Here's a concise stab: a set of "weightings and preferences" - that are unique to you - scoped to particular contexts about other people and topics, sourced from your own and your trusted contact's opinions and interactions.

example 1: "do all the people who's opinions I trust on 'human-ness' think a given npub is a bot?"
example: 2: "do all the people who's opinions I trust on 'best pizza' say I'll enjoy a given restaurant?"
example 3: "among the people I trust for hiring talented software developer freelancers, has anyone had good experience with freelancer Y?"

Note that for all of the above, the "answers" are unique to you. You'd expect someone else to have entirely different - even opposite - results!