A note on the notion of truth in fuzzy logic (Q5940144)

In a 2-valued logic, once we assign truth values to atomic formulas, we can use logical operators to determine those of arbitrary formulas, and thus get an interpretation of the underlying theory. It is well known that in every interpretation, every provable formula is true, i.e., has truth value 1. In general infinite-valued logics, when we start with truth values for atomic formulas, the truth values of the corresponding formulas may become undefined -- due to the lack of some infinite supremum or infimum in the underlying structure of truth values. In particular, the truth value of a provable formula may become undefined. It turns out that in fuzzy logic, if we restrict ourselves only to total (``safe'') interpretations, i.e., interpretations in which every formula gets a truth value, then in each such interpretation every provable formula is true. The authors prove that this restriction to safe interpretations is necessary: specifically, they prove that for each continuous t-norm that is different from \(\min(a,b)\), there exists a provable formula whose value in some (un-safe) interpretation exists but is different from 1 (= ``absolutely true'').