Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Ava Labs x CBER Ep 2: Governance of Decentralized Autonomous Organizations
114 Plays11 months ago
Brett Falk (University of Pennsylvania) and Gerry Tsoukalas (Boston University) discuss voting protocols in Decentralized Autonomous Organizations. Brett Falk (University of Pennsylvania) and Gerry Tsoukalas (Boston University) highlight weaknesses in existing voting protocols. An important point is that many voting protocols weight votes based on token holdings but a voter’s token holdings does not, in general, reflect the level of information that the voter has with regard to a voting proposal. The consequence of this misalignment between a voter’s weight and a voter’s information is suboptimal welfare outcomes.
Recommended
Transcript
Introduction to the Podcast Series
00:00:06
Speaker
Hello and welcome to Crafting the Crypto Economy. I am Silvia Sanchez, Project Manager at OWL Explains by Avallabs, and today we bring you a transformative podcast series in partnership with the Crypto and Blockchain Economic Research Forum. This series features leading faculty from renowned global universities exploring various elements in the blockchain ecosystem. These episodes are a bit longer than our usual hootenannies, since we will be getting very deep.
00:00:33
Speaker
and also each episode will have its accompanying paper posted on our website for further reading. And with that, I will hand it over to our moderators Fahad Saleh and Andreas Park.
00:00:43
Speaker
Hey, everyone, and welcome back to another edition of the Crafting the Crypto Economy podcast series organized by the Crypto Economics and Blockchain Research Forum, CBER and OWL. We're here to discuss recent research papers and trying to get you to understand better what people are working on, what kind of open questions there are in terms of crypto economics.
00:01:05
Speaker
in the blockchain space, and we have two great speakers here with us today who will talk about their work on DAO voting. Now, maybe we should start to introduce these speakers first. So, Jerry, how about you start with yourself?
Key Participants and Focus Areas
00:01:24
Speaker
Yeah, thanks, Andreas. It's good to be here. Appreciate the invitation. So, I'm Jerry Zucallis. I'm an Associate Professor in the Information Systems Group.
00:01:34
Speaker
in the business school at Boston University. And I work mainly on digital platforms, and blockchain being one of the things that I'm looking at. Excellent. And Brett, how about you? Yeah, hi. I'm Brett Falk, and I'm a research assistant professor in the computer science department at Penn.
00:01:51
Speaker
So as you can see this time around, we don't have actually economists in the room. We have people who have a very big background in operation system, computer science, and so on. The advantage hopefully will be that in contrast to us economists, you have a better understanding about many of the nuances of how DAOs are organized. But maybe we should start by taking a very high road and big picture overview because it's probably
00:02:17
Speaker
very important to understand first what a DAO is, so a decentralized autonomous organization.
Understanding DAOs: Concepts and Challenges
00:02:24
Speaker
I'm not sure who wants to take that question, but Jerry, maybe you can just give us a big picture overview of what this is all about. Yeah, sure. So I guess in my view, a decentralized autonomous organization or a DAO is basically a platform like you find in the economy,
00:02:47
Speaker
some examples Amazon, Uber, etc. But there's no one really running the show. There's no central hierarchy running the show. It's organized as a decentralized structure where all the participants, by that I mean both users, the people who are providing service and the people who are buying service, are coming together and somehow deciding the decentralized fashion, how to govern and operate on this platform. And the central entity
00:03:16
Speaker
that usually makes decisions. For instance, in the case of Uber, you know, who sets prices and who matches what we're looking for rides, et cetera, is replaced by things like a smart contract that automates these types of economic decisions. And I'll just say one more quick thing. You know, economists have looked at decentralized systems for a long, long time. But the fact that blockchain technology really enabled that
00:03:46
Speaker
you know, much more than before, I think has really launched a new research area from all different sides, including business schools, computer scientists, sometimes some mathematicians and all over the place, really looking at this, this really interesting new area. And, um, you know, in my view, there are, there are just say four major obstacles, um, that one needs to consider when creating these systems. Uh, one of them is how do you collect information from
00:04:14
Speaker
from different people on the system. So you don't have a central
Governance in DeFi: Aave and Compound
00:04:17
Speaker
decision maker. So how do you actually collect information from everyone? And that's what primarily with the papers that I'm assuming we're going to talk about today. But there are other things like how you record the information. So that has to do with consensus mechanisms and other things like that. The third thing is how do you coordinate user behavior? So outside of the information aspect, how do you get people to act in a way that not just in their self-interest, but also helps the platform as a whole,
00:04:45
Speaker
And maybe a fourth thing I would throw in is how do you raise capital for these types of platforms? So that would be, in a nutshell, my definition of what I think a Dow is.
00:04:58
Speaker
That's a pretty broad definition. I see Brad wants to jump in, but let me just quickly say, just for the audience's sake, maybe just a very brief summary here. Really, what you're looking at is there's an attempt to organize if you want a firm without an address, which is what blockchain really is, and it's a firm without managers. It's a fully democratic, organized entity.
00:05:22
Speaker
And that sounds actually something which can be very difficult to get off the ground. We know in a democracy, it's actually hard to get people to vote, to do anything. And then I presume, as you say, so I think one of the clever things about it now is, OK, so maybe people have the input, but then there is a particular contract feature. So that means an algorithm which then implements the choice that people have.
00:05:49
Speaker
Now, maybe it's useful for the audience to first actually take one or two examples. For instance, if I may suggest one, one of the major applications that we also discuss here in this podcast are DeFi applications, decentralized finance applications.
00:06:09
Speaker
And the biggest ones that, you know, the most important one I would say, because they're central in the operation are exchanges and borrowing and lending, right? Because that's really what most of finance is about in some form or another. So maybe we should, can you maybe pick one of the major platforms as I understand they are in lending and borrowing that are organized as DAOs and say what it is that they actually do? Maybe Brad, do you want to take that question? Yeah, that's exactly what I was going to jump in and say before.
00:06:38
Speaker
Like on Ethereum, two of the biggest lending protocols are Aave and Compound, and in a lot of ways they work fairly similarly.
00:06:47
Speaker
And so they have a bunch of on-chain contracts that work as escrow contracts, and people can go and deposit funds into these contracts as lenders, and then people can come and put up collateral and borrow from them. And the contracts are decentralized and will kind of keep running. And so in some sense, you could imagine naively that you wouldn't need any kind of governance once you put these contracts up. They just sort of will their code, they're just going to keep running without any human intervention.
00:07:14
Speaker
But actually, you need lots of decisions to be made. So for example, things like what type of collateral would be a good collateral, right? So if somebody puts up collateral, should we allow wrapped Bitcoin as collateral to borrow?
00:07:27
Speaker
You know, USDC and the different new tokens are made at all different times. And so, you know, you can't just have a single contract that knows in advance what are all the good collateral types because you don't want to accept bad collateral. So some somebody has to make decisions on the fly about should we accept this as collateral? So that's like one type of thing. Another question is about interest rates. So.
00:07:48
Speaker
they can have some algorithmically determined interest rates. And the way it works on Aave is the interest rate kind of depends on how much of the token has been borrowed from the platform, what they call the utilization rate. So if it has a billion dollars that it could lend out and 70% of that has been lent out, as the amount that's lent out goes up, the interest rate goes up. But how much it goes up is determined by a governance parameter. So there's this slope of this interest rate curve is determined by a governance vote.
00:08:16
Speaker
Then there are sort of even bigger picture things of like, should we take AVE and deploy it on another chain, right? So initially it was deployed on Ethereum. Should we take this code and deploy it on Arbitrum or on Optimism or some other layer too? This is something that the code can't do by itself. Somebody has to actually vote to do it and go do it. So these are the kind of governance decisions, even though the contracts are kind of running by themselves and can make some decisions themselves, there's a lot of need for human intervention. And to keep things decentralized, we keep
00:08:46
Speaker
this as sort of as a DAO, where token holders vote. So let's go with the first part then. So the ones where these decisions are really run by the contract. So that means by piece of code running on the blockchain. So as you described, there is the interest rate has, for instance, be determined. And that, you know, there's, there's probably a mechanism, I presume this is a mechanism which also runs, maybe not continuously, but at least there's certain cadence in which these decisions are being taken.
00:09:15
Speaker
Now, can you run us through how actually, let's say in the example of Aave, how would you describe the different rates? So I presume there is a cutoff rate for the utilization, so utilization being the fraction of the amount that is in a particular pool has been lent out, and then also the slope or the probably you would do this at what is the rate at the threshold point. So how do you determine these votes? How is this done?
00:09:46
Speaker
So for Abe's setting interest rate, there are two parameters that are chosen by governance, which is the interest rate at the target utilization, and then that kind of interest rate curve above it. So there's one slope going up to the target utilization, and then there's a steeper slope once you get above that. So it'll be like you're going towards 4% interest until you
00:10:09
Speaker
have lent out 80% of your reserves and then the interest rate curves jumps to like 60% and it heads towards 60% as you bend out everything. And so there's a big distinction in interest rates once you get above this target utilization rate.
00:10:25
Speaker
And so these two parameters are things that are chosen by governance. And the way Aave governance works is that someone has to propose, they make a binary proposal. They say, I think the two interest rates should be 5% and 65%. And then the Aave token holders get to vote yes, no. And this is taken as a token weighted majority average vote. And if the proposal is accepted, then those new interest rates get incorporated.
00:10:55
Speaker
So to clarify there, if I can jump in here, you're saying the current process is just an even weighting over AVE tokens amongst the people who vote. Is that correct? Yeah. So you just need a majority of it's a token weighted vote and you need a majority of the people who vote to have voted yes.
00:11:15
Speaker
By the outer curiosity, is there a required quorum or anything like that, or it's just literally whoever? No, there's actually a minimum that because, as you expect, there's sort of low voter turnout, you don't need 51% of all of the tokens in existence to vote yes. You need 51% of the people who voted, plus at least 10% of the token holders have to have voted. So there is a minimum cutoff.
00:11:43
Speaker
I see so, and then if there's no vote happening, the quorum is not reached and you stick with the status quo, is that correct? Yes.
00:11:52
Speaker
So maybe we can also, so we had one example would be the borrowing lending. And so it seems that there is a, and in a borrowing lending, you can imagine that you actually need to change interest rates regularly because of market changes, right?
DAOs vs Traditional Firms: Voting Complexities
00:12:06
Speaker
And I'm just gonna say this as an economist, I find it very interesting that we actually don't have a market mechanism there, right? So we have the voting mechanism where basically if you want the DAO sets a particular rate,
00:12:17
Speaker
And then we have the provision of collateral and the borrowing lending against the particular decision that is being taken. It's kind of interesting for me because you can imagine you can also have a market mechanism that organizes the whole thing. And with that, I would actually like to jump to maybe another application for DAO voting, which are decentralized exchanges. So AMMs also have potentially a way of how they can make votes.
00:12:45
Speaker
Do you have an example for that maybe? Yeah, so there's a lot of overlap in these AMMs, so there's still questions about should we deploy Uniswap on this new L2? Every time a new L2 comes out, that's an EVM chain, you have the option, should we deploy our contracts on this new L2 or not? So that's a common vote that you see coming up.
00:13:12
Speaker
Other questions revolve around treasury management. So this is also something that kind of happens off-chain, but a lot of times these protocols collect fees either from their initial token sales or from streaming fees coming in. And now you want to ask, should we use these tokens to fund people to build new front ends or wallets that interact with our protocol or should we
00:13:35
Speaker
fun people who are going to develop new smart contracts. So it's sort of like, how do you invest the platform's resources in sort of general growth? So these are things that affect basically all the DeFi platforms that used as. I see. So is there actually an example of an AMM, for instance, where the fee is actually being voted on? So in Uniswap version two, there was a fixed fee on every trade of 0.3%. But in Uniswap version three, there are different fee tiers.
00:14:04
Speaker
And initially there were three fee tiers, but new fee tiers could be added by governance. And so on every Uniswap trade on a specific pool, there's some percentage fee and governance could make up new fee tiers. And Uniswap governance works similarly to Aave and compound governance where someone has to propose a yes, no vote and then you vote and it's a token weighted vote. And if 51% of the voters vote for this, there's
00:14:32
Speaker
a new fee tier can be added. And again, there's some minimum quorum level that some fraction of the token holders has to vote for it to account.
00:14:41
Speaker
Okay, so now this is nice. So now we have two examples of mechanisms or decision problems which require the involvement of the DAO holders. Can I just ask a quick question here though? So more broadly, is there any restriction before the initial deployment as to what you could make sort of subject to the governance process?
00:15:08
Speaker
Yeah, that's absolutely right. So you could, right, people talk about smart contracts being immutable. So I can make a smart contract that just goes up and nobody can change it, right? Or I can make a smart contract which has some voting module as part of it, and the voting is allowed to change certain parameters, but not other parameters. So this tends to be pretty much baked into the system at the beginning of like what parameters voting is allowed to change. Now,
00:15:37
Speaker
Voting could always do off-chain stuff. You could vote to hire some developers to write you a new contract. But within the contract, it has to be specified in advance what are the things that voters can vote on.
00:15:50
Speaker
So we see here, this is already quite a complex problem. So we have two examples of protocols that have some need or potential for on-chain voting for particular features of an existing protocol. Then we have the decision of what you actually want to put in. So before the deployment of the contract, you have to think about what is it actually that you want to make subject to voting.
00:16:12
Speaker
And then we have decisions about what you can do off-chain. Now, I think for this particular podcast in our audience, the most interesting part is really the one which happens for the contract itself. So everything that goes from that happens on-chain that determines parameters of a contract to what it is that you actually should be deploying in the first place.
00:16:31
Speaker
This is just my take on this. Now, Jerry, maybe you can run us through what kind of complications arise and what kind of questions arise when we are asking, when we're looking into the specifics of how we do the voting, but also the parameters that are in there. What problems, economic problems arise? Yeah, a lot, actually, which was surprising when we first started working on this topic with Brett many, many years ago.
00:17:01
Speaker
I think my first question was voting is solved. My first comment was along those lines. So, you know, what could there possibly be to understand? But what's interesting is that these, you know, a lot of blockchains, even from the very beginning when people started talking about governance, there was a system that was adopted, which is the token-weighted voting system from the get-go, which seems like a very reasonable system.
00:17:30
Speaker
where people with more tokens, they have a higher voting power than people with less tokens. And it's similar to shareholder voting in that sense. It doesn't seem unreasonable. And one of the questions that we wanted to ask was, well, is that really the optimal mechanism? How does it perform? And the whole basis of our
00:17:59
Speaker
inquiry started from a very simple observation, which was the fact that it's very possible that people that have a lot of tokens and have a lot of power in the voting might not have the best information at the same time. And so that was the spark that really launched this whole investigation. What happens if someone has a lot of tokens but is misinformed? Then what?
00:18:25
Speaker
And so intuitively you can kind of extrapolate and say, well, that might actually be bad for the system. And so we, we dived into a whole investigation of, you know, does token weighted voting work? Should everyone instead just have one vote? And so there was all sorts of questions that came from that, but now I'll stop here.
00:18:43
Speaker
Okay, well, so the information part is the first, can I, just out of curiosity, and I think it's probably also worth pointing out, is that when we think of shareholder voting, it's actually pretty single dimensional,
Optimal Voting Mechanisms in DAOs
00:18:57
Speaker
right? Because a shareholder has a particular interest, which is on the profits or earnings of the firm, and they want to have those maximized. But a DAO is much more complex, right? Because a DAO is a platform, on a platform, you actually have two parties that interact,
00:19:10
Speaker
It's quite possible that somebody who is, for instance, if you take an AMM, is a regular liquidity taker, so somebody who goes to the contract, wants to trade with the contract, has an interest in low fees, whereas another Dow token holder could be somebody who is a liquidity provider and wants to have as much income from those.
00:19:31
Speaker
Their decisions could be conflicting or they could actually be aligned. It depends on actually how the system is set up, but it's worth pointing out that a DAO is much more than a firm in that sense. This makes, in my opinion, what you describe probably much more important than interesting, which is the question of information.
00:19:51
Speaker
So the one thing that we're concerned about in finance is asymmetric information or somebody knowing more than another and just simply the ability of a market to aggregate information. So that seems to be that this is the problem that you did you try to zoom in first.
00:20:07
Speaker
Now, can you enlighten us a little bit of how does actually voting contribute to information revelation? So how does information actually come about and is used in the most amount of information is actually revealed? How's that done? Sure. So I think sort of as a good running example, we can think about this voting for an interest rate on a lending platform. So this can be like a good thing to have in mind.
00:20:36
Speaker
is you want to vote on what should the interest rate parameter be. And if you imagine that Aave token holders are mostly incentivized to try to choose the interest rate that's the best for the platform, that's going to encourage the most activity and generate the most fees for the platform.
00:20:54
Speaker
You now have to choose how to aggregate things. In our first paper, we looked at a voting mechanism that you would call now maybe direct voting, or sometimes it's called instant voting. And it's not the mechanism that's used by AVE, but it is used by one inch. And this is a mechanism where if you're trying to set a parameter, which is like a real value, say a
00:21:14
Speaker
number between zero and one of what the interest rate should be, that everybody votes and then the platform is going to take some weighted average of everybody's votes and probably it will be a token weighted average so the people who have more tokens get more say in setting the parameter but everybody gets a little bit of a say in setting this parameter.
00:21:34
Speaker
Okay, so and then with that voting, so are you saying that the particular voting mechanism that is used there, which is essentially one token, one vote, is that optimal or is there a better way? In fact, actually, let's start with the question of what is actually the problem that people try to solve. I'm not sure about this, but I don't know if there was a lot of thought that was put in, or enough thought that was put in, in terms of information aggregation in these early blockchain systems. It's not entirely clear to me.
00:22:05
Speaker
that there were studies done on what's the optimal voting mechanism. I think shareholder voting to some extent works this way. One share, one vote. And so it's almost a natural extension for the blockchain space. We have one token, one vote. But one of the things in the blockchain space is that you have situations where voter incentives might actually be aligned, and this can happen quite often, but people just don't agree on what the best
00:22:34
Speaker
the best value is for a certain parameter, et cetera. And so the inquiry was really about, well, how good is this? But then the answer to that, we have to define a benchmark. So what are we comparing it to? And you could cheat a little bit and say, well, there's this optimal benchmark, which is, imagine there's some central entity, which again is the centralized system.
00:23:03
Speaker
that can somehow pick everyone's true information and aggregate everything and then make a decision on behalf of everyone on the platform. So imagine if every single user on the platform were to reveal their information to a central entity truthfully. The central entity would have a broad view of what everyone thinks and will actually be able to make a decision. That's what we call a first best optimality in this paper or more generally the best possible outcome.
00:23:33
Speaker
that you could achieve. But it's worth pointing out that even centralized systems may not work this way because you're assuming that people truthfully report their information. And so it's not clear. So it's really the absolute best you could ever, ever do. And of course, we find in this paper that a lot of times you don't get there. Actually, sometimes you do. We could talk about situations where you do and you don't.
00:23:58
Speaker
But you don't get there. And then the second benchmark is, well, is there just another simple practical system that could be implemented instead of token-weighted voting that could be token-weighted voting? And the one we examined in this paper, mind you, this was a few years ago, was simply one person, one vote instead of one token. And we call that a one-over-end voting mechanism.
00:24:23
Speaker
So there's not multiple things that came up here right away. So the first thing, just as a comment, it seems like what you're suggesting is essentially Vitalik's vision of you have proof of personhood and then a single person can actually make a vote. It's an interesting concept as it is.
00:24:41
Speaker
What I would like to go and maybe can you explain some of the reasons as to why we would not see the optimal outcome reveal itself? What are the obstacles there? What's going wrong that that doesn't work? Yeah, so I'll just go back. Maybe Brett has some thoughts on this as well, but I'll just go back to the original statement, which was the fact that there can be a misalignment between people with tokens. And the way the system is set up,
00:25:11
Speaker
It's not directly meant to deal with that misalignment. To deal with it, you have to go outside of that system. You have to find ways, for instance, selling your votes to someone else, proxying your votes. You have to find other ways to do that. The mechanism itself doesn't take care of that. For us, when we were looking at this problem, it was a clear point of failure. And so the first question was, well, what are the assumptions you need
00:25:41
Speaker
to actually get this system to work. So how would users have to behave for this thing to work in the sense that to recover the optimal outcome, which is the best possible outcome, the first best outcome? And one of the first results in that paper, where we were looking at people acting strategically, meaning that they're voting, they get some information about, let's say, the interest rate, but they don't just vote, let's say, I believe it's 5%. They're not going to vote 5%.
00:26:10
Speaker
Why? Because they're going to try to figure out what other people are going to vote. So we call these strategic points. Let me give you a really simple example just to make sure things are clear. Let's say I believe that the right interest rate is 5%. But I know that someone else who hasn't done as much work as me thinks it's 15% or let's say 10%. But I know that it should be 5. So I'm not going to vote 5.
00:26:36
Speaker
I'm going to vote, maybe I'm going to vote less to try and bring down the average because I know that someone else will vote 15, so I need to vote maybe zero. And it works the other way around as well. And so that's what we call, that's what we mean by strategic behavior. So it's a very, it's very common in the academic literature to treat people this way. It's much more debatable whether people actually behave this way in practice. But nonetheless, that was the starting point of the analysis was what if, you know, all of these players that are voting, let's say on an interest rate,
00:27:06
Speaker
What if they are aware that there are other people voting and they vote strategic? They have some idea of what they think, but they also see, you know, there's a broader sea of people there. They each have their own ideas. Now, you can't directly observe their information, but you can kind of understand, you know, there's people that are good. There's people that are bad. There's people that are informed and completely uninformed. And so you have that as your information. And it turns out that if everyone is strategic and they have this base layer of knowledge,
00:27:35
Speaker
You also need to know people's token holdings. You can actually recover the optimal outcome if people vote strategically. So there's an equation in the paper that basically describes exactly, you know, once you observe your information, this is how much weight you should put, or this is how you should send your vote to the smart contract that aggregates the vote. And so there's a closed form formula for that, for how you would go about doing it.
00:28:00
Speaker
And what's interesting about the solution is that one of the things you need to do is kind of undo the waiting mechanism. The fact that the aggregation mechanism is via token-weighted voting, one of the things you need to do is undo that part, just kind of take the tokens out of the picture, because it's really about information that people have. And so that was kind of one of the main results. But, you know, of course,
00:28:24
Speaker
All right, so really the key part here is, so the interesting part I see here is, so really you're interested, information doesn't depend on the token holdings, right? You can assume it doesn't as a first step. And then you can, you know, the second part of the paper goes into more details about that.
Challenges in DAO Voting Systems
00:28:42
Speaker
But you're right, as a first step, you can just make the assumption tokens are exogenous, the system, you know, token holdings are just randomly distributed, what happens?
00:28:49
Speaker
I mean, there are a couple of things I wanted to add. So one is I just want to highlight what Jerry said that it's actually really interesting that the first thing people do is undo. The contract has some weighting mechanism in it that says we're going to weight everybody. We're going to take the weighted average. And if voters are smart and they act in their own best interest, the first thing they do is they change their votes to undo that mechanism, which is kind of a funny
00:29:17
Speaker
And then they then try to re-implement a mechanism which awaits their votes according to their information. So it seems almost as if you are actually able. So one of the things that we learn from markets is markets can be quite good at revealing information, provided there is the right incentives for people to act in their own interest. So is that maybe something that works in your favor there? In some form, it's a habit market.
00:29:46
Speaker
Yeah, it does work in your favor. It doesn't get all the way there because you're very constrained in how you can unwind this mechanism because you know, the smart contract is only going to take a linear, you know, weighted average. And so you can't force the contract. You can change sort of the weights in the contract a little bit by, by changing your own boat, but you can't get it to do sort of arbitrarily different things. So you can't get all the way to the, to the sort of best outcome.
00:30:12
Speaker
In other words, what you're really finding is, actually, even though you do something silly, which is the one token, one vote voting, in this particular case, if you had one person, one vote voting, this would actually be better because of the information effect. In terms of revealing the right information to take the best possible decision, the latter would actually be better, as I understand it.
00:30:33
Speaker
But you can say that if everything works out well, then people have the right incentives, they probably have to face some cost of getting it wrong and the like, then you can actually recover that result if you want. But are there also situations where that may not work? I mean, I'm not sure if this is part of your particular paper, but is there also something where that could fail?
00:30:55
Speaker
Well, I would say that it does kind of fail and that you still can't get to sort of this best outcome. You can get closer to it when the voters act really strategically. If the voters just vote, if I think the interest rate should be 5% and I just vote that, we don't do very well. If I take into account that other people are going to vote wrong, we can get a little better, but we still can't get all the way there because we can't fully unwind the one token, one vote mechanism. I mean, the only situation where you can fully unwind it is if
00:31:24
Speaker
every single person is strategic and is voting strategically. And there are no issues about how much effort do I put into acquire information. It's really a perfect world where there's no cost of voting. There's not even a single person voting truthfully in the sense that they just submit their information truthfully. That actually undermines the whole system, which was an interesting result as well. But yes, in all other cases, you can't
00:31:54
Speaker
you can't recover fully the efficiencies that you would under a central planner with the ability to aggregate all that. Right. So basically what you're saying is, in principle, the bar is actually quite high that you have given the particulars of the mechanisms that are there. Because it seems a little counterfactual, as you say, because if what you were doing would actually be entirely correct, then not correct, but true, then you would have
00:32:22
Speaker
100% quorum each time, right? Because everybody would actually, there's no cost of voting, so everybody would do the voting, and we know that that's not the case. So if I can ask then, given the difficulty of maybe implementing this, and given the sort of fundamental nature of the point I think you're making, which is this misalignment between information and token ownership,
00:32:47
Speaker
Is there a role for information intermediaries in these markets? What practical guidance would you give given your fundamental finding about this misalignment, again, between token ownership and information? Yeah, that's a good question. I think what this result points to is, first of all, things that seem very intuitive don't necessarily work very well. And you might want to go outside of the system.
00:33:15
Speaker
and do things to adjust for the shortcomings of the system. And in this particular case, there are things like proxying your votes. If you know that someone, let's say you're voting on an interest rate and someone has expertise in that area, you might want to give your vote to that person. And so going around the system and enabling this type of information transfer seems to be quite important. And it's not unusual. There's a lot of situations where
00:33:42
Speaker
In real markets, you're using the model, you're using an algorithm, and people understand that the algorithm isn't perfect, and they do things around that algorithm to try and adjust. And just as an analogy, one thing that comes to mind is people are familiar with financial derivatives. This happens all the time in the derivatives market. For instance, just options on stocks are based on the Black-Scholes model, which makes assumptions about the distribution of the returns of the stocks, and people understand that those assumptions are
00:34:11
Speaker
wrong. And they go around that by, for instance, assigning different volatilities to different strike prices over the same stock. So you could say Apple in the next month is going to have a variance of 20%. Someone else might say it has a variance of 30%. And both are consistent in that model because people are making adjustments to address the shortcomings of the algorithm itself. So I think this is one direction that this analysis shows that actually proxies are important.
00:34:41
Speaker
And any type of mechanism you can have or intermediary you can have to enable better information sharing and kind of ignore that the token component of it would actually be better than the status quo at the moment. Can I just add that most DeFi protocols do allow some kind of proxying or delegation of votes. And so that's generally good. The one downside with that is that in terms of just straight information aggregation,
00:35:11
Speaker
if there's only a few people with a lot of information, when if too many people delegate to them, you lose some of these benefits you would get from like a law of large numbers type of thing. And it's actually kind of interesting, even when people are strategic. Once you delegate too much to one person, you imagine one person, you know, has 90% accuracy, and you have 1000 people who have, you know, 51% accuracy. Listen, like the
00:35:38
Speaker
The central limit theorem is going to basically say you're probably actually better just to take the average of all these thousand people. But all these thousand people see this one person is really good and they're all going to delegate to them. And you can end up with these problems of centralization, which are not just that centralization is bad, but actually you lose some of this aggregation effect.
00:35:59
Speaker
So I think that actually what the nature of the information is is actually quite important, right? Because, for example, if it's just something that, given sufficient analysis, it's sort
00:36:14
Speaker
revealed to everybody.
Role of Information and Analytics in Voting
00:36:16
Speaker
And this relates a bit, I think, to like Jerry, when you're referencing the Black-Scholes model, right? Like that's, we know it's wrong, but it's essentially a framework that allows us to start thinking a little bit more deeply about the problem. And we kind of make these ad hoc adjustments. But so related to that, then, is there some potential value for platforms like Aave or Compound or Uniswap to provide sort of essentially analytics
00:36:40
Speaker
to reduce the cost of information acquisition for the various token holders. Yeah, and I think a lot of these platforms provide at least some kind of analytics that they have more and more complex dashboards that show sort of everything that's going on. And then a lot of them contract with outside analytics providers who provide guidance. And these end up being a lot of the biggest delegates.
00:37:10
Speaker
You can hire somebody to do simulations, basically, and then voters can delegate to that person. So we see this happening. In some sense, if I may say so, the fascinating part that I'm just observing there is we're moving from a full democracy, everybody has to do their own research type of world, to one where we actually, in some form, centralize information acquisition and information provision in some way, simply because it is more efficient.
00:37:40
Speaker
Is that probably fair to say? I mean, this is something probably which the paper is not really covering, but it seems like that's essentially what we're getting at here. Yeah, I think we do see that. And it's very interesting to see those sort of delegations either to individuals whose primary job is basically like risk analyst or
00:37:59
Speaker
to companies who are doing this professionally. Or we see a lot of delegation to university blockchain clubs where a lot of the top universities have a blockchain club that acts as a big delegate on a lot of these DeFi platforms. And the hope would be that they are sort of unbiased and willing to do some research about. So I'm just going to say this for if there's any management students listening here. So it seems like intermediaries and managers still have a role to play in the world, but it's probably important.
00:38:28
Speaker
Unfortunately, yeah. Hey, aren't you employed in a business school? That's right. Yeah, I mean, I'm just the vision. You know the vision of I don't claim to speak for that for the decentralized community systems community, but the vision is that you could eventually do away. With any any type of intermediaries, but if you're going to get there, a lot of the research that's been done the last few years is increasingly showing that it's pretty hard to. There's a lot of obstacles all over.
00:38:59
Speaker
Yeah, I mean, you know, we know also from just from from the economics literature, voting is costly, and that alone creates problems and information acquisition is costly. So if you if there could be more efficient ways to organize this as a market. Now, maybe we should move on to discuss your second paper. So what's that about? What kind of problem are you studying there? And how does it relate to the one that we've discussed so far? That was a big narrative in the early days of these
00:39:29
Speaker
and different types of platforms was token-weighted voting is so good because big token holders will be incentivized to go and figure out what the right decision is. And so we tried to model that as well. So since we talk about intermediaries and information intermediaries, one of the things that could be relevant here is that there may be also an incentive for token holders to acquire information. So how does that relate to this problem?
00:39:59
Speaker
Yeah, so this was a big narrative around the original sort of decentralized information aggregation and DAOs that the more tokens you have, you should be incentivized to do some work to figure out the right decision for some kind of vote. And so we tried to model that as well. And so we imagined a situation where you could exert some effort that had some cost to you to improve your information.
00:40:25
Speaker
And so every user could choose whether or not to essentially do some research that would make their vote more informed and make it better. And then we try to see, with this possibility, would this be enough to get you back to this sort of best possible equilibrium that you'd have with a central planner? And what we found is that it does help a little bit, but it can't get you all the way there because, again, there's sort of a mismatch about how much effort you want to put in.
00:40:55
Speaker
So what is the mismatch? Where does that come from? What prevents you there? That your incentive is not exactly, it's aligned with the platform incentive because you have some tokens and the platforms revenues we imagine are essentially redistributed to token holders and your good decisions increase platform revenues, but you bear the direct cost of doing this information acquisition.
00:41:22
Speaker
there's essentially like a free rider problem that you don't want to put in quite as much work as the platform would like you to.
00:41:29
Speaker
Right. In many ways, if I may summarize this, this almost seems like the same problem why janitors don't get paid in stock options. Because yeah, sure, by doing their job well, they increase the value of the firm, but not sufficiently relative to the cost that they have of doing their job well. The benefit is just not large enough from their particular activity to provide the right incentives. I think this is really an important message and an important insight.
00:41:54
Speaker
that is often overlooked is just because there's an incentive, that incentive is often not large enough relative to people's costs. I think this is, I think there's a general insight here for the, often for blockchain communities in DAO or DAO designers, if you want, that this is just not enough. Yeah, just to add to that and to what Brett said. So what we found with this, you know, we call it endogenous information acquisition model is that you actually do get some
00:42:24
Speaker
some of the benefits that people were talking about on blogs and various forums and whatnot. Things like people who actually have more tokens will exert more effort than others to acquire better information, which you could argue that might be a desirable outcome in some sense. But what was interesting was that when we looked at why this was happening, it actually turned out it wasn't because of the token-weighted mechanism itself.
00:42:54
Speaker
Even if you take that away, and so, for instance, you ignore people's token holdings and you just say, you know, everyone gets a vote, doesn't matter how many tokens you have, you still get those beneficial effects because the only thing you really need is the fact that people with more tokens benefit more when the value of tokens goes up, and that's by definition the case. And so that's the only mechanism you need to ensure some of these desirable properties that were listed in marketing materials and blogs.
Quadratic Voting: Application and Limitations
00:43:24
Speaker
And the other thing that came out of this analysis was, okay, so you can't get all the way there. But there was a question of whether, you know, there are other mechanisms like this one over N, you know, one person, one vote, one talking about it can do better. And the answer is not straightforward, but, you know, we did a lot of numerical simulations to try and answer that. And in general, what we found, but I wouldn't necessarily, you know,
00:43:54
Speaker
quote this as valid for every situation. But in general, what we found was that even in this complicated model where people are, you know, making decisions about acquiring, improving their information, everything, you know, there's all the bells and whistles, we find that one over one usually outperforms at least more than 50% of the time. And it gets much worse as
00:44:22
Speaker
the dispersion in people's token holdings increases. And so just intuitively to think about that, let's say there's massive differences between various people's token holdings. You can define dispersion in many different ways. But generally speaking, it's an asymmetry between people with not a lot of holdings, people with a lot of holdings. So as this measure of dispersion starts going up, what you end up getting is the fact that token weighted voting gets worse.
00:44:50
Speaker
Um, and, um, and you don't, you don't want to do that. You either want to try and limit token dispersion. You don't want too many symmetries in the various holdings of people who are voting. Um, uh, or if you can't do that, then, you know, the analysis just, well, actually you'd be better off simpler mechanism, which is to ignore people's token holdings and just give them, you know, one account, one vote.
00:45:15
Speaker
That's actually quite fascinating. So the subtleties of the holdings and the incentives that they create is, I think this is often, it seems to be something that people probably overlook and that sometimes the beliefs of it is just great if everything is fully decentralized. That may not actually work in the best interest of everybody, right? So that's, I mean, when we talk about fully decentralized,
00:45:38
Speaker
You can define, it's actually not, that's probably not a well-defined term itself, right? So, you know, simply, you know, power can still be decentralized when there's a few people who have more power than others. But, you know, the fact that individual holders, a few individual holders that have a high stake, if you want, in what's going on, could be beneficial for everybody.
00:46:02
Speaker
Um, now speaking of this and, and speaking of the, of the particular voting aggregation mechanism that you had. So you refer to it. We have, we have talked a lot about the one share or one token, one vote, um, arrangement, but there are some other ideas out there that, that the people have thought about them, people have used. Um, and I think you've done some work on that too. And you may want to talk about that. Yeah, sure. So, um,
00:46:30
Speaker
What a voting mechanism that people have talked about a lot in the blockchain space is this thing called quadratic voting that was proposed actually outside of the blockchain space originally by Lally and Lyle. And the idea of quadratic voting was in the context of voting or in funding something you can imagine instead of paying linearly for your vote weight or for the outcome that you want.
00:46:59
Speaker
What if you have to pay for the square? And so this is where this quadratic comes in. So for token-weighted voting, you can imagine this.
00:47:08
Speaker
If I want to participate in a token-weighted voting system, where it's one token, one vote, I potentially lock my tokens up in some voting contract. And so there's some opportunity cost of staking in this voting system. And that's proportional to the number of tokens I have, which is now proportional to my voting power. So you can think of this as costing. My votes are costing me linearly in how much effort or how much influence I want to exert.
00:47:37
Speaker
But you could ask this question, what if I staked V tokens? I didn't get V votes. I got square root of V votes. Then this would basically, another way to say that is the cost of voting at some weight is the square. This is where this quadratic comes in.
00:47:58
Speaker
sort of one argument for this is that this can sort of flatten down these big disparities in token holdings. Now, if I want to, if I have 90% of the tokens, I don't have 90% of the vote anymore, I have much less, right? And so it really kind of pushes down on the big whales. And so the early votes or the early works on this quadratic voting showed that in some settings where people have kind of
00:48:27
Speaker
perfect information about their own preferences that quadratic voting can be optimal. And so we wanted to ask a similar question to what we asked before about as an information aggregation mechanism. So if you have some DAO platform and you're imagining, again, that you're voting on some parameters, and as a simplification, we imagine that most of the token holders are aligned on what they all want to choose the parameter that will result in the highest revenue for the platform.
00:48:57
Speaker
And we want to figure out a mechanism that will aggregate information. Well, will this quadratic voting aggregate information sort of better than either linear voting, so one token, one vote, or one person, one vote?
00:49:11
Speaker
Yes, if I may jump in. As I remember the early results, I think there was a paper in the AER a few years ago that studied that particular problem. The kind of question that people had at the time was, is this a good way how you can aggregate or bring out the decision that actually everybody prefers? Some people may have stronger preferences.
00:49:32
Speaker
for a particular outcome. And that is a way how you can express what it is that you're really after. And so what you're actually asking a related but interesting question, which is, so people have information. And so is quadratic voting actually a better mechanism maybe than one token, one vote to bring out the truth? Because that's what you're after, in particular in the light of incentive problems and so on. So is it? What's the outcome?
00:50:02
Speaker
So the answer is, sometimes, in the problem with quadratic voting, you don't get a clean result like you got in the previous works, where in the previous works, as you just described, where people have different
00:50:24
Speaker
strength of their preferences. Quadratic voting is a way for them to sort of exert the strength of their preference in a nice way. If we imagine that people have
00:50:36
Speaker
again, sort of this overall preference to increase revenue for the platform, but they have different sort of strengths of information. Now the quadratic voting again has the same problem that we saw in the first paper, which is that your votes maybe don't align with your information. And so you really want a system where the people with the most information have the most voting power. But now under quadratic voting,
00:51:05
Speaker
Actually, it costs, if you have a lot of information, it costs you way, way more to exert NEF influence on the vote. And so quadratic voting does all right when the information is sort of well distributed because quadratic voting is kind of forcing the vote weights to be more distributed. It's saying we're not going to, it's going to be too costly for a whale to exert a lot of influence. But actually, if somebody has a lot of information, you kind of want them to exert more influence.
00:51:34
Speaker
Um, and quadratic voting sort of kills that. So it almost seems like instead of having a proxy vote or giving your votes to the whale, you actually have to do it the other way around the whale would have to give you their vote so that you can get this out. And yeah, but, but this tends to be disallowed in quadratic voting because quadratic voting essentially always assumes there's some identity mechanism because even in the case, well,
00:52:02
Speaker
In the case where people have really distinct preferences, I can always get around quadratic voting by kind of splitting my votes. So instead of buying 10 votes at a cost of 100, I can split myself into 10 people and make 10 accounts and buy one vote in each account and get vote weight 10. And so to avoid this kind of end running quadratic voting,
00:52:30
Speaker
basically every quadratic voting system assumes there's some sort of identity management. And if you sort of allowed people to split their identity, then you kind of completely undermine quadratic voting because everybody can just vote one now and one squared is one. And so now you've recovered token weighted voting.
00:52:46
Speaker
But if I look at this, so if I take your first paper and I look at this paper, it seems that the number of cases when quadratic voting, even though we touted it as something which is superior and clever and new, is actually superior in terms of information aggregation. That seems very narrow, right? It almost seems like a harmful approach. So I think that's for the design of voting seems really quite important to understand this, right?
00:53:11
Speaker
Well, I would say it's just caveat, not always, because there are situations where it works. Sure, but if you do an exam test, I mean, it's important, Jerry, to say this, but if you have to design a system ex ante,
00:53:23
Speaker
You don't know what the possible outcomes are, Exante, in terms of distribution of information and the like. So if you had to make that choice, if you had to make a choice, Exante, would you go with, I'm going to push you now, put you on the spot. Would you actually go with one token, one vote, one personal vote or a quadratic? Yeah. I mean, that's, I'll preface my answer by saying, I don't know, but I'll give you an answer anyway.
00:53:48
Speaker
Just to make, before I answer, just to make a quick connection to the previous paper. The previous paper, so one of the takeaways is dispersion in token holdings is bad when you have this token-weighted aggregation mechanism. And you'd want to dampen that dispersion. Why? Because you might reduce the inherent mismatch that probably exists between people with tokens and people with information. So dampening dispersion, in general, as a theme, when it comes to these voting mechanisms,
00:54:18
Speaker
might be the way to go. And you can think of it, you can think of quadratic voting in that lens as a mechanism that dampens dispersion, in the sense that people with more voting power under the previous mechanism, now you'll take the square root of that, so their voting power is reduced. But then the question is, why would a square root function be the optimal function or the optimal way to dampen dispersion?
00:54:42
Speaker
And what we find in this paper that it isn't. It's not necessarily the optimal way. There might be other dampening functions that you could use that might not be exactly the square root, might be something else. And so I'm not going to discount the concept behind quadratic voting. What's the concept behind it? It's that it shouldn't be one token, one vote. There needs some adjustment. There needs to be some adjustment.
00:55:08
Speaker
Now, that adjustment might not be, might not optimally be the square root when you're coming at this purely from information aggregation perspective, which is what we do compared to all the other papers. And it might be something else. And so I'll give it credit there. But, you know, a lot of the analysis from the previous paper, you know, tends to hint in the direction of one over n voting seems to work a little bit better.
00:55:36
Speaker
if those are your two options, but I wouldn't bet money on that.
00:55:42
Speaker
Do a lot of work. Wait, so then, if I can just ask here, it sounds like the way information is formed ends up being fundamental, right? So if you kind of assume that ex ante, everybody is equally likely to have the same distribution of information or something like that, then I can kind of see this point that you're saying, like, we want to essentially dampen what the actual exposed token distributions are and kind of get back to that initially.
00:56:11
Speaker
Is that channel behind a little bit of what you're describing or am I sort of missing over here? Yeah, I think that's partially true. It is a little bit more so in the first paper it turns out it is a little bit more complex than, you know,
00:56:32
Speaker
Does it only work when everyone is homogeneous? I think is what you're saying in terms of their vibe of information. Yes, the ex-ante. That they have. And yeah, I mean, there's some of that. And a lot of the interesting properties are coming out from the fact that people aren't. And so there's a lot of, in practice, people just aren't.
Security Concerns and Decision Alignment
00:56:56
Speaker
And so there's a lot of weird things that can happen, including things that are counterintuitive, like
00:57:02
Speaker
even if you just take two different people, one's better than the other. Should you give them all the power? The answer is no. As long as both people have a little bit of independent information, you are better off by using both, even when one's better than the other. And in fact, if you give too much power to the, even if the person's 10 times better than the other, one person's 10 times better than the other person's,
00:57:29
Speaker
There are situations where if you give that person too much power, you're still worse off. So heterogeneity in the type of information is extremely important. It's weird, it's hard to analyze, and it's responsible for a lot of the unexpected outcomes that are happening. And so when you don't have that, things tend to simplify. But I just want to add there's sort of another angle to think about this. So we've been focusing on this information aggregation angle.
00:57:58
Speaker
You know, a lot of the discussions around these things are you're worried about some kind of maybe hostile type takeover or some kind of malicious attack. Right. And that's where things like potentially token weighted voting or quadratic voting do sort of naively do better than one person, one vote. Right. If you are worried about somebody coming in and buying a lot of tokens, a competing protocol coming in and buying a bunch of tokens and specifically trying to, you know, choose the worst parameters for you to tank your protocol.
00:58:28
Speaker
One person, one vote potentially does badly there and token-weighted voting provides some protection there because you have to buy some amount of tokens and then quadratic voting potentially provides even more protection there. That's kind of a separate motivation. You have to think about how much are you trying to protect yourself against this kind of external attack versus how much are you trying to elicit information from
00:58:59
Speaker
Yeah, but you can separate that. So for instance, when you're designing governance systems for Ethereum, for instance, this would be probably your primary concern. The security of the system would be your primary concern. You'd have to go about that a different way. When you're voting on an interest rate for a specific protocol, you might care more about the information aggregation aspect. And so I think it's...
00:59:27
Speaker
It's hard to talk in generalities with these systems and DAOs in general, because different DAOs have different objectives and might want to focus on different aspects. One of the things that I find really fascinating about this space is that it is also a giant playground or experiment, if you want, of trying out different tools and mechanisms in order to simply aggregate opinions about a particular policy that has economic implications.
00:59:55
Speaker
We do have a very large area of finance where this is actually already used, which is you have shareholder voting. And I think many of the problems that you describe here would translate potentially also into that space. So for instance, you have the question of takeovers. I think a much bigger concern these days is that of common ownership, where common ownership can lead to collusion, where firms actually make policies which are
01:00:23
Speaker
are anti-competitive simply because that allows them to extract more rents from the market. And so there's a lot of these questions out there. And I think in some ways what you're describing, we can use these insights possibly to say more about that. I mean, there's clearly some parallels between shareholder voting and token-weighted voting. Mathematically, if you abstract away from some of the nuances, you end up with a similar structure, basically. But shareholder voting might have some
01:00:53
Speaker
differences, the aggregation mechanism might be a little bit different. You might require a different type of majority threshold or other things like that. And the interesting thing for shareholder voting is that I don't know how much power shareholders actually have to run the business. The type of stuff they tend to vote on is not necessarily day-to-day operations. And in the blockchain space, it could very well be that.
01:01:22
Speaker
And so there's a strong difference there, in terms of the amount of information people need to have, the amount of effort they need to spend to get involved, understand the operations of a platform, the blockchain space is not comparable to what's currently going on in shareholder voting. Of course, that could change, but the lessons that we're getting from these, I should mention, these are some of the early studies that we're aware of,
01:01:51
Speaker
really formally looking at voting systems, specifically for decentralized blockchain systems. The insights we're getting is that it's not easy in general. If your objective is this perfect situation where you have a central planner that knows everything, yeah, it's not easy to get. If that's your goal, which doesn't mean that it's necessarily bad, but you can't just, it's just not easy. So I don't know if there are lessons to be learned
01:02:21
Speaker
But if there are, you know, it would have to be the case that shareholders become more active in the day-to-day types of operations of firm, which is hard to envision when you have an entire corporate entity in the bigger firm, the more difficult it is to envision. In many ways, the blockchain space. Think about startups. There are less things to kind of necessarily know, and it might be more adequate for
01:02:48
Speaker
that type. All these are really interesting. I just want to go back to one of the things that Jerry said about there are different types of votes. So we modeled this information aggregation. And again, I think this running example of setting an interest rate is a good one. And that's something where an attacker really probably doesn't have that much power to do a lot of damage, to buy a ton of tokens to set an interest rate to be suboptimal. Whereas a vote for something else like
01:03:17
Speaker
You know, a lot of these DAOs have an emergency shutdown feature or something where there can be a governance vote to just shut down the whole thing. Right. And that might be something where you're much worried about, much more worried about a sort of hostile takeover. Right. And so you potentially want a different voting system for these different types of things, maybe for the day-to-day operations versus for, you know, some kind of, you know, something that really could damage the system.
01:03:44
Speaker
One thing we hear a lot from the blockchain community is the advantages of decentralization.
Improving DAO Governance
01:03:50
Speaker
And in the context of what you're doing, the connection is the fact that you can have more people providing with independent information potentially. So you have a result on that. Could you talk a little bit about that? And maybe practically speaking, whether you do see potentially advantages to sort of more participants in these sorts of systems?
01:04:13
Speaker
across many of the different papers we've written, it turns out that the more people you get to participate, and this will not be a very surprising result, the better it is for the system. And so getting people to participate is quite critical. And some blockchains have taken an active view of this. They actually incentivize people, even via rewards, to get them to participate in governance. And so I think that's
01:04:40
Speaker
You know, if I were to summarize all of this research into one sentence, it would be more people get with independent viewpoints.
01:04:50
Speaker
Okay, well then, since we're, there's one more question that I wanted to ask though. So we've been talking concretely about, for example, interest rate setting protocols at say compound or AVI. So staying on the sort of concrete side of that, what guidance would you give for either compound or AVI in terms of actually how they should think about modifying their governance processes going forward?
01:05:17
Speaker
Well, okay, so I think one thing that I've been generally excited about, about the blockchain space, is that it's kind of an open slate about how you design your governance system. You have huge amounts of flexibility when you deploy a DAO about choosing a different kind of voting mechanism. And we have started looking at different ones.
01:05:38
Speaker
the paper we've talked about the most, we focused on this thing of what you call like direct voting, where the resulting interest rate would be some weighted average of the votes. That's not the system that AVE or Compound currently use, although it is used for other platforms. Compound and AVE both use this sort of yes-no vote
01:06:00
Speaker
call initiative voting, amendment voting, where somebody proposes, I think the interest rate should be 4%. You get to vote yes, no on that. And you have some window, you have a seven-day window to vote yes, no on this proposal. There's also alternative things. MakerDAO does what's called approval voting, where anybody can propose, say, an interest rate. And they're all active at any time. There's no seven-day voting window.
01:06:30
Speaker
And you get to approve of saying, I would approve of a 3% interest or a 4% interest or a 5% interest. And then the one with the most approvals is selected. Not already is a very different mechanism than saying you have seven days to vote on this yes or no, and then somebody can vote something else. And this is still a very small fraction of the type of voting mechanisms that have been explored in practice. And so I think this is still kind of a wide open space. And so we've been trying to kind of go through
01:07:00
Speaker
and characterize things. And so I'm sort of hedging here because I don't think anybody really knows right now what the optimal mechanism should be. But I think there's room for a lot more thought about this. And so I think a lot of people, you know, Compound came up with its governance contracts and a lot of people just forked those contracts because it was easy from a coding standpoint. But I think we should be thinking a lot about sort of what different voting mechanisms work better.
01:07:30
Speaker
And personally, I think this for some things like these system parameters, these direct voting does seem to be better than the binary kind of voting we see. But that doesn't mean it's sort of the optimal. There's still all kinds of other mechanisms.
01:07:46
Speaker
Well, this was certainly a very fascinating discussion, very much in depth. There is clearly what I learned from this. There are a lot of nuances in the organization of voting for DAOs. I kind of feel like now I have to adjust my slides that I have for my students.
01:08:02
Speaker
So yeah, but that's great to do because it shows that we learned the book is not closed. There is much to figure out. And I'm most grateful to you, Jerry and Brad, for enlightening us and the audience on this important topic. So thanks so much for listening.
01:08:24
Speaker
We hope you enjoyed this podcast. Thank you for listening. As a reminder, you can find additional materials on owlexplains.com and can stay updated by following us on social media. That's all for today.