> But humans don't work like that. When we think "innocent," we don't think "I have insufficient data," we think "they didn't do it". It is arguably a flaw in human thinking, but it's a flaw we have to live with and work with.

This interestingly maps to computer science. When talking about accusations for some reason we expect outcome to be binary, while the outcome is ternary. Binary logic is difficult, ternary logic is even more so.

Given ternary input state of accuser (false accusation, real accusation, ¯\_(ツ)_/¯ accusation) and ternary input of state of accused (guilty, not guilty, ¯\_(ツ)_/¯), I try to think of logical (relational?) operator/function which combines these inputs and yields ternary verdict. Given that output should be undefined unless both states are known, there should be AND between the states. Maybe something like Verdict = Accuser AND NOT Accused which would yield a verdict only if both accusation and guilt can be proven. But we still have a can of worms unopened regarding evidence and what evidence actually proves, because more likely than not evidence will prove or rule out possibility that something has happened