Logical consensuses for case-based reasoning and for mathematical engineering of AI

Abstract

We claim that computing forms of consensus among several agents about their solutions to past problems can play a useful pre-treatment role in case-based reasoning. Intuitively, we define a consensus as a subset of the plain accumulation of all the agents' individual past discovered solutions such that every agent can agree on all the information in this subset. A consensus can be expected to form a more reliable basis for further re-use or generalization than the knowledge from which it is extracted. We define various forms of logical consensus in this context: the focus is on computational issues about the automated extraction of consensuses in an extended Boolean logic setting.