A DISCUSSION OF UNCERTAINTY HANDLING IN SUPPORT LOGIC PROGRAMMING

Support logic programming, as introduced by Baldwin, is a technique for dealing with uncertainty in expert systems which significantly improves over early approaches such as certainty factors of MYCIN. Analyzing this technique is the opportunity to point out problems that are still unsolved, and issues that want investigation. The links between support pairs expressing uncertainty and belief functions are exhibited through Baldwin's voting model. Some difficulties pertaining to the representation of uncertain or fuzzy rules in the framework of PROLOG-like languages are pointed out. Propagation and combination rules are also discussed and the SLOP minimum rule of combination is improved, and related to an already known rule for upper and lower probabilities.