I think the premise is fascinating. It sounds like it has the same benefits and pitfalls as a cryptographic web of trust, with the caveat that it’s more accessible to the average human and therefore easier to game. It’s also sociologically problematic in that it tries to rate human beings, a process that seems like it would have an unacceptable margin of error due to inconsistent behavior of the subject, change over time, and subjective bias.
I’ll keep an eye on what you’re doing. I’m interested to see how you mitigate the inherent problems of trust networking.
As to the more general discussion, I do think that knowing somebody can help you judge their reliability — but it can also harm your ability to do so, depending on how their biases align with yours. It’s not something I would recommend to people in general.