Building an explanation generation mechanism in probabilistic knowledge-based systems

This paper proposes a Multiple-run Interactive Certainty Network (MRICN) that integrates human decision-making heuristics into probabilistic approaches for knowledge-based systems. MRICN is built upon ideas drawn from “Opinion Pooling”, “Probabilistic Network” and “Interaction”, allowing for reflective searching for the optimal sets of knowledge with the maximal certainty gain. It is implemented and experimented in an coastal lake sustainability assessment scenario.