TOTAL UNCERTAINTY REVISITED

A recently proposed [Pal et at, 1992] measure of uncertainty for evide nce assignments is analyzed. Termed total uncertainty, it becomes, for assignment m: A bar right arrow m(A) TU(m) = -Sigma m(A) logm(A)/\A\. We review its axiomatic characterization and properties, establish it s relationship with probabilistic entropy, and discuss its extremal pr operties. We then demonstrate that TU(m) is a representative instance of a class of related measures TUk(m) = -1/k Sigma m(A) logm(A)/\A\(k) , which share the same axiomatic properties. Finally, we apply TU(m) t o possibilistic assignments, find its maximum, and compare to that of pos sibilistic information U(p).