Department of
Computer Science
695 Park Ave.
NY, NY 10021

Susan L. Epstein

The CUNY Graduate School, Department of Computer Science and

 Hunter College, Department of Computer Science



     Media Appearances 







FORR is a satisficing architecture for learning and problem solving. Its premise is that agreement among reactivity, heuristics, planning, and search is a valid decision- making principle. A FORR-based program learns to specialize general domain expertise into problem-class specific expertise. An implemented FORR shell supports domain-specific development. FORR explores the relationship among reactivity, heuristics, and search. Applications include Hoyle for game playing, Ariadne for robot path finding and ACE for constraint satisfaction. Each of them has demonstrated substantial, learned expertise after relatively little training.


Current Work

- The integration of planning with high-level reasoning.

- Learning new heuristics.

- Autonomous restructuring of a decision hierarchy.

- Metaknowledge for heuristics.

- Multiagent solutions to multiple goal problems with (CD).


Application Domains

Adaptive Constraint Engine (ACE)

Game playing (Hoyle)

Wayfinding (Ariadne)

Human-Multi-robot Teams (SemaFORR)

Spoken Dialogue Systems (FORRSooth)


Key references

Epstein, S. L., & Petrovic, S. (2012). Learning Expertise with Bounded Rationality and Self-awareness. In Metareasoning: Thinking about thinking: MIT Press.

Epstein, S. L., E. C. Freuder and M. Wallace 2005. Learning to Support Constraint Programmers. Computational Intelligence. 21(4): 337-371.

Epstein, S. L. 1994. For the Right Reasons: The FORR Architecture for Learning in a Skill Domain. Cognitive Science, 18 (3): 479-511.


Additional references

Epstein, S. L. (1995). On Heuristic Reasoning, Reactivity, and Search. In Proceedings of the Fourteenth International Joint Conference on Artificial Intelligence, 454-461. Montreal: Morgan Kaufmann.

Epstein, S. L., Gelfand, J. and Lesniak, J. 1996. Pattern-Based Learning and Spatially-Oriented Concept Formation with a Multi-Agent, Decision-Making Expert. Computational Intelligence, 12 (1): 199-221.


This material is based upon work supported by the National Science Foundation under Grant Nos. #IIS-0328743, 9423085, #IRI-9703475, 9222720, and #9001936, by the New York State Technological Development Graduate Research and Technology Initiative, and by the PSC-CUNY Research Foundation.

Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation, New York State, or PSC-CUNY.

[top of the page]