Quadratic Voting for Polkadot Governance

With OpenGov landing on Polkadot in a few days I’d like to take a step back and see where we actually want to go as an ecosystem. I’d like to make the case that OpenGov is a plutocracy and doesn’t fit the requirements of a desirable governance model for web3. I’d like to argue why quadratic voting would provide a healthier middle ground on the spectrum between plutocracy and one-person-one-vote democracy.

Governance should derive its authority from legitimate sources. PoS and token voting as in OpenGov are legitimate in the sense that those who have more to lose have more to say. In an unbounded and pure market the capital will accumulate at very few individuals. A global one person one vote democracy, on the other hand, would not be legitimate for Polkadot because it would give people a vote who don’t care about Polkadot, are not affected by its decisions and have nothing at stake.

Governance should encourage the active participation and inclusion of all individuals and groups affected by its decisions. This is where OpenGov will fail. On Kusama it became obvious that referenda are decided by very few individuals and that an assumed 95% of all votes casted are absolutely irrelevant (not knowing how many humans are behind those 95% - nor the other 5%). This will lead to resignation and a lack of representation of many people who are affected by governance decisions. Vote delegation and conviction voting may help ease the plutocracy a little bit. Delegation helps with meritocratic representation and conviction voting may help protect the interests of minorities. But I don’t believe that we can be satisfied with this.

Quadratic voting applies the square root to the weight of committed tokens used for voting per account and therefore reduces the effect of capital vs. personhood while still giving capital a strong (enough) voice. By adjusting the exponent, one could even to some extent tune the balance between one-person-one-vote and one token one vote.

Quadratic voting, however, can only work in sybil-resilient system. The following simple example shall show why this is the case (for simplicity, conviction is 1x):

Uncle Scrooge votes Aye with 25 DOT
4 other people vote Nay with 4 DOT

  • In token voting, it’s 25 vs 16, so it is an Aye
  • In one-person-one-vote it’s a 1 to 4, so it’s a Nay
  • In quadratic voting its sqrt(25) to 4 * sqrt(4) = 5 to 8, so it’s a Nay

Now, Uncle Scrooge is clever and distributes his funds evenly on 25 accounts

  • now it’s 25 * sqrt(1) vs 4 * sqrt(4) = 25 to 8, so it’s Aye

So, a sybil attack can change the outcome, given the same capital distribution

How can Polkadot governance become sybil-resilient?

  1. State issued ID’s could be used for KYC (i.e. as recently announced by KILT&deloitte)
  2. Biometric Identity based (I’m not gonna go there and argue elsewhere why this is the least desirable option)
  3. Social Graph based (i.e. BrightID)
  4. Pseudonym Parties (in-person, like Encointer or virtual)

For more details I recommend to read this amazing paper by Bryan Ford: Identity and Personhood in Digital Democracy: Evaluating Inclusion, Equality, Security, and Privacy in Pseudonym Parties and Other Proofs of Personhood

9 Likes

At a high level, I am in pretty strong alignment with what you stated here @brenzi.

First, I want to call out that calling the current Polkadot governance system a “plutocracy” is not inaccurate, but perhaps too negative of a connotation. The simple truth is, as you mentioned, that we do not have an identity system on Polkadot (yet) that would facilitate systems like quadratic voting. So given everyone can be an anonymous and potentially malicious actor, the only voting system that makes sense is one which incentivizes token holders to make decisions which are beneficial to making their token value grow. Yes, if you have 1M DOT, you will have a significant weight on the outcome of votes, but you are also proportionately invested in the success of the network. This is exactly the kinds of systems which business are run on, where founders and high level leaders have voting shares and large amounts of stock in their companies, which keeps them invested in the success of their business.

Polkadot has even gone one step further here with conviction voting, which I haven’t really seen anywhere else. That is, a person can increase the power of their vote by locking their tokens for a longer period of time than a normal vote. So actually, a group of smaller passionate voters can have much more weight than an impassionate whale.

There really are no deep flaws with the current governance system from what I can see.

However, there are certainly a set of world problems that can be solved with alternative voting systems.

For example, if you are like me, you might believe that national election systems today are flawed. Especially in the USA where gerrymandering and other voting schemes attempt to twist the outcome of elections. I do believe that blockchain technology could act as one part of a solution to these problems. I also definitely agree that we should NOT vote for the next president of the USA weighted on how many tokens someone has. This kind of election is perfect for a 1:1 voting system based on a robust web3 identity system.

However, I don’t think it is right to assume all votes on the blockchain should be of one type.

Instead, we should be thinking hard about incentives and outcomes, and pair the voting systems appropriately. For example, I probably think runtime upgrades should remain in the control of the majority token holders. I think some technical decisions like increasing number of parachains should stay in control of a meritocracy like the fellowship and their internal ranking system. I think funding for local Polkadot events could use a quadratic voting system, so that there is increased chances of funds making their way outside of the main financial tech hubs, but also resistant to a simple sibyl attack. If we were to ever help national elections, we should do 1:1 voting.

I think quadratic voting is the right technical direction, but we should be careful about what tools we use to solve what problems.

12 Likes

this is perhaps the line that best sums up the vacuum in pretty much all discussions / arguments about pretty much any topic in this ecosystem. in general most people see the status quo as it is, and then work forward from there, rather than asking if anything, other than the historic state is sacred…

For example crowdloans? They’re an incentive system, they produce an outcome. It delivered good outcomes in one dimension, but is that outcome the one we are now optimising for?

Venture capital creates an incentive system, that drives an outcome - is that the outcome the one we want?

@brenzi you are currently enduring with Encointer the challenge of explaining public / common goods to a bunch of people who really want number go up aka:

When we reduce to these binaries, we can see the heart of the problem.

  1. Public goods - not sexy, slow burn, R&D driven, experimental, loss-making, hard if not impossibe to assess ROI in the short and even medium term, but ultimately the foundation on what everything else is built.

  2. Private goods - 99.9% of the crypto industry, sexy, change the world stuff, dopamine hits, short term wins, startup culture, influencers, overnight fame, celebrity - generally this is the tip of the iceberg though, its just most people are not exposed to the work beneath the surface that made that overnight success work.

Public goods are upstream from free markets. It doesn’t work the other way around.

People have short memories - Bitcoin birthed this novel public infrastructure after decades of R&D by many unknown individuals and groups with no funding, and motivation being curiosity about a problem… intrinsic rather than extrinsic motivating factors.

The other dimension we can consider related to this is simply about understanding that what we have here is a continuum - public goods —> money making products.

Rather than flip flopping from voting system to voting system in search of the one ring to rule them all, we should develop an fully on-chain innovation process that covers the spectrum.

  • The treasury is good for some stuff, but not everything. That’s ok.
  • The burn is good for some stuff, but not everything. That’s ok.
  • Delegation works in some areas, but not others (it opens up lots of legal issues).

You can work through pretty much anything this way, with both primary and second order effects.

Lets get stuck right into the weeds, what kind of process do we want?

What does the ideal journey look like over a five year period (or even a decade) for an idea, a team and an initative?

The truth is, fostering an sustaining innovation culture is a subtle, and delicate thing.

Right now we have no such process, but we can figure one out together.

We just need to be clear what incentives we are working with, and what outcomes we want at each stage.

This will help square the circle of the competing demands of public and private incentives.

I’ll echo that I’m very much in agreement with Shawn on this. One-person-one-vote systems (loosely-termed “democratic”) that include the voice of “everyone” (define that?) equally are not typically associated with arriving at the best outcome, however that be measured (economic, technological, …).

However such systems have two major advantages over stake or merit-based voting schemes:

  1. All individuals feel equally heard and included. This mitigates the possibility of “revolt” in the case of a controversial decision.
  2. Enables and empowers a lot of “minor actors” in decisions, thus massively reduces the possibility of the targeting of individuals by malicious actors.

I can certainly imagine a number of governance primitives which could be adapted with the arrival of a pervasive, strong and private Sybil resistance scheme to better ensure the resilience of the network and realise a number of important services much closer to their ideals.

Quadratic voting may be an element of this, but I see it by no means as the most exciting thing should the problem of Sybil-resistance be cracked.

10 Likes

In this video, I delve into the concept of quadratic voting and its implications. I discuss how capitalism naturally leads to inequality and how quadratic voting aims to address this issue. However, I highlight the potential drawbacks of quadratic voting, such as the shrinking of the pie and the difficulty in preventing Sybil attacks. Join me as I analyze the pros and cons of this voting mechanism and its impact on governance.

1 Like

Thank you @giotto for your reflections on the topic. I’ll attempt to summarize your arguments in text form to reply one by one

[@giotto paraphrased by @brenzi] The size of the pie matters because it is what secures the blockchain

Agreed. We both aim for a good market cap of DOT. Quadratic voting is not a means for redistribution, therefore it’s really not about “communism vs capitalism” on my POV.

[@giotto paraphrased by @brenzi] Quadratic voting gives more power to small accounts and less power to big accounts. So, why would anyone buy a large amount of DOT? […] People will “vote with their feet” and sell their DOT […] Why would they take a risk which is higher than the representation they get?

I don’t believe that people buy DOT because it gives them representation in DOT governance. They buy it because they expect a return on investment.

Let me explain why QV is in the interest of every DOT holder, not only the small ones:

Your hypothesis seems to be that the outcome of QV voting will be systematically worse for whales than linear token voting. I disagree. As you explained nicely, people “vote with their feet” as well. This means that small DOT holder will be very careful about scaring the whales away.

I can give you multiple real world examples from Swiss democracy which is commonly understood as one of the world’s most direct democracies.

The Swiss citizens voted against an inheritance tax, although >95% of the voters would be net beneficiaries of such a tax.

  • a very plausible interpretation is that the voters were afraid that many wealthy citizens may “vote with their feet” and leave the country

The Swiss citizens voted against increasing mandatory holidays from 4 to 6 weeks

  • a very plausible interpretation is that businesses may “vote with their feet” and leave the country, or SME’s may earn less revenue, pay less taxes and offer less employment in consequence

Therefore, I think you’re hypothesis is wrong. Voters are not stupid. Or at least their stupidity doesn’t necessarily correlate reciprocally with wealth.

Do not forget that 1 DOT may be the same “skin in the game” for one person as 1 MDOT for another, if you measure the percentage their DOT holding represents in their overall wealth. Why would retail investors join Polkadot if they feel ripped off and ruled by early adopter whales?

Quite to the contrary of your POV, I believe it makes Polkadot much more resilient if more people feel represented in its governance. Not only because of the “wisdom of the crowd”, but also because a “sense of belonging” can be so much stronger and more lasting than pure economic (greed) incentives.

Then comes your very valid argument around corruption:

[@giotto paraphrased by @brenzi] Sybil-resilience doesn’t solve the problem becasue there is no way to prove a voter is voting as a free agent. Therefore, whales may be incentivized to buy votes.

First of all: vote buying/bribing is highly incentivized in today’s plutocratic OpenGov. So let’s measure governance designs with equal measures.
Like you, I’m very concerned with this and I don’t have an encompassing solution. In another post I have explored the opportunities of private voting, which may give us a hint on a solution: If voters have no way to unambiguously prove how they voted, vote buying is futile.

6 Likes

Answers:

1 Like

So you’re letting me play your secretary again and condense your 40min talk into a couple of lines of text? You’re bringing forward some very prototypical libertarian arguments which I’d like to contemplate, so I’ll be your keyboard once again for the points I deem relevant. If you fancy continuing this lovely discussion, I’d be in favor of text.

[@giotto paraphrased by @brenzi] Linear is the natural state. Therefore, everything else needs to argue why the added complexity is merited due to advantages.

Many things in physics and psychology are more meaningfully analyzed on a logarithmic scale (like loudness and sound pressure levels are measured in dB, so is light intensity)
Arguing with “the natural state” has been used to defend many horrible events in human history, we should be careful with that - even If I’m not implying bad intentions in your case. Societal coordination is a choice. Civilization is defined exactly by its difference from pure evolutionary natural forces. Killing another human is considered unethical by humans only. The rest of nature doesn’t care.

[@giotto paraphrased by @brenzi] Reducing voting power is a factor decreasing the value of the token, even if it is not the main driver for buying (or not selling) the token

I can agree to this beeing one factor influencing token price. And, as you say, other factors may offset that influence

[@giotto paraphrased by @brenzi] Maybe you could do an experiment where a (non transferrable but delegatable?) derivative of DOT only has voting power and see if it will have value because people will pay to get your tokens delegation

This is a flawed experiment IMO. You will not measure what you try to explore.

[@brenzi] This means that small DOT holder will be very careful about scaring the whales away.

[@giotto paraphrased by @brenzi] The whales also have to be very careful not to scare the small holders away, therefore your argumentation is flawed

Yes, of course whales don’t want to scare small investors away. But that’s not changing the status quo, is it? Currently, whales need retail investors to bring in their return on investment. Hopefully, we will talk about “users” soon, not retail investors. No matter if we vote linear or quadratic. Your counterargument does noting to invalidate my argument

[@giotto paraphrased by @brenzi] Countries have 1p1v and companies have one share one vote because that makes the most sense. Tokens are probably more similar to companies than to countries.

Polkadot is neither a country not a company. Polkadot can make a new choice how it wants to be governed. It seems we have different ambitions for web3. If you reduce web3 to a for-profit company, I’ll get bored out. To me, web3, it is a potential tool for optimizing human coordination on a global scale. If you are happy with the former, we can just let our disagreement stand as it is. If you share the latter, then we really need to find a way to make many people feel represented. Capital has proven to accumulate at the few, so it is a very sub-optimal primitive for governance of an important societal fabric like “my vision” for web3.

[@giotto paraphrased by @brenzi] Democracy in Switzerland is just the simplest possible thing

So that’s why it took that long to be established and why it is that hard to defend?

[@brenzi] Voters are not stupid. Or at least their stupidity doesn’t necessarily correlate reciprocally with wealth.

[@giotto paraphrased by @brenzi] You would have to prove that stupidity is correlated with wealth if you argue that the wealthy should have less voting power. If intelligence is equally distributed, why not let the wealthy vote as a smaller sample of everyone involved? Probably, people with large amounts of money tend to be smarter, because if they managed to accumulate wealth and not lose it, even if 95% of the influence may be luck. The can likely spend more time evaluating referenda and accumulate more domain-specific knowledge

Just to clarify: In QV, the wealthy still have more voting power than the less wealthy, just not linearly.

But here we’re getting to the core!

Intelligent votes are not the only goal of governance. Governance by intelligence and knowledge is a technocracy and there would certainly be better ways to assess intelligence than wealth. Here, again, it depends on what your vision is for web3. My goal is to use web3 to coordinate humanity in better ways than we do today, and I’ll leave “better” undefined here not to go off-topic. Still, the world I want to live in lets all humans participate to a certain degree and feel self-actualization and being enfranchised. If people don’t feel represented, they don’t feel responsible and won’t act in the best interest of the whole. They also may revolt against the elite, eventually. In web3 terms, that would probably be called a “rage-exit”. So here you get your argument why whales should probably support QV out of self-interest.

[@giotto paraphrased by @brenzi] How can a whale rip off a retail investor? This is impossible

Well there is plenty of evidence for market manipulation by whales, but that seems off-topic to me.

[@giotto paraphrased by @brenzi] companies are run much better than governments

Companies solve a significantly simpler problem than countries and they need not care about externalities unless there is a state enforcing rules. You’re comparing rocket science to building a house.

[@brenzi] vote buying/bribing is highly incentivized in today’s plutocratic OpenGov

[@giotto paraphrased by @brenzi] This is not true, because power is proportional to ownership, so power is aligned with what you could lose. QV is much more likely to fall victim to corruption because voting power is less aligned with ownership

Today, whales can easily collude with a (so far) reputable project for a treasury spend referendum and the whale would not risk to lose their tokens and not even depreciation of the token - as long as the corruption goes undetected. They would just be a parasite on the treasuries funds to the loss of all the other stakeholders.

I still don’t see why corruption is more likely to happen or being more harmful in QV

3 Likes

Feedback:

2 Likes

Thank you @giotto, I have listened to your entire post. You certainly make good points and I respect your position. Mine is different, most likely because our ambitions are not the same. Therefore, I currently have nothing new to add without risking defocusing this thread or repeating myself.

1 Like

I totally agree with Gav here.

Quadratic voting may be an element of this, but I see it by no means as the most exciting thing should the problem of Sybil-resistance be cracked.
QDV can only be an accepted solution IF you can ensure that on’t lead to multiple accounts creation.

Because if you can’t prevent it, then you’re back to the initial issue.

Currently, any solution leading to multiple accounts creation may just be ignored imho.

1 Like

We have sparked a discussion on X to collect more opinions and voice on this topic.

Generally, thanks so much for opening this as i believe Quadratic Voting has potential, but it truly depends on a case.

For Donation-based models i think it works fairly well as Gitcoin has already demonstrated, but can it work for Ttreasury-based models?

Anyway…

Here’s a recap on X, feel free to engage :point_down:t3:

PS: We are planning twitter spaces, would love to get you @giotto @shawntabrizi @gavofyork and @brenzi to have a moderated discussion on the topic of quadratic voting and sybil resistence in OpenGov

I already tried to reach you, but in just in case it didn’t get to you, here’s my telegram or matrix.

3 Likes

@giotto Do you believe we can have a midpoint where voting is not a function of (conviction weighted) balance but

balance^x

where x is some number like 0.9?

I think this is a decent compromise between the linear model (x=1) and the “one address one vote” (x=0), if you keep the existential deposit annoyance.

I would love a mechanism where new x is decided by the community using the previous new x. From your videos, I’m guessing you would keep x at 1 (no?) and then a lot of @brenzi type people would be at 0, and then we’d end up something like 0.9 with this mechanism which is a decent solution that keeps both sorts of people slightly unhappy. You’ll still find people leaving the country, but enough x=0s hanging around the x=1 to achieve decent outcomes.

We want Sybil resistance for many reasons, including byzantine assumptions. In aggregate, these outweigh any benefit of quadratic voting, both for polkadot now, but even more for distance future plans. A mixnet really wants this information, for example.

If anything, I’d suggest “slightly the opposite” of quadratic voting: You rewards “plutarchy” in very small ways designed to convince the “plutarchs” to reveal themselves.

We could then exploit this information about which dots were controlled by “plutarchs”, maybe somewhat in discussions on on-chain votes, but even in protocols without financial implications. As an example, if we’ve multiple relay chains then you split each “plutarchs” validators up evenly among the relay chains. We should not require this per se, but if we did it then we’d have greater confidence.

Could you explain this further detail @burdges?

16DG, Ivy, ChaosDAO are quite active in governance feedback. I doubt new Substrate-based relay chains would cannibalize each other so much – instead they will reduce activity on non-Substrate chains (especially Ethereum L2s), so I think there is a really important idea to derive from what you’re getting at. Please explain?

Thank you for this important discussion on a key topic.

The decision to introduce QV for Polkadot Governance really comes back to one question: What does Polkadot want to be? As @brenzi said:

The original goal of the Internet was to establish a peer-to-peer infrastructure for sharing information. This vision aimed to create an interconnected network for seamless information exchange without a central authority. However, as the internet evolved, it became more centralized, with big tech companies controlling vast amounts of user data.

Today’s desire for a peer-to-peer internet stems from the ethos of decentralization: direct communication without intermediaries. This approach ensures users retain control and ownership of their data, promoting privacy and security.

With the emergence of blockchain technology and decentralized platforms, there’s a renewed push to return to this foundational idea. The aim is to reshape the internet to prioritize user control, open communication, and democratized access to information.

It is clear as day: We can not use the same primitives that enabled Big Techs exploitation, for the foundational layer of a system that is supposed to be about decentralization.

To achieve our vision of a decentralized web, we must boldly innovate in governance. This requires addressing the challenge of sybil-resilience and establishing a scalable proof of personhood in order to diversify governance systems.

The question we need to answer is not “Should we introduce QV for Polkadot Governance?” It is “When and how are we introducing it?

1 Like

It’s true “sharding” not “cannibaliz[ing]” anything, and nothing to do with governance. It layers on top of our existing roll up to scale to more paracahins, without increasing the gossip overhead.

It’s first discussed by Bryan Ford’s team in

https://eprint.iacr.org/2017/406

It’s conceptually simple compared to roll ups, but involves extra protocol level security analysis using concentration inequalities. Afaik everyone else screws this up when they design committee systems, which makes me laugh.

Anyways quadratic voting wastes the Sybil resistance “resource” that’s better “spent” scaling.

Can you explain how are the two use cases rivalrous? If we have strong sybil-resistance, why not use it for QV AND your byzantine assumptions for scalability? I believe this discussion might be even more relevant in this thread Happy to add your described use case there

We’ll never have good Sybil resistance if we make being Sybil extremely profitable, or give the impression of it being profitable, ala governance. It’s too easy to pay people.

At this point, quadratic voting people say “yay basic income”, but once they’ve incentivized this payment infrastructure, then they’ve lost Sybil resistance. Aka quadratic voting wrecks Sybil resistance in exchange for a “small basic income for the criminally minded”.

We want the Sybil resistance elsewhere where incentives look much lower for typical users, but which improve security and performance, or in future maybe in privacy ala mixnets.

Example 1. We’ve some minimal number n of validators per relay chain, which we determine by the on-chain randomness quality, the number of relay chains, the honesty assumption ala 80% or 85% or whatever, and optionally who we imagine colluding. We improve this last part if we spread validators evenly across the relay chain. Now n being larger makes sharding secure, but gossip is O(n^2) so n being smaller saves resources, so this should impact the number of parachains per relay chain.

Example 2. Approval checkers tranche zero assignments could be adjusted so that individual operators rarely duplicated their own work, not unlike Example 1. It’s possible this permits fewer approval checkers, which makes polkadot slightly more efficient.

Example 3. A mixnet depends upon users choosing one honest node, out of five to three random choices. If an operator honestly declares its multiple nodes, ala MyFamily in Tor, then users can chose only one, which ups their chances in theory.

Example 4. A referendum passes or looks likely to, but people complain about how Wally the Whale pushes it. This could prompt less involved whales to take action, but they might not do so if Wally set up Sybils beforehand.

All of 1-3 have some small cost for large operators, although only after they make dangerous code modifications. And nominators matter here too, even if they do not incur costs. We could compensate this cost by paying larger operators or nominators very slightly more, or giving them slightly more say in governance.

1 Like

:wave: We think that is a good step to overcome the pop myth of the Internet as a decentralized technology and acknowledge that it is a distributed technology that never had decentralization as a goal.

Please, read the “The Myth of the Decentralised Internet” for just one obvious analysis: The myth of the decentralised internet | Internet Policy Review

Even FidoNet that is more open to participation is far from being a decentralized network or aiming for it.

2 Likes