ASSUMPTION ANALYSES IN COMPILED LOGIC-BASED DECISION SUPPORT SYSTEMS (LDSSs)

We describe a LDSS which extends and improves on prior proposals. It uses a connection graph to compile programs for handling ordinary queries, thus eliminating the need for deductive search at query time. It provides for assumption analyses (like 'what-if' questions) by allowing changes to the rules (clauses) of the system and tracing the effects of such changes, thereby avoiding the need to recompile the whole system. It allows certain existentially quantified formulas to participate in the LDSS as constraints on answers produced by rules. Such a LDSS exhibits desirable characteristics of general decision support systems like handling semistructured problems and being supporting, descriptive, effective, and evolutionary. Future work includes interfacing to large statistical packages and dealing with aggregate functions like average.