Off-line Reasoning for On-line Efficiency

The complex i ty of reasoning is a fundamenta l issue in A I . In many cases, the fact tha t an intel l igent system needs to per form reasoning on-l ine contr ibutes to the di f f icul ty of this reasoning. In this paper we investigate a couple of contexts in which an i n i t i a l phase of off-l ine preprocessing and design can improve the on-l ine complexi ty considerably. The first context is one in which an intel l igent system computes whether a query is entai led by the system's knowledge base. We present the not ion of an efficient basts for a query language, and show tha t off-l ine preprocessing can be very effective for query languages tha t have an efficient basis. The usefulness of this not ion is i l lust rated by showing tha t a fair ly expressive language has an efficient basis. The second context is closely related to the ar t i f ic ia l social systems approach introduced in [MT90 ] . We present the design of a social law for a mul t i -agent envi ronment as p r imar i l y an instance of offl ine processing, and study this problem in a par t icu lar mode l . We briefly review the ar t i f ic ia l social systems approach to design of mul t i -agent systems, introduced in [MT90 ] . C o m p u t i n g or coming up w i th a social law is viewed as a p r imar i l y off-l ine act iv i ty tha t has ma jo r impac t on the effectiveness of the on-l ine act i v i t y of the agents. The tradeoff' between the amount of effort invested in comput ing the social law and the cost of the onl ine ac t iv i t y can thus be viewed as an off-l ine vs. on-l ine tradeoff.