That exists now btw, with verification based on physical passports. It will soon plug into other forms of verification (social network, fingerprints, stake-based bounties, etc) to give higher certainty of one person one vote. But yeah I would bet any distant-future consensus mechanism has proof-of-humanity as its basis, using a democratic decision making layer to verify transactions. https://coinpassport.net/
This is fairly new tech academically btw, and it's only possible anonymously since the ZSnark cryptography tricks.
Passports - issued and verified by centralised entity
Fingerprints, I infer that's basically digital signatures in use today - also controlled by centralised entity
Social networks - to base your auth on the bot moderated and bot infested 3rd party platform with zero support functions is a laughable idea
Stake based bounties - what does this even mean? Inferring from stake based, you meant that the auth will be based on the stake, meaning on the wallet with tokens? Then that's precisely the problem I've described above.
There is no clever technical way to solve social problem. Human identification is a social problem. You can't anonymously identify a person, unless dragnetting through his life with spyware scripts and fingerprinting (great idea for privacy, I'm already dreaming about it).
Passports - the cost of creating fake passports or removing valid ones to most lawful countries is prohibitive enough and visible enough to watchdog groups that we're not under any serious threat from governments going rogue here anytime soon. It's a decent start. At least a few thousand $ cost to forge security wise - increased the better your security market is watching for it.
Social networks - I mean furthering social verification by having networks of users (probably bootstrapped with their passports, and whatever other verification data - video, fingerprints, documents, etc) vouch for each other as real-world connections. You dont need Facebook etc for that - you build that into your identity system as a second layer which compounds the trust. A network discovering a fake user gets financially penalized for vouching (stake based bounties) so there's incentive to really do what you're asked - verify in person someone you've known for some time.
We don't need to anonymously identify a person. In fact, I'm expecting people will be willingly supplying some ungodly amount of data to prove they're real for the financial incentive alone (securing larger loans based on trust). But we can make the verification functions taking all this data go through smart contracts with zero knowledge proofs that can ensure said data doesn't leak to anyone the user doesn't want to share it with - and the protocol can establish a trust score.
And even if we failed to keep personal data off the internet, regardless of how public your profile is you'll always be able to use those zero knowledge proofs to setup an anonymous avatar with a proof hash that it belongs to (exactly) one existing profile from the set of verified profiles, allowing you to vote with the avatar while the system can still guarantee one-person-one-vote. So - anonymous voting.
This is fairly new tech academically btw, and it's only possible anonymously since the ZSnark cryptography tricks.