The generalized modus ponens considered as a fuzzy relation (Q1187436)

The paper is an analysis of the rule of generalized modus ponens in fuzzy logic, which was first formulated by L. A. Zadeh in his early works. Several criteria taken from the literature are imposed on the rule. The fundamental assumption is the following: The conclusion \(Q_ 2\) is computed from \(Q_ 1\) and the conditional statement `If \(P_ 1\) THEN \(P_ 2\)' using the formula \[ \mu_{Q_ 2}(y)=\bigvee_ x (\mu_{Q_ 1}(x)\wedge \mu_ R(x,y)),\tag{1} \] where \(R\) is a fuzzy relation derived from \(P_ 1\) and \(P_ 2\) using some kind of implication operation taken from many-valued logic. Further criteria for strengthening or weakening \(Q_ 1\), leading to analogous requirements on \(Q_ 2\), are further imposed. The membership functions are supposed to be continuous. The main conclusion of the paper states that there is no fuzzy relation \(R\) fulfilling all the imposed criteria. The author considers also linguistic modifiers (hedges), e.g., ``very'', ``roughly'', etc., which are modeled either by powers or shift of membership functions, and again, no suitable relation \(R\) exists to fulfil all the criteria. One of the conclusions given by the author is that formula (1) should be changed, and he proposes to find some functional relation. From the logical point of view, we may hardly agree with all the criteria since they cover also counterintuitive conclusions such as ``IF it is warm THEN I am fine; it is very warm, hence I am very fine''. Of course, such a monotonicity may sometimes be desirable but it can easily be reached by introducing suitable special axioms and using common modus ponens of fuzzy (many-valued) logic. The analysis from the point of view of a proper formal many-valued logical system is neither presented nor cited though such a system does exist.