Computer Science Department). find
  2. Rosenbloom, P. & Newell, A. The chunking of goal hierarchies: A generalized model of practice. In Michalski, R. S. (Ed.), Proceedings of the International Machine Learning Workshop. Urbana-Champaign, IL: Department of Computer Science, University of Illinois, 1983. . find
  3. Rosenbloom, P. S., & Newell, A. The chunking of goal hierarchies: A generalized model of practice. In R. S. Michalski, J. G. Carbonell, & T. M. Mitchell (Eds.), Machine Learning: An Artificial Intelligence Approach, Volume II. Los Altos, CA: Morgan Kaufmann Publishers, Inc., 1985. In press (Also ava… find
  4. Rosenbloom, P. S., Laird, J. E., McDermott, J., Newell, A., £ Orciuch, E. Rl-Soar: An experiment in knowledge-intensive programming in a problem-solving architecture. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1985, 7(5), 561-569. (Also available in Proceedings of the IEEE Works… find
  5. Rychener, M. D. Production systems as a programming langaugefor artificial intelligence applications (Tech. Rep.). Computer Science Department, Carnegie-Mellon University, 1976. find
  6. Rychener, M. D. The instructable production system: A retrospective analysis. In Michalski, R. S., Carbonell, J. G. & Mitchell, T. M. (Eds.), Machine Learning: An artificial intelligence approach. Palo Alto, CA: Tioga, 1983. find
  7. Rychener, M. D. & Newell, A. An instructable production system: Basic design issues. In Waterman, D. A. & Hayes-Roth, F. (Eds.), Pattern-Directed Inference Systems. New York: Academic Press, 1978. find
  8. Sacerdoti, E. D. The nonlinear nature of plans. In Proceedings of the International Joint Conference on Artificial Intelligence. IJCAI, 1975. find
  9. Sacerdoti, E. D. A Stmcturefor Plans and Behavior. New York: Elsevier, 1977. find
  10. Shortliffe, E. H. Computer-based Medical Consultations: MYCIN. New York: American Elsevier, 1976. find
  11. Simon, H. A. Search and reasoning in problem solving. Artificial Intelligence, 1983, 21, 7-30. van de Drug, A., Rosenbloom, P. S., & Newell, A. Some Experiments with Rl-Soar (Tech. Rep.). find
  12. Computer Science Department, Carnegie-Mellon University, 1986. (in preparation). find
  13. VanLehn, K. Felicity Conditions for Human Skill Acquisition: Validating an A I-Based Theory (Tech. Rep.). Xerox Palo Alto Reserch Center, November 1983. find
  14. Waltz, D. L. Understanding line drawings of scenes with shadows. In Winston, P. H. (Ed.), The Pyschology of Computer Vision. New York: McGraw-Hill, 1975. find
  15. Waterman, D. A. & Hayes-Roth, F., (Eds.). Press, 1978. find
  16. Wilensky, R. Planning and Understanding: A computational approach to human reasoning. Reading, MA: Addison-Wesley, 1983. find
  17. Winston, P. Artificial Intelligence, 2nd cd.. Reading, MA: Addison-Wesley, 1984. find
  18. PAGE 59 Pattern Directed Inference Systems. New York: Academic find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/laird-1987-soar-architecture-general-intelligence/