Inference-based constraint satisfaction supports explanation

Conference · Tue Dec 31 04:00:00 EST 1996

OSTI ID:430674

Sqalli, M H; Freuder, E C ^[1]

---

1. Univ. of New Hampshire, Durham, NH (United States)

Constraint satisfaction problems are typically solved using search, augmented by general purpose consistency inference methods. This paper proposes a paradigm shift in which inference is used as the primary problem solving method, and attention is focused on special purpose, domain specific inference methods. While we expect this approach to have computational advantages, we emphasize here the advantages of a solution method that is more congenial to human thought processes. Specifically we use inference-based constraint satisfaction to support explanations of the problem solving behavior that are considerably more meaningful than a trace of a search process would be. Logic puzzles are used as a case study. Inference-based constraint satisfaction proves surprisingly powerful and easily extensible in this domain. Problems drawn from commercial logic puzzle booklets are used for evaluation. Explanations are produced that compare well with the explanations provided by these booklets.

• OSTI ID: 430674
• Report Number(s): CONF-960876--
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
ALGORITHMS
ARTIFICIAL INTELLIGENCE
DECISION MAKING
INFORMATION RETRIEVAL