Decentralized Governance Part 2: Governance and AI Kitties
http://www.pinaryoldas.info/WORK/The-Kitty-AI-Artificial-Intelligence-for-Governance-2016
In Part 1 I said I’d start with Dfinity’s Blockchain Nervous System (BNS). But the goal here is not to outline the BNS implementation in detail, as it’s pretty well covered here. My intention is to use the proposed Dfinity model to explore various governance issues. Since my last post, Cardano have added more detail to their proposed governance model here, Fred Ehrsam provided a comprehensive view here with Vlad Zamfir’s response here. Luke Duncan’s post here also covers similar ground to this post. It’s great to see so much activity in this area.
By the way before I go on, if you are interested in decentralized governance for permissionless blockchains then please join us our discussion forum on Gitter.
Let’s assume… it’s like a repeat of Bitcoin late 2013 … where nocoiners have won the day.
In a previous post, I outlined why governance is key for the traditional VC model to work and speculated how liquid token markets might open up new market-based approaches to governance. But what does market-based governance even mean? Well firstly, in our current climate of over-heated cryptoasset markets, one reason why VCs may no longer need their governance rights is that they now have the most basic market-based approach available to them: the right to sell on the first sign of trouble. Before VCs / investors had this liquidity path open to them, if some adverse circumstances arose, say there was a lack of faith in the leadership team, or stronger competition entering the market or changes in the regulatory environment, they had built in enough leverage in the boardroom to influence and enact change. Now they can just sell their tokens. However, this breaks down in a weak market, where this option would lock in losses and so is not very attractive.
But let’s assume we are running a token owner-centric, code-is-law, on-chain governance model. And let’s assume we are not in a frothy market: say it’s like a repeat of Bitcoin late 2013 and we are in brutish bear market, where nocoiners have won the day. Now a market-based governance approach could potentially give our VC a way out. In this situation where the governance rules give token holders voting rights, then the VC can buy up tokens to build a controlling stake. They hire a new development team, design a new strategy and use their freshly purchased governance rights to push through their changes. This is what we mean by market-based governance. Activist tokenholders, market discipline or what is really just a fairly typical private equity play.
One person’s hostile takeover is another person’s 51% attack.
Now one person’s activist tokenholder is another person’s 51% attack (or 34% attack for Proof of Stake / BFT). Maybe it’s not VC / hedge funds / PE trying to making a buck. Maybe it’s a nation state trying to undermine the network. Maybe it’s a corporation who doesn’t like being dis-intermediated. Now we can start to see why on-chain governance starts to look like a risk. And we have lost our censorship resistant network. But luckily we still have some protection. If a rogue investor / nation state / corporation pulled a stunt like this they are likely to cause a network fork.
If we play this scenario forward, then post-fork the governance rights temporarily improve for users who have moved to the forked network. Assuming the new network is smaller in terms of active users, and users on the old fork forgo their right to vote on the new fork, the user now gets a bigger say in decisions. The small community of forkers have forked for a reason, they tweak the governance rules to their liking, leaving the the on-chain voting as is, and sit back happy… Until the users on the old network sell up, the token price drops, activist investors move in… back to the above. So it’s clear that the first change by the forkers would need to be to remove on-chain governance. So could it be that on-chain governance is incompatible with the idea of these networks being publicly-owned utilities?
Over time good governance will be like brand value for equities: an intangible that could add a huge premium to network value.
It is clear that if reliable on-chain governance could be made to work then these governance rights would have real economic value. Notation Capital have an interesting estimate of the value of various investor rights here. In this model they see VC’s moving to more short-term horizons (and so haven’t put a value on governance rights), but for long-term hodlers (and in less exuberant markets) a governance premium would be factored into the token price. Over time good governance will be like brand value for equities: an intangible that could add a huge premium to network value, as it lowers risks for investors. And once it’s value is quantifiable then token holders would be unlikely to diminish these rights, as it shrink their asset value.
But what governance rights would investors value? Is there such thing as a typical investor e.g. hedge fund vs pension fund vs retail investor? And are interests of miners, investors, users, developers, entrepreneurs and corporations likely to align when it comes to governance? These are big issues that we’ll come back to in future posts. But first let’s start to build up a basis for answering these questions.
Let’s have a guess on what a hypothetical governance model might look like. In the early days, we may want to have a period of rapid experimentation and learning and maybe this needs a more flexible and lightweight governance model. Maybe the risks are lower, as the new network has little value and maybe a small group of self-appointed leaders can drive progress. Now let’s take a few steps forward to the end goals. We probably want to reach a steady state where things are very stable. The network has a clear, predictable roadmap with clear transparent processes around what gets onto that roadmap. Maybe the token issuance is fixed and if these tokens are issued to fund development work then this implies the resources required to deliver the roadmap are fixed. If token issuance needs to react to market conditions, then the process by which “central bank” issuance decisions are made should be transparent. Coded as rules in smart contract? Well, maybe. Same goes for mining fees. Ideally these are fixed up front but maybe this needs some governance role where there is an existential threat to the network e.g. mining has become unprofitable and miners are jumping to other networks. There are some serious decisions to be made here which probably needs a mature set of tools and processes. We need visionary technologists, architects, researchers, developers and product managers. We need cryptoeconomists and plain old economists. We need data scientists, business analysts and strategists. In short, we need experts. So the handover from project founders to formal community-lead governance needs a model that recognizes and empowers experts.
Now “expert” in the context of governance is a dirty word these days. Decades of corporate PR from various industries (with tobacco and energy companies leading the way) have ensured that expertise has been completely devalued. Liquid democracy (planned for use by Dfinity and Cardano) could be the governance model that restores expertise to it’s rightful place. In this model, good governance is about making sure decisions are routed to the people best qualified to make them, a meritocracy where someone can build up a reputation for being well-informed and the general populations reward them by empowering them to vote on their behalf. And if you are happy that you have the insights required to vote yourself, then you don’t have to delegate. But there is much more to this bigger idea.
Blockchain has made massively scaled up voting systems possible and this would allow every individual to vote on every issue. We are now starting to see implementations using blockchains that are secure, scalable and practicable (e.g. Patrick McCorry has a nice presentation here on the voting / verification side of things). BNS allows ultimate flexibility: vote yourself on issues you are passionate (and informed) about. Delegate when you don’t have a strong opinion but understand that it’s an important decision that needs careful consideration. Abstain if you don’t think it’s important enough to delegate. And you can have preset rules. Maybe you are a developer and you vote yourself on technical decisions, but delegate economic decisions. And delegates can delegate their votes too, which allows for a different voting strategy: I can now delegate my vote not to the person I think is best placed to make the decision, but instead to the person who is best placed to know who the best person to make the decision is. So in theory, this seems like a great way to reward the people who invest the most time in understanding the trade offs involved in a decision. They get rewarded with the voting power of hundreds or thousand of followers.
In the Dfinity BNS model token holders vote in proportion to the number of tokens they hold. This is a form of timocracy, a system where only property owners get to vote, but in this variant of timocracy the voting is weighted proportionally to property ownership. This same model is also built into Proof of Stake designs and is also proposed by as a governance model for many other DVNs. Of course, this is not about lambocracy (thanks Alex Stokes). There are very good practical reasons why these models are being proposed. DVNs are designed to offer pseudonymity and a “one address, one vote” model wouldn’t work, as it would be very easy for a large token holder to spread their tokens across a large number of addresses. We are also not advocating the introduction strong, verified, public identities (“one citizen, one vote”) to solve this, because it breaks important principles like privacy, censorship resistance etc. and leaves us prone to tyranny of the majority issues. We will, however, need to come back to this subject, as maybe there are additional checks and balances that can help this model avoid sliding into plutarchy (particularly in the context of the hostile network takeover scenario outlined above).
But back to delegation. Don’t we delegate already? Shouldn’t we be looking to avoid the pitfalls of the governance model we are most familiar with i.e. representative democracy? In typical national, state or local-level politics we delegate all our choices to a person we feel can best represent our views. In turn these delegates outsource many decisions to a second tier of governance, that of their own political party. The shortcomings of this model are plain to see. So is there a risk that liquid democracy will evolve into something similar? To understand this we need to understand the “hacks” that subvert representational democracy today.
A naive analysis of an imaginary, yet typical, representative democracy would indicate we have an appropriate level of validity and transparency. The candidates / parties say what they stand for. Everyone has a free, anonymous vote. We have a free-ish press. Anyone can put themselves up for nomination. You can argue about “first-past-the-post” vs proportional representation but the voting process is mostly transparent (even if sometimes convoluted) and mostly robust (with the odd e-voting hack for sure, and maybe a little gerrymandering). So far so good (and yes naive).
But this is the tip of overall governance iceberg. The biggest real-world governance issues we face today happen at the edges of this process. The hacks are on pubic opinion. They are on the selection process of who gets a shot at power. They are around funding. They are on what ideas are part of public discourse vs. things that are controversial and therefore not for mainstream consumption.
The obvious social hack is corruptibility of delegates through incentives, threats or blackmail. When you have a small number of people empowered to make very significant economic decisions, there are obvious gains to those who can successfully corrupt delegates. In the real world, this might involve decisions regarding radio spectrum allocations, real estate planning permission, network neutrality, financial regulations, energy subsidies etc. For DVNs the economic issues include decisions on fees, token issuance, treasury decisions, bounties / developer rewards etc. While we may consider these as forms of 51% attacks, many hotly-contested decisions may only need a tiny vote swing to sway a decision. So this is the first thing we need to watch out for with liquid democracy / BNS: ensuring that we can detect and respond adequately to corrupt delegates. It may not be enough to assume that in a repeated game (in the game theory sense of the word) such behaviour will be punished. After all, the inertial forces that keep politicians in place, even after blatant policy u-turns, never ceases to amaze.
And it’s not just about corrupting the voting process. In a parliamentary democracy only a small number of laws are reviewed in any one term and the ruling party or coalition has full control over what issues are addressed. So we need to understand the process to get issues onto the agenda, how issues raised might be dropped even before a vote takes place, how the wording of proposals was crafted / influenced and what biases may be at play when drafting these. This is an important part of the governance process that tends to get overlooked.
While we look for decentralized governance solutions, most networks have opted to keep their decision-making processes centralized, usually within a tight-knit group. And for most, there is a genuine intention to work towards a decentralized and more transparent process. This approach is the right one for where we are now and remains robust in the face of corrupting influences. Newcomers may have dubious intentions and for many networks the people who were involved early (long before the crypto economy was worth $500B) are best placed to represent the interests of the community for now. They have proven their dedication by building up advanced knowledge of the technical, economic and political issues. They are the right people to take this to the next stage. But the eventual shift from centralized to decentralized governance has huge risks. Any mistakes that are made will be very, very hard to change later and we could end up living with a deeply flawed governance model for many years to come. Like in the real world.
In the next post I’ll cover Nick Szabo’s tweet stating that overthinking the design of governance would lead to weaker solutions, as opposed to using tried and trusted open-source governance models. So are open-source governance models a good fit for the governance of decentralized value networks and where might these approaches break down?
To wrap up, one final comment. Dfinity BNS is designed with AI in mind. The intention is that an AI bot would learn your preferences over time and eventually end up voting on your behalf. I can only refer you to Pinar Yoldas work referenced above. In a bizarre twist, this artistic work from last year has perfectly predicted the rise of the CryptoKitties phenomena, especially as we move to liquid democracy. We must soon decide whether the next logical step is to delegate governance to these AI Kitties.
Once again, if you are interested in decentralized governance for permissionless blockchains then please join us our discussion forum on Gitter.
Brendan Dillon is CEO at Reflexivity. We provide consulting, advisory and software development services, with a particular interest in governance-related projects.
Please contact us at info@reflexivity.network.