AGP4: System for resolving subjective and value based disputes

Goals

I would like to propose a system of judging issues that are subjective and based on shared personal values in a limited population.

Description

Suppose an Aragon company CEO has a clause in his contract that states the board members can terminate him at any time, without any sort of compensation, if he publicly engages in "inapropriate behavior while representing the company". Now this is usually a typical clause in a traditional contract, but something very hard to enforce in digital blockchain companies.

Now let's imagine that the CEO organized a party with strippers on the company headquarters, which was amply photographed and shared on social media, and the board has decided this violates the "inappropriate behavior clause". The CEO, believes that the board wanted to get rid of him any way, and that this is just an excuse to avoid paying him the million dollar compensation for being fired, so he will try to fight it.

How can a contract define "appropriate behavior"? This will deeply depend on personal moral views on employment of sex workers: while some group of people might see this as extremely innapropriate, others believe that this is a perfectly honest job. Even among those, the line is blurry: what if it was a prostitute instead of a stripper? What if the CEO actually paid them for a private service?

This is where this sort of oracle could come in:

The system picks a mersenne prime (2^n-1) number of judges to make a decision
If one of the party disagrees, they can increase the bond and call a higher instance that increases the number of judges.
The judges of each instance will be rewarded/punished by comparing their decision with the instance in which the trial ended. So the rule of each level of judges is not only to follow what they believe but also what they think other judges will think, so there's an incentive for them to look at past judgements and a system of precedents. If their decision was overturned then they'll be punished (by loss of reputation, payment or some other way) in accordance to how much their group's proportion was out of bounds.
If an issue is very controversial, or if one of the parties has a strong incentive to fight, the number will increase so much that it would be more similar to a jury system, and then finally a popular vote. Since larger votes also set precedents, there is an incentive for third parties to work as "rights group" and help pay the bond of what they believe to be an unjust trial.
The ultimate vote would be a set of judges so large that it becomes effectively a popular vote. Since this is the last appeal, then each voter can vote according to what they believe should be the new precedent. A large vote means that if the population (at least the population's) belief and attitude towards that particular matter can change and contradict past judgements.

Uncertainties

The system depends a lot on how voting is decided or weighted, as it only works if voters are a diverse set of people that have a stake in the future of Aragon. It might also fail if ultimate voters are swayed by external events or can be bribed via smart contract

Bribing is going to be a challenge indeed. I wonder if it would work to use the same court system to open a case against a judge when bribing evidence is found.

Also, can't find what is special about mersenne primes for this kind of system. Wouldn't any odd number work?

Nothing special about mersennes, it's just a nice way to create large odd numbers. Prime can be even better if the decision is not primary

This seems in line with the judicial/arbitration system talked about in the Aragon white paper which essentially creates a game where judges try to pick the consensus option among a random subset of judges, in the whitepaper the number of judges is fixed for various levels of appeals and starts at 5. With this proposal you do not have a fixed number of judges, but the number of judges scales with the bond that the parties contribute. With more judges, it is more difficult to provide significant incentive to bribe them and risk taking a reputation hit.

The question I have is if the bond is associated with getting a higher number of judges to review the case, when does it make more sense to increase the bond versus creating a smart contract that rewards any judge that can prove they voted in favor of the briber. I think the best strategy for someone in a dispute is to not increase the bond or number of judges, but always try and bribe them.

Also if this mechanism allows bonds that are not equal on both sides and the bond for the losing side is given to judges, it actually creates the reverse incentive for judges to vote against the person placing the higher bond.

Those issues seem to lead me to think that a more static appeal process makes more sense, but the issue of bribing judges is still present. I'm not sure if there is a feasible way for the reputation system to take into account judges potentially taking bribes (how could this be proven if bribes happen externally?), perhaps this is integrated into the appeals process so that if judges take bribes at the earliest level and its subjectively obvious they are likely to be punished for it in an appeal?

I would say, instinctively that there is a correlation between the amount of stake the judge has on the Aragon system and the likelyhood that he is to accept a bribe. If a judge is part of a panel of 5 judges on a case about a USD$10k deal, a judge that has USD$50k on Aragon tokens is unlikely to be bribed: it makes no sense for one of the parties to offer more than the value of the deal and if Aragon is seen as an unfair system then the judge will lose more than that. I remember reading about stake-betting on particular cases, but are there requirements for long term locking of tokens? Maybe a judge can only be a part of a panel if they have an order of magnitude more Aragon tokens at stake (locked in general, not tied to any particular judgement) than what the judgement is about.

Of course, there's a scenario for irrational actors or even externalities that make the value hard to measure (maybe there's a $1000 dollars at stake but one of the parties would be willing to burn $100,000 just to hurt the other party)

On Wed, Aug 2, 2017 at 5:16 PM, Luke Duncan notifications@github.com wrote:

This seems in line with the judicial/arbitration system talked about in the Aragon white paper which essentially creates a game where judges try to pick the consensus option among a random subset of judges, in the whitepaper the number of judges is fixed for various levels of appeals and starts at 5. With this proposal you do not have a fixed number of judges, but the number of judges scales with the bond that the parties contribute. With more judges, it is more difficult to provide significant incentive to bribe them and risk taking a reputation hit.

The question I have is if the bond is associated with getting a higher number of judges to review the case, when does it make more sense to increase the bond versus creating a smart contract that rewards any judge that can prove they voted in favor of the briber. I think the best strategy for someone in a dispute is to not increase the bond or number of judges, but always try and bribe them.

Also if this mechanism allows bonds that are not equal on both sides and the bond for the losing side is given to judges, it actually creates the reverse incentive for judges to vote against the person placing the higher bond.

Those issues seem to lead me to think that a more static appeal process makes more sense, but the issue of bribing judges is still present. I'm not sure if there is a feasible way for the reputation system to take into account judges potentially taking bribes (how could this be proven if bribes happen externally?), perhaps this is integrated into the appeals process so that if judges take bribes at the earliest level and its subjectively obvious they are likely to be punished for it in an appeal?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/aragon/governance/issues/5#issuecomment-319785474, or mute the thread https://github.com/notifications/unsubscribe-auth/AAG5ApinYpQMLVICUfRAce6SNgLWAp-pks5sUNkcgaJpZM4Ob-hw .

-- Alex Van de Sande Lead Designer Ethereum Foundation

I would say, instinctively that there is a correlation between the amount of stake the judge has on the Aragon system and the likelyhood that he is to accept a bribe

I think this is right, but I'm not sure how strong the correlation would be. The expected impact on the overall price of a single judgment on a 10k deal is likely negligible, and if it's not negligible it means that the price of ANT would be highly volatile (as the action of a single judge in a 10K judgment has a large impact on the overall price of ANT). What I think is likely is that each instance of misbehavior has little effect on ANT, but if there is consistent misjudgment it would have a significant cumulative effect. This is sort of a tragedy of the commons issue, as it's in every judge's best interest to take a bribe, but the cost of taking a bribe is distributed across the entire network.

I agree that long term locking of tokens could help resolve the issue, but that would have to be balanced with the need to make participating as a judge attractive. If judges face liquidity risk in general, they much also be compensated significantly more in order to ensure we have enough judges participating.

aragon / governance