Concrete example: a man disappears. We can speak of the possibility that he
has been murdered, the possibility of finding a weapon, etc. But we would
use the term plausibility when discussing motives or patterns, as in "How plau
sible is it that he was killed by a robber versus a jealous lover?" or "How
plausibly does this crime fit the pattern of the Bengal strangler?".
In terms of modeling, plausibility may be more like a measure of consistency o
ver groups of statements. It could be rigorously defined to be separate from
the actual possibilities of the component statements, so we could consider
Poss(A), Poss(B), and Plaus(A=>B) separately. Of course, if Poss(B)=0 and
Poss(A)=x>0, this should place an upper bound of 1-x on Plaus(A=>B), I would
think.
------------------------------