Algorand Governance vs Vitalik
The Algorand Foundation is decentralising governance, we examine to what extent this lives up to Vitalik Buterin’s ideal.
Introduction
Coin voting is the process by which people holding the native tokens or governance tokens of a given blockchain can vote on governance issues for that blockchain. Usually coin voting is set up in a way such that one token equals one vote.
In 2017 and 2018 Vitalik Buterin, wrote two blog posts about blockchain governance and its decentralisation, the crude takeaway from these being his contentions that coin voting:
- will lead to corruption of the voters (“governors”) causing poor outcomes for the blockchain, and;
- will be unrepresentative of the community.
Then earlier this year he wrote another piece about the challenges blockchain technologies will have to overcome in order to provide privacy and coercion-free voting in democracies.
The Algorand Foundation (“the Foundation”) sees things differently and believes that the benefits of decentralising governance outweigh the negatives. The Algorand blockchain (“Algorand”) is governed by Algorand Inc (“Inc”) and the Foundation with Inc responsible for improvements to the protocol and the Foundation responsible for the 3.2bn Algo token (“Algo”) Algorand Ecosystem Resource Pool (“AERP”). This currently goes towards participation rewards and grants for a range of purposes.
The Foundation will soon begin the process of decentralising and democratising control of the AERP to Algo holders, so if (like me) you are an Algorand enthusiast and want to see where the potential pitfalls of this project lie then read on as we compare these plans to Vitalik’s critique of decentralised governance.
Corruption is Possible if Privacy is not Enforced
“The voting process has four important security requirements that must be met for a vote to be secure: correctness, censorship resistance, privacy and coercion resistance. Blockchains are good at the first two. They are bad at the last two.” — Vitalik 2021
In western democracies, politicians are elected via secret ballot as a measure to reduce corruption. Just as this is needed to prevent people from getting threatened or bribed into voting for a political candidate, privacy and coercion-resistant voting will prevent the same from happening in on-chain voting.
From a practical standpoint what this means is that the protocol used for voting needs to prevent others from discovering what the accounts held by governors voted for, or even if they voted at all, they should also prevent governors from being able to prove what they voted for, even if they wanted to. This allows people to vote based on their personal needs without fear of retribution, and it also prevents people from being able to solicit bribes as those distributing bribes will have no idea who to send them to or whether to trust any claimants of bribery.
Unfortunately, on-chain votes for allocating AERP funding will allow anyone to see the account numbers of governors and most likely what they voted for. Although this doesn’t mean the votes are bound to be corrupt, it does provide an attack vector for dishonest actors.
Why Bribery and Corruption is Bad
Now that we have established that bribery and corruption are (at least in theory) possible for governors to partake in, we need to remind ourselves why they are not ok.
“The second part of this article will involve me, an armchair economist, hopefully convincing you, the reader, that yes, bribery is, in fact, bad.” — Vitalik, 2017
The first reason is that like corrupt politicians buying votes and taking kickbacks from government contracts, the resources don't go into building up the ecosystem as intended, instead, they go into back pockets. In the 2017 blog post Vitalik quotes from the document EOS an Introduction:
“we suggest that the bulk of the value be returned to the community for the common good…Known as Community Benefit Contracts, the mechanism highlights the importance of DPOS as enabling direct on-chain governance by the community”- EOS an Introduction
In 2018 there was a scandal involving vote-buying at EOSIO where some of the 21 supernodes were said to be buying votes and colluding with each other in cartel-like behavior, going against the principles outlined above.
The second reason is that it leads to suboptimal outcomes. I.e. people not voting for the most efficient node, but voting for the node that would pay the biggest kickback or promise their own node votes as part of a cartel.
This would be especially damaging for the governance of the AERP because it would lead to grant money funding kickbacks and the wrong projects being built (if at all).
Why Bribery and Corruption Happens
“ people, at least in their capacity as crypto token holders, are profit maximizers, and seem to see nothing evil or selfish about taking a bribe or two.” — Vitalik 2017
The 2017 article makes short thrift of the idea that people are good and therefore will not take bribes. Furthermore, even if people think they are being honest, experience has shown that it is fairly simple to create a process that looks legitimate, but that is being used to harvest votes, for example with some sort of lending pool.
In reality however it is clear that the Algorand ecosystem is currently too small for corrupt but innocent-looking vote-harvesting operations to go unnoticed because legitimate DeFi options like Yieldy are still so thin on the ground. This means that ceteris paribus as the Ecosystem expands attacks such as the one described above become more likely.
High Voter Participation Makes Corruption Expensive
“One of the main criticisms of coin voting mechanisms so far is that, no matter where they are tried, they tend to have very low voter participation. The DAO Carbonvote only had a voter participation rate of 4.5%.” — Vitalik 2017
In the above quote, Vitalik cites low turnout as a consistent theme in on-chain voting. He cites 2 reasons why if you hold a vote low turnout is undesirable:
- Decisions lack legitimacy because even the winning outcome lacks popularity.
- An attacker with only a small percent of the voting tokens can sway the vote. For example, it is easier and cheaper to bribe fewer people than many.
In the context of the Foundation’s governance proposal low voter turnout could happen for two reasons, firstly it could happen if not enough Algo holders applied to be governors or secondly it could happen if not enough governors (i.e. those with the right to vote) actually voted once becoming a governor.
If we are to take Vitalik’s earlier pronouncement that “crypto token holders are profit maximisers” then it stands to reason that the first scenario could be avoided by offering a large enough reward for people to become governors. Whilst we won’t know for sure how much of an inducement people will need to take in order to get involved with the governance scheme the Foundation have announced that the approximate amount of Algos that will be given in rewards will be 75 million Algos per quarter, and they expect between 1 and 4 billion Algos to be staked in this time equating to somewhere between 18 and 73% of the freely available tokens (5.5bn). This seems to successfully rule out a turnout as low as the examples Vitalik cited in his blogpost.
Gladly the Foundation’s governance proposal handles the second possible scenario of governors not voting extremely well. In order to collect governance rewards for participating in the Algorand governance program, governors (Algo holders enrolled on the governance program) must make a decision in every single vote held. If a single vote is missed then all rewards are forfeited, this will strongly incentivise participation.
Vote for the Good of the Community or you will Lose Money (Perhaps)
“In DPOS, the reward is constant, and it’s the voters’ role to vote for pools that have good performance, but with the key flaw that there is no mechanism to actually encourage voters to vote in that way instead of just voting for whoever gives them the most money without taking performance into account. Penalties in DPOS do not exist, and are certainly not passed on to voters, so voters have no “skin in the game” — Vitalik 2018
“Skin in the game” refers to a situation where the protagonist has a big incentive to see a specific outcome, thus inducing them to do all that they can to see that outcome come to fruition. In this case, it refers to the need for voters to gain financially from decisions that are beneficial to their blockchain’s community and lose if they vote against this.
The way that the Foundation has found to force governors (voters) to face the consequences of their votes is that one Algo committed is worth one vote and that they must hold these committed Algos for at least 3 months after voting in order to receive their governor rewards. The votes of governors who hold their coins for less than 3 weeks will not be counted. The theory goes that if a governor makes bad votes his coins will lose value and he will lose money.
This provides somewhat of a protection against careless decision-making on the part of governors because governors can sell their tokens if they think a bad decision has been made, forfeiting their status as a governor and any rewards but escaping without capital losses. However, it could also have the opposite effect; hefty rewards might incentivise governors to keep their tokens locked up in sufficient numbers to sustain the price of Algos (even when poor decisions are made) in order that they may keep their rewards. They would do this safe in the knowledge that other token holders are unlikely to be following the results of votes as closely as they are, and that the short-term exchange rate for USD/ALGO is not so much influenced by their vote as by its correlation to the wider market.
The three-month commitment period also makes it less practical for attackers to set up a conspiracy. It is unlikely that an attacker could create an honest-looking scheme for harvesting Algos which could generate an APY high enough to justify locking a user's coins down for that amount of time, especially not as people lending to such a scheme may as well just become governors themselves.
To summarise we could say that the way governor rewards are structured does ensure that governors have skin in the game as mentioned by Vitalik, and it makes it more expensive for attackers to corrupt governors. However, there is a conflict within the reward structure which may cause governors to hold their committed Algos even when they think a bad decision has been made. This could be a problem if a large proportion of the Algos in circulation are committed to the governance program because bad behaviour would not be punished swiftly enough by negative price action, leading to much larger collapses later on when the lack of progress becomes evident.
Human Beings can say no to Corruption
Vitalik makes a distinction between tightly coupled, and loosely coupled governance. Tightly coupled governance is that which is implemented on the chain and is automatic, such as when Ethereum miners vote for gas fees. Loosely coupled governance is done through decisions that signal to humans that their action is required, humans can of course ignore orders.
“there is a key difference between tightly coupled and loosely coupled votes. In a loosely coupled vote, direct or indirect vote bribing is also possible, but if the community agrees that some given proposal or set of votes constitutes a game-theoretic attack, they can simply socially agree to ignore it.” — Vitalik 2017
Governance of the AERP is a case of loosely coupled governance. Grant recipients (as opposed to governors) are not just account addresses, they are known entities either companies, institutions, or individuals. There is due diligence done, and grants are paid in tranches based on contractural milestones being met. This means that even if a fraudulent grant proposal was made, governors bribed and the vote won, the recipient would still have to make good on their promises to build something useful in order to get paid; for this reason, it is unlikely that anyone would take the risk of offering kickbacks.
It is therefore a much more compelling argument that coin voting won't be the cause of any bribery or corruption because the people holding the purse strings at the Foundation can simply refuse to indulge it.
De Facto Centralisation Prevents Corruption
“this kind of centralized trust model, while arguably useful in some contexts in a project’s early stages, is clearly one that is not sustainable in the long term.” — Vitalik 2017
Vitalik’s point about centralisation was that coin voting would not necessarily lead to corrupt practices because for many blockchains the coins are held by founders or foundations with large premines, and these people are not vulnerable to bribes and hold enough coins to outweigh most bribe attacks. This is not sustainable in the long run because control of these tokens will eventually move on to other people with less personal attachment.
For Algorand this is not so much about what the founding members hold, we know for example that the Foundation will not be voting in these polls, however, this does not mean that power is not centralised in other ways. The Foundation will control:
- What proposals make it to the shortlist.
- What type of vote will be held.
- What the contractual terms of any successful proposal will be.
- What the governance rewards will be, and;
- there will be a button making it easy for governors to vote for all the proposals recommended by the Foundation.
The point being that Vitalik's comments about centralisation with regard to decentralised governance still ring true with the plans of the Foundation. It is true that some control of the process will have been ceded to Algo holders, but all the power to direct and influence the vote will still remain with the Foundation.
Vitalik’s comments absolutely ring true with regard to the plans of the Foundation, as you can see from the ordered list above, the grant-awarding process is still very centralised and this will prevent corruption. To their credit, the Foundation knows that there is more that can be done to further empower the Algorand community. To this end, it is looking at ways to further democratise control of the AERP, but as we have seen above decentralisation could open the door to corruption, so stakeholders should keep a very close eye on how more powers can be decentralised.
Summary: is it possible that coin voting will lead to corrupt outcomes for the AERP?
No.
The plans of the Foundation for decentralising the AERP align with Vitalik in the following ways:
- Turnout: The number of Algos committed will be between 18% and 73% of the total float with almost all of these coins casting a vote.
- Loose Coupling: If a corrupt grant application is detected then votes in its favour can be disregarded by the Foundation.
- Centralisation: The Foundation still has most of the power meaning that (at least in the short term) grant recipients would be the same whether voted on by governors or directly awarded the grant by the Foundation.
In the course of the article we have also uncovered these factors which would make corruption more expensive:
- Grants are paid in tranches: This makes bribery pointless as (if successful) the grant would need to be used to fund the proposed work rather than bribes.
- Governors have to hold their committed Algos for three months: Although it is up for debate whether this will force governors incentives in line with those of the wider community (see “skin in the game”), what it will do is make it prohibitively expensive for those seeking to bribe governors.
So what could make this process less secure?:
- Further decentralisation: More power will inevitably have to be pushed outwards from the Foundation into the hands of the wider community, how and for what purposes this is implemented will need to be done in a way that maximises community engagement whilst minimising corruption or undue influence from minority interests.
- Issues other than grants to be voted on: Anything tightly coupled would remove checks and balances imposed by the foundation.
Are Voters Representative of the Whole Community?
“Another important objection to voting is that coin holders are only one class of user, and may have interests that collide with those of other users.” — Vitalik 2017
Moving on from the robustness of the governance program from attacks, Vitalik makes the point that coin voting is unrepresentative in the following ways:
- coin-holders aren't the only stakeholders in an ecosystem, and;
- whales may have power that is disproportionate to their benefit to the community.
Within any blockchain community, there are many types of stakeholder and the needs of these can conflict in different ways, therefore giving one group of stakeholders disproportionate power over the blockchain could lead to a fracturing of the community.
In the case of the 10bn Algos minted at the inception of the Algorand blockchain this is how they have/ will be distributed:
Assuming the groups which have been allocated Algos are largely still holding them, you can see from this schedule that many of the coin-holders are also representatives of other stakeholder categories, e.g. relay node runners, as well as developers, and entrepreneurs who have received grants to improve the ecosystem and tooling.
In our case therefore the Foundations plans render Vitalik’s criticism moot. Coin-voting cannot be unrepresentative of the wider community if the coin-voters are themselves also the wider community.
Conclusion
The Algorand Foundation seems to have learnt from the past insofar as most criticisms made by Vitalik against coin-voting governance prior to 2018 have been skillfully mitigated. Bribery and corruption would be too expensive, and coin voting will give more facets of the community a voice over how the ecosystem progresses. The main challenge confronting the Foundation is how to decentralise their power and influence further whilst maintaining the integrity of the vote. I for one am very excited to see how this pans out and am looking forward to the future.