aragon
AGP 3: AN Ranking/rating system
Goals
- Create a ranking/rating system that tracks a voters accuracy
- A ranking/rating system that can also act as a reputation system
Description
Since the governance is based a lot on Game Theory in many aspects, why not use a ranking/rating system that has been created for games.
I would suggest Glicko-2 as a viable option for a ranking/rating system that can keep track of how well voters perform and that value can act as a kind of a reputation system. The reputation could then also be used for validating the Judges for ANJ Supreme Court, or even have the 5 random Judges on Decentralized Court have a higher change of getting picked by weighting the odds a little by the reputation system.
Glicko-2 could be implemented for all Yes/No votings at least.
The Glicko rating system and Glicko-2 rating system are methods for assessing a player's strength in games of skill, such as chess and go. It was invented by Mark Glickman as an improvement of the Elo rating system, and initially intended for the primary use as a chess rating system. Glickman's principal contribution to measurement is "ratings reliability", called RD, for ratings deviation.
The RD measures the accuracy of a player's rating, with one RD being equal to one standard deviation. For example, a player with a rating of 1500 and an RD of 50 has a real strength between 1400 and 1600 (two standard deviations from 1500) with 95% confidence. Twice the RD is added and subtracted from their rating to calculate this range. After a game, the amount the rating changes depends on the RD: the change is smaller when the player's RD is low (since their rating is already considered accurate), and also when their opponent's RD is high (since the opponent's true rating is not well known, so little information is being gained). The RD itself decreases after playing a game, but it will increase slowly over time of inactivity.
The Glicko-2 rating system improves upon the Glicko rating system and further introduces the rating volatility σ.
Further reading about Glicko-2: Example of the Glicko-2 system -paper with formulas: http://glicko.net/glicko/glicko2.pdf Glicko rating system on Wikipedia: https://en.wikipedia.org/wiki/Glicko_rating_system Glicko-2 Javascript implementation: https://github.com/mmai/glicko2js
Uncertainties
This was a thought and an idea of what might work, i haven't had time to really get into thinking how this would actually work, so further research and feedback on the matter from others too would be appreciated.
When you say "create a ranking/rating system that tracks a voters accuracy", what do you mean exactly? Is this for AVT holders, or is it for the decisions made by the decentralized court?
If this is for AVT holders, then I assume that accuracy is a direct measure of the most popular vote, which would lead to some odd and undesirable side effects.
@onbjerg There could be two different systems, one for governance and one for the Jurisdiction that could or not, intersect somehow.
So most popular would be the winning side yes, but how much resistance there were could also be measured, how many losing votes, which would create the result of how much does the result affect the persons reputation in proportion to the popularity of their choice.
Glicko-2 is interesting, though I'm not sure how exactly it would be applied to governance (rather than games) for either voting or for judicial situations since there is no objective correct outcome. There is the winning side in the decision, but that does not mean that the other side was objectively wrong, just that at the time the winning side was more popular.
I'd be worried that this might create a bit of a recursive feedback loop where popular ideas become more popular simply because they were popular in the first place.
Essentially, I don't like the idea of tracking a voters "accuracy" as the goal of voting is not to guess the outcome of the vote but rather to voice an opinion.
This is really interesting Idea that may gave merit for some governance. not to echo too much the previous comments but it depends what kind of votes the the ranking is being applied to.
Where voting is not value based, as in the voting on something measurable and objective a ranking makes sense.
For example if the goal of proposal X was to increased net profit in Q4 by 10%. The vote passes 60:40 and profit had increased 12%, those who voted no or abstained would loose RD while those that voted yes would gain.
Glicko-2 is interesting, though I'm not sure how exactly it would be applied to governance (rather than games) for either voting or for judicial situations since there is no objective correct outcome.
As far as judicial decisions go, the white paper it mentions decisions being based on a prediction market. This can work if that is the case
For example if the goal of proposal X was to increased net profit in Q4 by 10%. The vote passes 60:40 and profit had increased 12%, those who voted no or abstained would loose RD while those that voted yes would gain.
This is describing something closer to a prediction market than a vote towards governance, assuming the intention of the vote was to determine if the the proposal would meet its goal (prediction) rather than wether it was worthwhile to do it (governance). Someone could have voted "no", not because they didn't think that the proposal would increase net profit by 10 percent, but because there was a better proposal that they thought would increase net profit by 15 percent and doing one meant not doing the other.
The Futarchy model attempts to blend governance and prediction markets, and perhaps a reputation system could be useful on the prediction side.
As far as judicial decisions go, the white paper it mentions decisions being based on a prediction market. This can work if that is the case
This is true, and I think in the case of judges picking the most popular interpretation of some rule is a fairly close approximation for "accuracy" so this might work well.
However, it still worries me that it might result in "groupthink" among judges. I'm unsure whether that would have a net positive or negative effect.
The Futarchy model attempts to blend governance and prediction markets, and perhaps a reputation system could be useful on the prediction side.
Exactly "vote on values, bet on beliefs"
Someone could have voted "no", not because they didn't think that the proposal would increase net profit by 10 percent, but because there was a better proposal that they thought would increase net profit by 15 percent and doing one meant not doing the other.
This was just a naive example. But the only way I see a ranking system working in governance is by some futarchy system.