RET - A LOGIC FOR RELATIVE EVIDENTIAL SUPPORT

A formal logic-RES for Relative Evidential Support is proposed based o n the following ideas: 1. Arguments are represented directly and are t aken as the terms for RES. By arguments we mean the relationships betw een two judgments expressing that one supports or refutes the other, i e, evidential supports; 2. Comparisons between arguments are used as c arriers for their relative strengths. By comparisons we mean relations hips between two arguments with respect to their strengths. 3. Uncerta inty reasoning is viewed as a process of three phases of: evidence str ucture construction; evidence accumulation; decision-making. Some exam ples are presented showing how RES can be used to represent various ki nds of uncertain information such as: the relative strengths of eviden tial supports, eg, evidence e, supports conclusion p, better than evid ence e2 supports p2; belief functions, eg. evidence e1 is exhausted by stating a belief function with mass function m1; necessity assertions , eg, the necessity function derived from the membership function f or predicate F on domain D; probability assertions, eg, the probability that statement s is true is .99. These examples illustrate the advanta ges of RES over other representations of uncertain information and evi dential reasoning, eg, 1. it is based on relative strengths of argumen ts that cannot be represented using any absolute measures; 2. It is ca pable of representing explicitly the design of evidence [21]; 3. It ca n represent many kinds of absolute measures-in doing so it has the mer it that it explicates the assumptions and requirements for using diffe rent kinds of measurements of uncertainty; 4. It provides a natural co mmon base for a hybrid system. Our conclusion is that RES is a suitabl e formalization for uncertain information.