Generating Facts from Opinions with Information Source Models

Abstract

This paper describes an attempt to model the method of generating a fact from the opinions of other persons or of institutions as a process which is based on knowledge about these opinion sources. AI Techniques of reasoning, such as rule based deduction, require an input of facts with a definite truth value. In domains like economics or administration, decisions are often based on a set of diverging subjective opinions when the fact which is needed is not available. The credibility of an opinion is considered dependent on a number of information source attributes producing the opinion as competence and goals. A model of the information source containing these attributes is used as a knowledge structure in a cyclical process of subsequent combination, doubt and reinterpretation of single opinions. Information Source Models can be used like User Models to represent an intelligent system's knowledge about its social context.