Multi-service Graph Reputation protocol

The problem

  1. Users inside centralized services need to know reputations of other users they’re interacting with;
  2. Building reputation with ratings imposes a big burden on the user and still accomplishes nothing, can be faked, no one cares about these ratings etc.

The ideal solution

Subjective reputation: reputation based on how you rated that person previously, and how other people you trust rated that person, and how other people trusted by people you trust rated that person and so on, in a web-of-trust that actually can give you some insight on the trustworthiness of someone you never met or interacted with.

The problem with the ideal solution

  1. Most of the times the service that wants to implement this is not as big as Facebook, so it won’t have enough people in it for such graphs of reputation to be constructed.
  2. It is not trivial to build.

My proposed solution:

I’ve drafted a protocol for an open system based on services publishing their internal reputation records and indexers using these to build graphs, and then serving the graphs back to the services so they can show them to users when it is needed (as HTTP APIs that can be called directly from the user client app or browser).

Crucially, these indexers will gather data from multiple services and cross-link users from these services so the graph is better.

https://github.com/fiatjaf/multi-service-reputation-rfc

The first and single actionable and useful feedback I got, from @bootstrapbandit was that services shouldn’t share email addresses in plain text (email addresses and other external relationships users of a service may have are necessary to establish links from users accross services), but I think it is ok if services publish hashes of these email addresses instead. At some point I will update the spec draft and that may have been before the time you’re reading this.

Another issue is that services may lie about their reputation records and that will hurt other services and users in these other services that are relying on that data. Maybe indexers will have to do some investigative job here to assert service honesty. Or maybe this entire protocol is just failed and we will actually need a system in which users themselves will publish their own records.

See also