"Trust" Considered Harmful

In the context of a web of trust, it takes linguistic gyration to make the word "trust" useful. The reason it does get used is historical -- these ideas came out of the security community. But in a modern context the word "trust" is either is so vague that it's useless or is outright misleading.

If you trust someone to do a particular thing or be a particular way, it means that you believe you have an acceptably good prediction. The idea breaks down into useful components:

1. You believe: Tit-for-tat interactions are based in the self. There is a You to profit, and in this case it does trust its own judgement.


2. in a prediction: The judgement is an analysis of existing data for the purpose of deciding whether some transaction will be profitable.


3. that you made: Nobody but you shares your self interest, but others do affect it. So you have a logical system based on induction. I extend the ability to affect me to parties that I directly interact with (the 0 step). Parties interacting with parties that have the ability to affect me have the ability to affect me (the n+1 step).

The only "trusting" going on is between you and your judgement -- a hopelessly vague and unquantifiable interaction. All the useful action is in forming your judgement, which is not part of the trust action, so focusing on trust is misleading.

Rather than "I trust Alice to do foo because Bob trusts Alice," I would say "I believe it is X% likely that Alice will take this action because Bob believes it is Y% likely and Bob's prediction is Z% likely to be correct. X% is a high enough likelyhood that the risk of being wrong is acceptable." The first statement is folk wisdom, the second is quantitative reasoning. One is fuzzy shorthand, the other is clean and utilitarian.

If your prediction is good enough to meet a threshold of acceptable risk, you decide to trust. That's the only simple and binary thing here: do you do it or not?

Transitive Trust

In the context of webs of trust, you often hear discussions of whether trust is transitive. Someone might say "I don't believe that trust is transitive." Does trust transfer intact as it ripples through a a chain of relationships? Yes, if and only if every possible prediction is either 0% or 100%. But if any prediction is between those values, no. There is a transitive relationship, but it is not syllogistic, because syllogism is a purely boolean operation.

Transitivity and syllogism are not the same thing.