imagine a political economy in which every five to ten years after a government institutes a policy/set of policies, a decision oracle (everyone in the country) votes on whether they were good or not.
There is already this thing in politics where if you opponent proposes a good policy change, you vote against, because you do not want your opponent to be associated with a good thing. Two weeks later, you propose the same thing, using different words, maybe with some modifications that benefit you.
This mechanism would give even more incentives in the same direction. First, even more motivation to vote against what your opponents suggested, because granting them success would not only make them popular but also increase the power of their voters. Second, if your opponents succeed to pass a law, you have even more motivation to sabotage it using any means available; doing so will reduce the power of their voters.
I see this proposal as inferior to the original formulation of "vote values and bet beliefs" - largely because it conflates these two stages.
So I'd prefer to address the problem by separating voting for desired outcomes from prediction markets about how to achieve them - and then ignoring pundits who express views about the latter without putting their money where their mouths are!
Confidence level/s: very speculative, low
The problem
It seems that one of the key optimization problems in democratic decision making is that we often have little incentive to vote for policies that we truly think will increase the public utility.
As per Jess Whittlestone at Vox:
The solution
A possibly workable idea to improve the utility-maximizing effectiveness of our politics is based on the premise that we might better at deciding as a demos whether a policy was good or not in retrospect than we are at deciding whether a policy will be good in the future.
Thus, imagine a political economy in which every five to ten years after a government institutes a policy/set of policies, a decision oracle (everyone in the country) votes on whether they were good or not. If they are found “good” with a high majority—say 70%—everyone who voted for them gets paid; if they are indecisive (i.e. 50-50) no one gets paid. If they are found “bad”, everyone who voted against them gets paid. We could fund this with a zero-sum prediction market, in which citizens could optionally bet on whether the policy would eventually become to be considered successful. Or, we could set up the system as a way to distribute “social dividends/UBI”, which everyone seems to agree are soon coming. Either way, this should have the intended effect of encouraging everyone do a bit more research into the policies they are voting for; voting for the “right” policy is a money-making opportunity. Likewise, there is an additional incentive for a government to make their policy decisively work, as once initially voted for, if the benefit of a given policy is unclear in retrospect (that 50-50 situation), then no one gets to make any money.
(How do we manage this without giving up voter anonymity? Probably cryptographic digital voting, something along the lines of what Benaloh 1987 initially proposed. Five years after the fact, you cryptographically sign a message confirming you own the key that, according to the public anonymous voting ledger, originally authorized the favorable policy.)
But perhaps there is a better way, incorporating my feedback system with Eric Posner and E. Glen Weyl's "Quadratic voting system," in which citizens are distributed a yearly budget (or per referenda) of "voting credits" that they can spend or save. If a citizen doesn't feel strongly about a given vote (providing they only consider its effect on them personally), they can save their votes for a future election in which they do feel strongly; the power of their voting credits decreases with a quadratic penalty, however. That means that 1 vote credit will buy you 1 vote, 2 vote credits 1.4142… votes, 100 vote credits 10 votes and 150 vote credits 12.247 votes, etc.
What if we ran the same feedback system as above, with citizens retrospectively judging the success of policies—except this time, citiens who voted wisely are rewarded with more voting credits, and vice versa. With this system, we could also better apply the “negative feedback” aspect of the process, penalizing citizens for voting unwisely. How do we do this? Well, we could alter the square by which someone is penalized for committing additional vote credits to a particular referenda. For example, if you voted more frequently in favor of “good” policies than in favor of “bad” policies over a period of ten years, you find your votes per policies decreased by: 1.5√n While if the opposite happens, your next ten years have the greater penalty: 2.5√n Thus, we may have a sort-of solution to John Stuart Mill's quandry over how to assign more voting power to wiser voters, an idea which he ultimately rejected as there is hardly an objective metric by which to sum up "wisdom"; the system would become aristocratic. This way, however, a democracy restrospectively decides its most prudent voters in a decentralised and anonymous fashion.