The way Benefit Tokens work is really interesting if you think about it:
Everyone can vote to give the tokens to himself. However, he knows that would be useless, because he just has one vote.
The community of agents as a whole can’t converge on any of their selfish desires. The voting mechanism means I have insufficient power to direct the tokens to my personal benefit.
The tokens are assigned collectively, so can only be assigned to something the community converges on. This leaves only some sort of common good, or shared vision of ‘Benefit’. That’s the only thing a community of agents who are not collaborating can converge on without conspiring together.
This is somewhat like the way in a Schelling Oracle, multiple agents can converge on the truth without communicating, except that here we’re converging on the ‘Good’ rather than the ‘True’.
It reminds me a bit of the Awesome Foundation, where participants pool their money and vote to give it to the most ‘Awesome’ project. They could vote to give it to themselves, but the game-rules render such votes ineffective.
Is this a recognised principle in formal Game Theory? I can think of all sorts of uses for ensuring governance is benevolent. This mechanism could be twisted to assign public grants, to target&penalise industries that have negative externalities (like pollution), and to fill a lot of gaps opened up by the selfishness of the standard capitalist game.