[Muromec]

From the point of view of the bank the problem is usually defined as

"how do we asses a complex situation where identity of the person X is one of the signals (but maybe not the strongest one) with enough certainty to balance a probability Y of bad something happening that will cost us Z and still make money"

Most of the time Y and Z are defined because the other department said so and we trust our colleagues, dus the answer is computable (somebody somewhere has it open in a spreadsheet right now).

If you add a transitive property to the system, then, unless there is some regulatory magic that caps the possible value space of Y and Z, the answer is (by default) no.