There’s definitely our fair share of disagreements I had with @paulaberman in the past around governance issues. Some may remember that we held different views around UBI DAO governance for example. We are not accountable for other people’s views @Koki.
I don’t think I’m avoiding any question. I believe that we need to protect our DAO against power centralization. Period. My actions are guided in a direction that aims, hopefully, to put Proof of Humanity over my own personal benefit.
You might agree or disagree with my tactics. That’s fine. I hope we can still agree in the common goals. I sincerely appreciate how you connected prode.eth to ubi.eth. Let’s keep working on the buidling.
I’ve asked for your view, not Paula’s or others view.
So far, you have pointed out only one disadvantage of the quadratic delegation: requires education. I reasked about the disadvantages because I don’t trust in magic solutions. There is always risks or drawbacks when you make changes in complex systems, like in the governance of our DAO. And I’m here asking for information because in my own research I could found any paper about quadratic delegation, only about quadratic vote.
So I will re ask again so all the voters have the complete information about the implications of the change:
According to your own opinion, @santisiri and I invite to the other authors of this HIP too, what are the disadvantages/drawbacks of implementing quadratic delegation?
I think many have exposed better than me the drawbacks @Koki. But I tend to value in a positive light that we move forward with mechanisms than aim to incentivize more direct participation; delegating to those with less power rather than concentrating in those with most; and limiting the effect of potential malicious clientelism and farmed delegations. Those are existential risks that face a hard limit with a quadratic formula on delegations.
No one is claiming there’s magic here. Of course there are always trade-offs. I simply hold a positive view about this step and a healthy evolution for our DAO to remain democratic and not corporative.
1p1v is the base of decentralisation. We should not abandon this for whatever reason. There are many more other solutions we could explore but as the baseline 1p1v should remain.
I advise everyone to read this article as it has a lot of insights on the current situation.
Reducing the impact of delegations will not lead to more decentralization, it will be the opposite, it would lead to a centralization of power in the hand of people who can handle the logistic (splitting between addresses, presigning and copy voting programs).
This will suck all the time of participants who will be able to get higher voting power, not because they can convince more people, but become they are better organized. This may even lead to the birth of the equivalent of political parties.
Even worse, reducing the vote impact of delegations will put the protocol at risk of attacks.
There is around 100k$ in the DAO treasury up for grab by bribers and farmers.
Those attacks are not theoretical:
There has been an attempt to bribe POH voter into giving them shares of the treasury. You can see that all delegated votes went into protecting the project while the briber managed to get 56 votes (while the payout was conditional on the vote passing, now imagine unconditional bribes).
I’m just coming back from ETHBerlin and the “winning” project basically just bribed people into voting for it.
You said it right there. Handling the logistic is expensive, so it creates an added cost. With the current system, the logistic cost of a delegation whale is zero, you just create pre-delegated wallets (let’s say, about 200) and then you start registering people with this delegation kit. After that, you forget about ever reaching those people to vote ever again. There’s a huge risk of a single person with the finger on the button “Vote”.
The disadvantages I see is that there is an extra step in what was earlier a simpler process, the difficulty of people to grasp what non-rounded numbers in voting power mean, and that people might delegate less if they find it hard someone to delegate to that has not already a large amount of votes. These are trivial in comparison with the rest of the advantages.
With Quadratic Delegations there are incentives for many people to rally and organize around the concerns that are a priority to them. We empower more minorities rather than keep feeding delegations to ourselves. I don’t know about you Clement, but I want to decentralize power from myself in this DAO.
The signatures that Snapshot takes requires the input of the content available in the proposal, the space ID and the voter’s choice. There not much that can be achieved with “presigning” and copy-voting… unless you assume people are somehow incredibly manipulable and you can hipnotize them at a distance the second they use their Metamask.
I don’t believe people should be regarded as tools in our ecosystem. The votes cannot be copied with an alternative delegator solution since all delegations are talliend on an on-chain smart contract.
Delegations don’t disappear, we will have more people empowered to protect the protocol in the same way we did there. Which by the way, that vote was also won on the direct vote count.
Bribing and collusion are a problem that we can definitely sort out the day we begin implementing Secret Voting… Sismo’s ZK Badges are a great step forward and there are already 82 ZK PoH accounts that we can pilot a secret vote with.
QD as is proposed here greatly disincentivizes that.
Minorities that don’t have the time or resources to coordinate voting directly effectively (and these are the minorities we want to protect, probably) will be obliterated, instead of delegating and trust a minority leader with less effort. Minority leaders will have their voting power quadratically reduced.
When you turn the switch on, immediately 400 votes that are delegated will now require an extra effort of coordination to be counted.
Moving forward, even if we incentivize everyone to delegate when onboarding, it will still be a lot harder to fight attacks from farmers/bribers than with the current delegation system, since the delegated VP will be quadratically reduced and you need more individual votes to defend.
There are vocal minorities, and non-vocal minorities. Non-vocal minorities with low resources to coordinate will be obliterated and it will be very hard for them to have an impact.
An actual farmer/attacker will not delegate and keep people’s key to have exponentially more voting power than honest delegators.
Non vocal minorities will be obliterated long terms if they are small and not vocal, Shin, I do not know where you are going with that. Neither policy protects them.
Logistic (like spinning out a new address for each person who want to delegate) is not engagement.
I think this remark is extremely interesting.
Let’s look at the basic state of governance, the one we have without democracy:
Without democracy, there is still a struggle for power, but this struggle for power is not based on votes, but on military actions (from full fledged armies fighting each others to sneakier form like coups killing/capturing key leaders).
This is extremely wasteful, you have to buy arms, coordinate, plot, etc to take power. And even worse: people die.
Democracy in some ways approximates the military struggle for power, but does so in a less wasteful manner:
It approximates it. As assuming that all sides are similarly coordinated and humans of similar military strength, the same side would win.
It does so in a less wasteful manner, no need to have some many military personnel and no need of life costing battles. Those are simulated in the urns.
The approximation is however unperfect, in the real world some people are stronger/richer than others, some more willing to take risks to govern, some people are more coordinated and there are countless cases where a strong minority manages to impose its will over the majority.
Here the point of Luis is that in the new system, people who are “more vocal” (i.e. spending more resources in governance), will get an advantage.
This shifts more the governance toward the militaristic end (note that I don’t mean the governance will be mainly militaristic, but currently we may be at 95% democratic - 5% militaristic and it would shift it to 75% democratic - 25% militaristic).
Indeed, people coordinating to split delegations between addresses would get an enormous advantage over those who don’t.
If this proposal passes we can play this game. Actually I even think I’ll be better playing this game than the people supporting this proposal.
However I prefer not to have to waste resources and stay in a democratic system.
Indeed, it is interesting to compare quadratic delegation with quadratic voting as it is normally proposed. To recall, quadratic voting is typically relevant in elections that can produce multiple outcomes, such as choosing a budget where multiple budget items can be accepted. Then, participants are encouraged to split their vote credits over multiple options as each option they vote for receives a number of votes equal to the square root of the credits they use for that option, so concentrating their credits on a single option leads to diminishing returns in how much votes that option receives. This has the effect of favoring outcomes that have a broad base of community support rather than just a core of intense supporters that would want to give all of their credits to a single outcome, despite the diminishing returns.
While the goals of this quadratic delegation proposal aren’t necessarily quite the same as those of quadratic voting, it is nevertheless somewhat ironic that it seems to have an opposite effect. You would expect that the people who make the effort to vote individually on a given issue are disproportionately going to be the people with the most intense opinions. So by reducing the weight of delegated votes, outcomes that are supported by a core of intense supports seem like they would be more likely to be adopted.
If persons that delegates with QD finds themselves extremely compelled to vote because they are passionate about that particular subject, their vote is not reduced.
If a voter couldn’t care less about anything their vote will be reduced.
The quadratic delegation is as representative of the group sentiment towards a subject as quadratic voting is.
Even though his is already at Phase 3, we could revisit it soon to go for a better formula.
We’ve been having some actual constructive conversations on twitter about other ways to mitigate delegators power, while trying to supress voters ast least as possible.
Turns out quadratic might not be the most optimal, but we still need something different than the current linear:
Enter: The exponential fraction. I rememer @ludovico at some point mentioning that other formulas less penalizing tha quadratic can be used, and maybe this is a good case for them.
I’m hoping that soon, a new HIP would modify quadratic for something that eases both sides concerns.
I guess these posts were prior to our twitter debate. I wouldn’t advise this kind of improvisation. Breaks credible neutrality; goes against the lindy effect of quadratic formula; does not take into consideration the academic literature and ultimately its making something even way harder for voters to understand. Those are some of the arguments exposed about this on Twitter today. The difference graphed is also far from being a relevant change. There’s a reason square roots are preferred for equilibrium. Please read: Penrose square root law - Wikipedia
Also, you don’t want constant changing rules with a parameter that can be tinkered with. You want predictability. On this matter it’s wise to stand in the shoulders of giants that have actively worked on the mathematics of this like Glen Weyl and others.
Quadratic works with simplicity and elegance and does the job we expect it to in order to empower minorities over bipartisanship.
The [1409.0264] Nash Equilbria for Quadratic Voting paper takes the perspective of participants being able to buy a given number of votes v by paying a cost of v^2. Then, as I understand, they see that in equilibrium the number of votes participants are willing to buy is roughly proportional to the utility they get from the outcome passing. (So, in that equilibrium you get engagement from participants whether get a lot or a little utility from the outcome.)
The closest parallel for quadratic delegation seems to me to be to take the delegates as the participants, who can spend effort to recruit people to delegate to them. So under quadratic delegation they get a number of votes that scales as the square root of the effort they make. You have to make some approximations for this to fit into that model. (For example, the number of participants is not fixed, and humans on PoH have the agency to vote themselves. Also an important assumption in all of these quadratic models is a lack of coordination between the participants which gets to comments people have already made about this system benefiting groups that can organize. Finally, this assumes that the effort spent by a delegate to recruit new delegees is proportional to the number of delegees, where in reality I could imagine there being economies of scale.) However, modulo those approximations, the theorems would seem to say that quadratic delegation will lead to an equilibrium where all of the delegates are incentivized to find people to delegate to them in proportion to the utility they get from a typical proposal. Is this how you are thinking about this/modeling participants, or is there some other model that is more appropriate?
An optimistic interpretation of that would be that a wider variety of participants will be incentivized to campaign for delegations. A less optimistic interpretation would be that the strategy of farming for new members who delegate to you now seems to be generally incentivized. The perspective of people paying to get a number of votes is natural when you are using the money raised to pay for public goods/something that benefits the community. Depending on how optimistic one is about the efforts people make being used as part of healthy campaigning or recruiting people for PoH who will be engaged versus farming delegations from people that are not engaged, the “costs” people would be incentivized to expend here may or may be beneficial for the community.
That paper, and several others that I referred to in debates throughout this week, were mentioned by me regarding why the quadratic formula (and not other variations of the squared root parameter) makes sense. That’s the paper I specifically remember Glen himself mentioning to me on why quadratic and not “cubic” or any other version of the formula makes sense. You can track down the theory on this as far back to 1946:
Penrose (1946) observed that the, if voters act randomly, the power of an individual with m votes relative to that of an individual with a single vote grows as m2
Which I think it’s a very nice fundamental that helps explain why squaring works so well.
Farming delegations (a. k. a. clientelism) is incredibly cheap —way cheaper than farming profiles— and is ultimately an action that rewards voters who take advantage of lazy participants unaware of the issues being debated in our DAO or even completely ignorant about what it means to delegate. This kind of attack was executed in the past few weeks and the fact that we move forward with a governance technique that rewards engaged users able to organize more than lazy participants (which they can still delegate if they want to btw), is by far a net positive for the democratic ideals we hold ourselves accountable to.
As you correctly point out: this leads to having more smaller delegates in the long run that can become activists around minority causes rather than an oligarchy or the very real threat of a monopoly of super-delegates that can become demagogues (a label often applied to critique myself) or authoritarians (our friend in common).
The very tenet of Proof of Humanity is sybil resistance. Farming profiles is not that easy to execute but as we have seen in the past, we have indeed seen sock puppet attacks going after the UBI or even 20 profiles all of them voting with a few seconds’ difference on Snapshot, rendering those profiles suspicious. The incentives for farm profiles are not strictly related to this proposal and are at the core of the very nature of what sybil resistance aims to solve. So I would definitely recommend a wiser approach to effectively separate the issues that we debate; on one hand those related to our protocol and on the other those around governance. These should be separate threads of thought and debate, that of course can always inform each other.
Thanks for your input @william, always a pleasure.
PS: Do keep in mind I come from a country where clientelism has completely hijacked our democracy.
Engagement 