Towards analogy-based decision - A proposal

National audienceThis short paper outlines an analogy-based decision method. It takes advantage of analogical proportions between situations, i.e., a is to b as c is to d, for proposing plausibly good decisions that may be appropriate for a new situation at hand. It goes beyond case-based decision where the idea of graded similarity may hide some small but crucial differences between situations. The method relies on triples of known cases rather than on individual cases for making a prediction on the appropriateness of a potential decision, or for proposing a way of adapting a decision according to situations. The approach may be of interest in a variety of problems ranging from flexible querying systems to cooperative artificial agents