Believability of Sensemaking Support Systems Using a Bayesian Network

摘要

We show how experts change their opinions based on recommendations by a computer-aided sensemaking system. A three-tier information level consisting of a prior expert judgment, computer generated hypotheses, and posterior expert judgment, are used to compute the expert 'believability' index. Believability index (BI) is defined here as the level or scale (0% to 100%) in which experts, in consensus, believe on the computer aided recommendations. Belief functions are used to model the expert notional beliefs using a Bayesian conditional probability network. The question of generalizing belief function is captured in our model using fuzzy residuum concepts.