A Fuzzy Model for the Accumulation of Judgments by Human Experts

Abstract

Abstract A formal model for the justification and checking of arguments is proposed for the development of expert systems. The two major parts of the model consist in the detection of violations of the rules governing rational discourse and in the evaluation of conclusions in fuzzy syllogistic reasoning. The application of the model consists in the interaction between the expert (‘arguer’) and the formal model. In the beginning the train of arguments is checked for rule violations and the expert (‘arguer’) is asked to mend his/her argument accordingly. Then the acceptable train of arguments, that is, a syntactically valid chain of propositions is evaluated according to its fuzziness. Upper and lower bounds for the acceptability of the train of arguments are calculated and fed back to the expert (‘arguer’) in either numerical or verbal form. After that the expert decides either to choose a level of acceptability in the given bounds or suggests an expansion of the justification. In the latter case the evaluation in both steps starts anew. First experimental results in predictions for everyday events indicate that the model is viable, fits into the subjects' cognitions, and helps the subjects to arrive at trains of arguments which are more acceptable for them than their initial justifications.