A lot of the problems with decentralised reputation systems (or indeed decentralised anything systems) stem from the difficulty of establishing a decentralised identity system.
I wish that the idea of "idchains" had been explored more, but it seems it didn't get much further than a (now unlisted) video on the inventor's YouTube channel:
The (perhaps controversial) idea is that you accept that your face is public data, and so use that as sort of the unique key in a global database of faces, or ideally, hashes of faces. A user could then carry around a digital copy of their face picture (in a mobile app, say), which other users could attest and check the hash of in a global blockchain.
There would need to be gamified incentives to do this attestation (effectively key-signing parties for non-geeks), and you could opt to have people take pictures of you each time and upload them to a decentralised file store (like IPFS). Organisations could then compete on using machine learning to generate confidence scores that a given identity really had been in use for a significant duration, vouched for by other individuals, and not a duplicate of another account somewhere else in the system.
There are various trade-offs in terms of security and usability and anonymity, but it seems like an idea that is worth exploring.
people's faces change with some regularity. Also there are people who are incapable of recognizing faces, or just especially bad at it, so the method is unfair to them.
The idea of faces changing is handled by the idea of regularly (perhaps once a year) taking a new picture of yourself and having it re-attested by witnesses (perhaps ones randomly chosen, within reason, by a third party). Algorithms or human judges can then give a credibility score of whether the person is genuinely ageing over time or reusing a batch of old pictures, for example.
As for the issue of prosopagnosia, that's interesting, I hadn't thought of that. I think that faces could be used just as a bootstrap, equivalent to a "forgot my password" process, to enable someone to associate a blockchain-facial identity to a public key generated on a device. (Also, apparently people with one type of the condition, associative prosopagnosia can have the ability to answer "same/different" about two pictures, even if they can't tell you who is in the pictures.)
I haven't seen a full threat model or a fleshed out protocol of how all the pieces in the system would interact, but it's good to hear some of the hurdles it would need to overcome, thanks.
If sequencing a DNA sample became as cheap and common as taking a selfie, then I could see that happening, yes. I also think that would be a very strange world to live in, but recent events have shown that strange worlds are hard to rule out.
Often it's useful to be able to trust an online identity, i.e. for peer-to-peer lending sites, peer-to-peer home exchange sites, work assignments, etc.
Most often the various platforms (like Couchsurfing, Uber, BeWelcome, Loanbase, Localbitcoins etc) will have some built-in support for ratings; user account A can leaving positive feedback for B after A and B has had an interaction on that platform. The ratings never leave the platform. There are usually no ways to see the difference between a conman creating yet another fake account and a newcomer to the platform. It is impossible for an outsider that have been scammed on one platform to leave a negative feedback warning users on another platform about the scammer.
What we need is a global network-of-trust system. It should be truly peer-to-peer, it should be based on open standards and open source. Just the risk of getting negative feedback on such a network could be enough deterrent to avoid certain types of scam.
I had a similar idea a few years ago but decided to make it wholly objective. Basically, individual actors could publish cryptographic proof that they had engaged in a certain transaction with a certain party and reached a certain outcome, without necessarily disclosing the full details of the transaction. This information could then be analyzed and collected by those considering entering in to a transaction with that party, or perhaps paid agencies who acted as caches to do this for the general public. Quite a lot of related thoughts at http://ifex-project.org/
> I need to confess something: ‘‘Whuffie’’ would make a terrible currency.
http://www.locusmag.com/Perspectives/2016/03/cory-doctorow-w...