Agents on the Web Probability and Agents

T o make sense of the information agents gather from the Web, they need to reason about it. If the information is precise and correct, they can use engines such as theorem provers to reason logically and derive correct conclusions. Unfortunately, the information is often imprecise and uncertain, which means they will need a probabilistic approach. More than 150 years ago, George Boole presented the logic that bears his name in An Investigation of the Laws of Thought on Which are Founded the Mathematical Theories of Logic and Probability. The title indicates the concern that classical logic is not sufficient to model how people do or should reason. Adopting a probabilistic approach in constructing software agents and multiagent systems simplifies some thorny problems and exposes some difficult issues that you might overlook if you used purely logical approaches or (worse!) let procedural matters monopolize design concerns. Assessing the quality of the information received from another agent is a major problem in an agent system. Here, we describe Bayesian networks and illustrate how you can use them for information quality assessment.