Inconsistent Belief In Common Sense Reasoning Quantitative Non-fix Approach

Since AI becomes a subject, scientists keep to such a definite idea. The idea is to study and simulate common principle of human thinking, which can be realized by computers. Scientists do a lot of experiments to simulate human intelligence and cognitive procedure. They indicate that most of the intelligence system cannot hold on the common sense knowledge. As for common sense knowledge the denotation and reasoning are both difficult. To deal with the inconsistent knowledge is the most difficult and important subject in common sense reasoning.According to different demands in kinds of applying fields, there were many kinds of revision methods to deal with inconsistent knowledge. Among them C.AlchourronN P.Gardenfors and D.Makinson's AGM theory is the most well known. The AGM postulates claimed to be satisfied by any rational revision method. This paper studies several revision methods and puts forward a non-revision method. The main work of this paper can be divided into three parts:(1) The first part of this paper presents several representative iterative belief revision methods, and compare them to AGM postulates .It can be concluded that most of the iterative belief revision methods obey such two rules: One is minimal change principle; the other is strong prime update principle.(2) There are several problems in all kinds of belief revision methods, for example loss of useful information or birth of unexpected information in reasoning procedure and also difficulty to draw a conclusion. Professor Deng has developed a non-revision method which describes the credible degree of belief by quantitative method. Since this method uses the quantitative information to describe the reliable degree of knowledge, it dose not deal with the inconsistent knowledge and uses the heuristic information to instruct reasoning that can make the result by reasoning among source knowledge the most reliable conclusion. So this method can avoid losing useful information and avoid facing the difficulty to select conclusion set.This paper compares the non-revision method to AGM postulates. It demonstrates that non-revision method is a good one.(3) It is well known that the determinant problem in classic proposition logic is a NP complete problem, and the first order logic is a half determinant one. So both the analysis of computing complexity and the realization of it are the most important fields.This paper presents the realization of the non-revision method which is based on proposition logic.