RCL: An A. I. Tool for Computing Maximal Consensuses

摘要

RCL is an original Artificial Intelligence tool for computing maximal consensuses among conflicting agents in the Boolean logic framework. It extracts one maximal set of information that does not conflict with any of the agents, taking all possible deductions into account. This form of consensus can go far beyond what is actually shared by all agents. It might be endorsed by every agent since it does not conflict with any of them. Several types of preference criteria can be selected and iterated in order to orient the computation towards some preferred consensuses. From a computational point of view RCL relies on an internal transformation of the maximal consensus-finding problem into one single instance of the Weighted Partial Max-SAT problem.