Deception Detection in Expert Source Information through Bayesian Knowledge-Bases

摘要

The goal in this effort is to automatically detect deception by an individual or expert who is contributing to an information knowledge-base consisting of multiple experts. Contemporary decision makers often must choose a course of action using knowledge from several sources. Knowledge may be provided from many diverse sources, including electronic sources such as knowledge-based diagnostic or decision support systems, through techniques like data mining. As a decision maker's sources become more numerous, detecting deceptive information from these sources becomes vital to making a correct, or at least, more informed, decision. This applies to unintentional misinformation as well as intentional disinformation. The authors are developing formal definitions for a deception attempt, the deception core, effective deception, and successful deception. A deception attempt occurs when the opinions returned to a decision maker by an expert agent are not those calculated by that expert agent through observations, but are intentionally substituted to influence the decision maker's actions. The deception core refers to those opinions that are manipulated to form a deception attempt. An effective deception is a deception attempt that succeeds in altering the actions of the decision maker, though not necessarily to the actions desired by the deceptive expert. Finally, a successful deception is an effective deception in which the alternate actions chosen by the decision maker are the actions desired by the deceptive expert. This ongoing research focuses on employing models of deception and deception detection from the fields of psychology, cognitive science, and artificial intelligence as well as implementing deception detection algorithms in probabilistic, intelligent, multi-agent systems.