Uneven outcome distribution possibly providing economic incentive for bad actors


Since there is an economic disincentive to start a dispute and be wrong, this means that inherently there should be an outcome distribution where the rulings will favor the disputers favorite.

Let’s say Juror A knows this, and decides they can automate their votes by looking at the favorited decision based on past distribution for general arbitration requests (Let’s use the Proof of Humanity Registry as an example, Juror A samples that 90% of rulings have sided with the actor involved not meeting the Proof of Humanity submissions guidelines, and after analyzing the relationship between between money lost from incorrect voting, money gained from correct voting, and gas fees at the current time, Juror A figures out that blindly choosing this outcome based on past distribution is profitable long term.

Let’s also say Juror B has a lot more PNK to stake, and is participating in the network looking to maximize profitability. Juror B has way too many votes that he can process honestly with the amount of PNK he is staking, and decides to take the same approach by automating his decisions based on past outcome distribution, as the current bottleneck from maximizing returns at this point would be spending the time to read through case documents.


So as a summary, whats stopping this type of behaviour from effecting the network? As participators have more PNK staked, there is a higher incentive for automating this process (the laziest and most common being to look at past distributions), which will have a chain effect on polarizing the outcome distribution even more, which in return create an even higher incentive for this type of behaviour (and the cycle continues…).


I think a key point to call out here is that a juror obviously cant choose the types of disputes he participates in, so there won’t always be an easy distribution to sample (Like predefined contracts such as Proof of Humanity), but a little more work can be done to figure out which dispute resolution option goes against the normal flow of the agreed transaction

I also think there are a lot of game theory questions to answer that I’m not qualified to comment on. It seems that if a growing % adopts a strategy like this (and with the incentive growing for larger stakers there is a non-zero percentage of this happening), it will force a strategy for other jurors to follow suit, as jurors are inclined to follow the same strategy to maximize their returns. Without coordination, the best strategy seems to be to “do what is told and assess the case accurately” as that what you expect others are doing, but if enough people follow a ‘lazy’ strategy like this, it could lead to a consensus where all jurors vote a certain way on dispute regardless of the case data? Just my thoughts, would like to hear any comments on the subject. I’m inclined to believe this is a very rare possible outcome, but was wondering about guarantees of something like this not occurring.

These are definitely issues that one has to think about. Note that one of the constraints in the calculator we have been using to make parameter update proposals specifically relates to this (look under “Lazy Voting Attacks”):

Essentially what the calculator tries to do is to set the deposits high enough so that if someone follows an automated strategy, they will lose money on average even if most of the cases have predictable outcomes. The more predictable the outcome is, the higher the deposits should be set so that this is still the case. The calculator takes in historical rates of which outcomes win the most in each court (when available, when setting up a new court, one more or less has to make an educated guess for this). However, it is certainly true that this is something one wants to stay on top of because, as you point out, if the lazy strategy becomes profitable the situation can snowball to the point where people vote for that outcome expecting the other jurors to be following this strategy.