Court concerns and hard questions

[Xeus]

Let me preface this with the following: I’m a long time Aragon follower. I want Aragon to work. I love the passion of the team, and I’ve been impressed by their capability to deliver well designed and functional software, as this is no small feat. From a technology perspective amazing things have been accomplished.

Now for the critical part: Personally, the participation from ANT holders so far has been a huge let down, and I have a very hard time imagining who would choose this network to uphold just decisions through the Aragon Court. How do others view this? What in your mind makes you confident that the court will be appealing to organizations needing the terms of high-value agreements upheld?

If the Aragon Court system has the participation levels of the AGP that voted to recognise the Network DAO and ANJ contract as official, then there will be 2.5% of ANT participating in the court. With 80% of that coming from 1 address.

Granted, 1 address does not mean 1 person, the keys could be controlled by a whole panel of highly educated and trained lawyers which people will be tripping over themselves to have their agreements adjudicated by. They just all let Bob manage the keys of their secret society of blockchain enthusiast law professors. But personally I don’t think we should rely on this.

Similarly, multiple addresses does not necessarily mean multiple people. It could be yet worse than 80% controlled by a single person.

Who would see value in having their disputes adjudicated by this kind of representation, i.e. 1 anonymous ethereum address? As it stands, this is the Aragon Court’s target market.

Furthermore, I’m interested in participating in this system. I’ve always had a fascination with justice and right and wrong. I tell people if I wasn’t writing software I’d be in law, so that’s why I’m here, at the intersection of the two in a community like Aragon. With ANT in my pocket to convert to ANJ. But as I understand it I face this:

The Aragon Court system is built to be a schelling game where a juror is incentivized to rule in agreement with the other 2 jurors, if they don’t their bond will be slashed and rewarded to the ruling majority of jurors.

As it stands, the other 2 jurors will overwhelmingly most likely both be controlled by a single entity holding the majority of participating ANJ. So as I understand it, if I do participate in this court my game will be anticipating and ruling in agreeance with 1 other entity, which isn’t what I want to be doing. And I have a hard time imagining organisations wanting to opt-in to this little jurisdiction either.

Am I the only one who has these concerns?

Thanks for reading.

[LouisGrx]

Hey @Xeus,

It seems that your concerns are the low engagement of ANT holders, the concentration of ANT holdings among engaged holders and the implications this has on a potential capture of the Aragon Court.

I don’t have in depth knowledge of the Court mechanics and release plan. I’ve been digging into some previous posts and the analysis here is interesting but seems to assume fairly distributed ANJ holdings? It would be cool to have an experts’ take. @lkngtn, @facuspagnuolo, @jorge

[Xeus]

Thanks for the reply and pointing me to that work done previously by Luke. It does look like the model assumed all jurors had the same amount of tokens, each Juror being intialised with an amount of tokens of (token_count / juror_count).

I hacked it a bit to add an additional parameter which specifies the number of tokens the largest juror has, then dividing the rest evenly among the remaining jurors.

To replicate my concerns that worry AGP participation may be a signal of Court participation, I set the Juror count to 10 (there were 7 unique addresses participating in the vote). Total tokens to 100, with 81 held by the largest Juror (82% of ANT in the vote was held by 1 account.)

Here are the results after 230 steps:

After 107 steps there was only 1 Juror with a token balance > 1 that could continue participating.

In the post Luke already brought up an eventual centralization problem and some possible ways to address it. I’m curious where this ended up and if it will be included in the upcoming Court release, as it could potentially be a problem from day 1.

[Xeus]

So, reasoning through it some more on my own, my conclusion is this:

My concern is valid: If the distribution of ANJ is similar to that of the AGP that voted the contracts official, this is bad news bears. For everyone involved.

However: The incentives are being setup to discourage this outcome. Because it’s in everyone’s best interest, including ANT whales, for this to not play out.

How did I reach this conclusion?

The ‘limited-time-only fixed rate of 1 ANT per 100 ANJ’ period.

Initially I thought this would make matters worse, as then a whale can get more ANJ as they don’t have to deal with each ANJ costing more than the last due to the bonding curve.

I’ve since done a 180 on this, and now understand this period to be the best option to discourage centralisation in a non-sybil resistant system like the Aragon Court. I can walk through my thought process more if anyone is interested.

Immediate dilution of stake by minting after activation period:

‘the Aragon Network DAO can mint up to a 10% of the initial supply of ANJ and transfer it to the DAO’ - https://blog.aragon.org/aragon-court-pre-activation/

The requirements of participating in the Aragon Court

The utility of the ANJ token is to rule on disputes and collect rewards through participating in the court system, but there are real-world time demands of this participation. This is not a passive investment which idle capital should be deployed to and sat on. This is something that the modeling above doesn’t capture, as resolving disputes cannot in reality be done programmatically, and thus there is an upper-bound on the amount of participation an individual can have in the system.

The ability to fork

To come back to the original sentiment of the concerns at the beginning, ultimately nobody would want to use the system if it is highly centralised. It’s certainly not in the best interest of the network for the Court to end up in this state, and due to the decoupling of ANT and ANJ, if the Aragon Court were captured by a dysfunctional majority then ANT holders could vote to drop this Court and recognise another as official.

The first 3 discourage a well meaning but profit seeking whale. The last also protects against malicious attacks.

What do people think about my take?

[lkngtn]

@Xeus

What do people think about my take?

Would love to dive into this a write a much longer response but for now I will just say that your concern is definitely valid but as you also point out the court is fairly different incentive wise compared to the AGP process.

It’s worth noting that we may need to tweak some of those incentives to improve things in the future, even if the protocol works reasonably well already.

One tweak in particular that I think is worth exploring further is to adjust the final appeal mechanism to use some sort of automatic forking or futarchy decision function instead of a simple majority decision among jurors.

[burrrata]

@Xeus

Initially I thought this would make matters worse, as then a whale can get more ANJ as they don’t have to deal with each ANJ costing more than the last due to the bonding curve.

I’ve since done a 180 on this, and now understand this period to be the best option to discourage centralisation in a non-sybil resistant system like the Aragon Court. I can walk through my thought process more if anyone is interested.

I’d be interesting in understanding your thought process more

[Xeus]

@lkngtn

Would love to dive into this a write a much longer response but for now I will just say that your concern is definitely valid but as you also point out the court is fairly different incentive wise compared to the AGP process.

Thanks for sanity checking and giving some feedback that I’m on the right track.

@lkngtn

One tweak in particular that I think is worth exploring further is to adjust the final appeal mechanism to use some sort of automatic forking or futarchy decision function instead of a simple majority decision among jurors.

What would automatic forking look like? Is it something like: it’s written into the protocol that the final appeal escalates to a vote for ANT holders, the result of which would either automatically fork the Court or deny the appeal?

That sounds pretty neat to me, especially since I was already thinking that the ability to fork is something that gives additional security to the Court. Having it hard coded into the protocol would make that much more explicit.

I’m not sure about the futarchy option, to pull from the article you linked:

the basic premise [of a prediction market] is that if you have a decision which needs to be made, and you can relate that decision clearly to some objective success metric, you can use prediction markets to pick the option which is predicted to be more positively associated with your success metric

I’m apprehensive about the nature of the ‘objective success metric’ used when dealing with disputes between humans. I mean, have you seen making a murderer? We could set up a market to bet on whether or not that sweet natured country guy has a secret evil streak to him or not, but how will we ever be able to verify what actually happened to Teresa Halbach in order to resolve the market?

I’m guessing the answer would be that of course we can’t verify the truth, if we could it would be trivial to resolve. And in reality, the objective success metric would be tied to something about the performance of the court. I.e. The court isn’t measuring itself against the correctly tracking the truth, that’s impossible. It’s measuring itself against maintaining the market’s confidence of tracking the truth, the measurable property of which being the amount of fees collected by the network, as this is measurable and should be a reflection of the court’s rulings tracking the truth.

I guess the human in me is just irrationally scared of the disconnect between the reality of tracking the truth, and the possibly best-we-can-do alternative of maintaining market confidence in tracking the truth. I.e. Something in me is scared of it devolving into more resources being put into marketing campaigns about how great we are at ruling fairly, rather than spent on ensuring quality and due diligence. Resulting in cases where the truth actually is verifiable, but due to incentive alignment being put on market perception, it’s not verified, and instead whatever outcome is chosen is marketed as correct, because that’s the success metric.

I feel like the applications of prediction markets I’ve seen so far tend to stick to more straightforward verifiable outcomes, like whether your favourite sportsball team won or lost the world concussion championship. Probably for these reasons, it gets complicated.

[Xeus]

@burrrata

I’d be interesting in understanding your thought process more

Sure!

The first was to change my thinking from:

We don’t want anyone to have too much ANJ.

To:

We don’t want anyone to have too much of a percentage of the total ANJ.

It’s an important distinction because the pool of ANJ is not static, and through the pre-activation period will be inflating. Each additional ANJ purchased dilutes the stakes of everyone else.

This means even without a bonding curve there are diminishing returns on purchasing more. To illustrate: Say one started today with a rate of $1 = 1 Token and 10 people bought 1 token each. 100% of the pool was captured for $10. Now a whale comes and buys 10 more tokens. The whale paid the same amount for 50% control as everyone previously paid for 100%. If the whale bought yet another 10 tokens, they’d pay the same amount yet again, but this time only increase their share from 50% to ~66%. They’re getting diminishing returns on purchasing a share of the market, because they’re diluting all existing shares as they do it.

I remember people having the same discussion during the Tezos ICO, as it was highly contentious to have an uncapped ICO. But capping puts a ceiling on the inflation and breaks this, creating an incentive to buy as large a share as quickly as possible before the cap is met.

I’m glad there is no cap being put on this pre-activation period. That’d be a problem.

I also thought through the alternatives that came to mind that might be better:

The first one that came to mind was setting a cap on the amount of ANJ that each individual could purchase before the bonding curve is available. This isn’t possible because the system is open and not Sybil resistant. For example, putting a cap on ANJ acquired per account would be meaningless, accounts are trivially cheap to create.

Having a bonding curve from the beginning rather than allowing a fixed rate period could also potentially decrease the ability to purchase a controlling share. Or it could make it easier. It’d depend on whether the whale purchased before the smaller holders or after. Odds are it’d end up being worse, as a whale can co-ordinate their buy timing much easier than a large amount of small holders, meaning in reality a whale would purchase as soon as possible with a very high gas price and increase the price through the bonding curve, making subsequent smaller holders capture an even smaller share of the court.

So that’s how I ended up at: Yea this pre-activation period is actually a pretty good idea, I can’t think of a better one.