Voting with time commitment by Vijay Mohan | Devcon SEA

[Music] uh hello thank you all for uh coming for today's presentation uh my name is Vijay moan I am an economist and an independent researcher uh in the space um I think following um Dr Hans Hansen is a pretty hard act but uh anyway I'd like to share some thoughts on voting with with time commitment um that is how you use time commitment for uh blockchain governance um so just to give you some background uh we are all familiar with token based voting which is that you have a certain number of governance tokens and based on the number of governance tokens you have you get voting power in proportion to that so this is something we all quite familiar with but when you go to the literature both the academic literature and in fact blog post by practitioners there are three problems that keep uh cropping up so the first one uh is a plutocracy so by a plutocracy I mean that there are certain token holders who are whales they have a disproportionate amount of those governance tokens and they're able to um influence the decisions uh from the voting process the second one is a civil attack and this comes because a blockchain is snonymous so it's sort of the blockchain equivalent of voter fraud um you basically create multiple accounts uh with fake identities and you do that because the voting mechanism you have gives you some advantage of doing so so that's the second problem and the third one is vote buying where you buy tokens just before a vote um the idea being that you buy those tokens to control the vote and you sell them immediately after right so you're you're sort of buying that proposal and it's all happens very quickly you buy it you vote you hijack the proposal and then you sell and what we are finding is that a lot of dows are adopting uh some voting scheme that's out there could be one token one vote could be quadratic voting they've heard about it none of these were you know necessarily blockchain based things um they adopted and then what they find is later on six months one year two years down the road you're left with one of these problems right either you find that so for example if it's one token one vote you would expect that you know you may end up with uh a plutocracy and that's the problems that you end up finding and then there's a lot of talk about opes we've got a plutocracy here or our voting mechanism has been Cil attacked so I mean the question we need to ask is what if we took one step back looked at the big picture and said if I wanted to develop a voting mechanism from scratch how do I think about it so that I minimize the chance of these problems uh coming up at the end um and that's what um I tried to do with with uh two co-authors payman kzer and Chris Berg we wrote a paper which was published with this year where we essentially asked that question um if we start thinking about these problems right at the get-go and build things from scratch um and we think about all the ways in which these things can be linked for example how is a plutocracy linked to a civil attack uh how can we reduce vot buying because we know we're going to end up with these problems at the end um how would that look so that's sort of the problem we attacked and what we ended up doing was well we proposed a new voting scheme um which said that well this might have a lot of potential in circumventing those problems and have even additional uh beneficial features so it's sort of a new voting mechanism we proposed I would love to hear your feedback at the end of it of course but uh that that was one of the main things about this paper and overall we really wanted to emphasize this idea of time based sufferage so there's this paper by I think it was KS and layers if if I remember correctly where they identified a lot of different voting scheme suffrages they said well there's token based suffrage there's meritocracy and there identity based conspicuously absent was time-based suffrage and we really wanted to emphasize that look time has value for a voting mechanism on the blockchain um so I'm going to get a little bit technical because but it's just basic technicalities that I need to get across um in any voting token based voting system you're going to have to find a way to connect the number of tokens you have to the number of votes you get and that's all this function is saying it's just saying that you've got to take the number of tokens and come up with some scheme to convert it to votes so the simplest one is one token one vote where that function is just V of Q is equal to Q every token you have you get one vote for it but you could go to town with this problem right you could say that I'm going to square it I'm going to cube it I'm going to take the logarithm of the number of tokens I have I'm going to take an exponent you can do whatever you like with it so depending on what you do you are going to affect these problems the Civil attack problem and the plutocracy problem and our goal here is to put some structure into this problem that if you go to town with this and you say I want to make a log of Q as my voting mechanism sure go ahead and do it but what problem are you going to end up with how do we think about it right from the beginning so what we show in that paper is that under very reasonable conditions by which I mean when you ruse out crazy stuff like collusion and things like that all you can do to to prevent a civil attack with that basic transformation from tokens to votes that you've created is to make sure that that function is linear in tokens any if you curve it the wrong way you're going to end up with a civil attack so it's got to be linear in tokens we also show and it's sort of makes sense right if you think about an intuity let's suppose you have 10 tokens and everything is linear if you take the 10 token and split it into two accounts of five each well 5 + 5 is still equal to 10 so as long as you maintain that linearity there's going to be no incentive for you to open multiple accounts so that's the simplest way to tackle that civil resistance problem however with plutocracy resistance you need to put some curvature on that function you need to sort of make it strictly concave by which I mean that every token that you vote with additional token that you vote with you get more votes but each additional token gets you less votes than the previous one it so it sort of tapers off and um and that also makes perfect sense if you want to control a plutocracy and make sure people don't you know maximize the power of their tokens too much every additional token that they put in the influence of that should taper off and that's what this says but if you think about it you cannot have both simultaneously you cannot have a straight line function which is linear the first one for Cil attack and a curvy function to get rid of plutocracy which is the second one and that's what we call the Civil um uh uh resistance or the protoc reible Dilemma which is that you got to pick one right you either you if you're going to make it straight if you're going to make that function a straight line you you'll get rid of a civil res attack most likely but you are going to end up with a plutocracy if you put a curve on it you'll control a plutocracy but be prepared for a Cil attack and that's the Dilemma that that you're faced with with a pure uh token based voting system so all blockchains that are based on token that that have token based voting uh have only two options they either have to the way they structure that conversion from tokens to votes they can either do it in a manner which gets them plutocracy resistance but they've got to find some other way to control the Civil problem because that plutocracy resistance only comes when you curve that vote function a particular way or you can make that vote function linear in which case yes you've made your vote function as Cil resistant as possible but then you got to find some other way to tackle the plutocracy so it's not enough to have v as a function of Q alone you actually need two resources or two instruments and this is very well known in Economic Policy right so when you're looking at an economy and you've got two national policy targets say you want to control inflation and you want to control unemployment you cannot achieve it with one instrument like interest rates because they're moving in opposite directions you need two instruments and that's exactly the first Insight we want to give is that you can't make just V uh v as a function of Q alone and Tackle both these problems you need a second instrument youve you tackled one problem you need a second instrument to tackle the other one so the second instrument that we are proposing apart from Q is time commitment where when you vote you lock up your tokens for a certain amount of time as well right so basically you've got two resources now you've got two voting instruments you've got q and you've got T and one attacks one problem the other attacks the other problem and we're actually quite familiar with this because curve does use that sort of um um logic in its voting function so for those of you are familiar with curve it's just Q * T is what they're really doing where T I'm measuring is a quarter of the number of years that you lock and of course you could go crazy with this right you could say that I'm going going to lock one token for a million billion gazillion years and I'm the king of curve you well to prevent that they've capped it off they've said that four years is the maximum and that would make sense right so every time you want to have a time commitment here we want to cap off the time commitment so that one person doesn't put one tenth of a token for a trillion years and somehow gets a lot of votes so the time commitment there is capped off at four years but curve also introduces a lot more strictness the the the vote escrow curve token that it introduces is non-transferable you cannot get out of that time commitment it's absolutely locked so they've got these other details that are happening behind that uh vot function but if you just look at it it's pretty nice right it's it's linear in Q so my guess is that this is going to be pretty Cil resistant and it's got time so if you don't have tokens you can use time um to to sort of you know fight with someone who does have tokens so it's sort of ticking the box there and and and I like this when I I first see it um and any type of time Dimension that you add to the voting process you are going to provide a disincentive for vote buying because the idea of vote buying is that you buy tokens you lock you vote and then you immediately sell them and your whole purpose is to control the voting if you put any type of dime delay even if it's a two-day dime delay or four years like um uh curve does it is a disincentive for vot buying because you've locked your tokens and they can't get rid of it immediately so there's an opportunity cost for them to keep those tokens locked so time commitment has two immediate benefits the first is that you've got this other instrument with which you can express the intensity of your preferences and second you've got protection against vote buying so out of those three problems time commitment almost immediately solves one which is vote buying um and the other nice thing about time commitment is that you cannot launch a civil attack with time if you're going to create multiple accounts you cannot popul populate those accounts with time you got to populate them with tokens so time cannot be exploited for a civil attack the way tokens can be so we don't have to worry once we've introduced time into that vote function we don't have to really worry about time being used for a civil attack your focus is always on tokens being used for a civil attack so the overall picture that comes now is let's structure our voting function in such a way that it's linear and we minimize the chance of a civil attack and then you use time to counter the plutocracy two different instruments acting in two different ways right so you structure it so that the one which can be used for a civil attack you eliminate that possibility the other one time which cannot be used for a civil attack you let that do the plutocracy so if I go back to that vote escrow that curve has like I said it's pretty nice at the face of it but the devil's always in the detail there are even though curve has got a lot of strict um rules behind it that um you you cannot transfer it you cannot get rid of that time commitment actually you can even if you don't sell the vote escrow tokens you can actually sell your uh claims over the lock tokens to someone it's over the counter and it's quite complicated right but it can be done there are some articles I think by open Zeppelin which describe exactly how you can sell your claims over those lock tokens even if you can't sell the lock tokens themselves that's one way to do it the other way to do it to get out of what you know that that time commitment for which is completely inflexible is to add another layer and I think that's where convex came in they said look it's completely inflexible let's add another layer which adds flexibility right and then after convex came vum after vum came someone else and pretty soon no one knew what was going on who owned what who was voting for what and we all gave up trying to figure that particular problem out so yeah you can circumvent that extreme time commitment in certain ways but wouldn't it be just nice to be able to get rid of your time commitment for whatever reason you need to appease your spouse you need you know funds for school you don't like the uh that platform anymore get rid of the time commitment and move on so we think that flexibility is a good thing that the strict thing that curve is introduced is too strict and that's why you've seen all these layers being added to that and also if you think about the fact how the way vot escrow is structured once you've committed for say four years you can vote again and again on proposals with those lock tokens there's a DK rate and all that but basically you don't have to think about what are important proposals anymore you just keep voting you know time after time with that power of wealth that you have uh without actually having to think about the proposals so wouldn't it be nice if all the voting took place on a proposal by proposal basis so that people actually have to think especially Wales and other people who have if you're going to put all your monetary backing behind one proposal what it going to meet for the next one right so and of course given the first point if you can get out of it well maybe you will incur some loss but if the next proposal is valuable enough you have to think about it at this point with curve you don't and of course this is a very this is linear in both ways it's linear with respect to Q which is a good thing which gives you Cil resistance but it's also respect with with respect to T which means that if you really want people to encourage you know encourage them to commit for the long term this doesn't scale well it's like every additional unit of time it linearly increases why not give that more power if you if your da is interested in long-term commitment why not give as much power as possible to long-term commitment so we started off with these three problems plutocracy civil resistance and vote buying but once we went into the um curve mechanism there seemed to be a lot more sorry um desirable features that we could add we could give that t function part of it more power to justify long-term commitment uh we could make it proposal by proposal so that everyone has to think about where are you going to put your money um if you don't like it at the end after 6 months and you think that oh my God this platform is going the wrong way you can get out of that time commitment right you don't have to rely on convex and this and that and complete you know chaos to be able to have flexibility it's built into the problem uh and there is some um mechanism for information to be generated from the voting process which I haven't talked about before uh and I'll talk about that briefly in a minute so the pro the the idea we came up with at the end is something called Bond voting and I won't go too much into detail into this um the paper is published and anyone who's interested can take a look or just come and see me and I'm happy to chat about in uh but what we did was we started off with um a zero coupon Bond like a treasury bill uh and said that well you get a coupon payment at the end you buy it at a discount um what if you did that with voting you get one vote and how much you commit determines how much of a discount you get uh in terms of how much you have to stake uh in order to vote so we played around with that idea for a while and this is the formula that we ended up with uh where it's linear in q but it's exponential in t right so it it should be fairly Cil resistant because of the linearity in q but that exponent in t gives you a lot of incentive for long-term commitment because you get a lot of benefit from it and R is an interest rate parameter that's fixed by um the bond governance uh the the the blockchain governance and intuitively the way we like to think about it is that suppose you commit a certain number of tokens Q for a certain amount of period t uh think of it as the as the platform issuing a voting bond for you and you can just think of that as an nft so you get an nft which says uh this is the amount of tokens you've committed this is the time for which you've committed it for and the reason I want to think of it as an instrument is because once I've got that nft if I don't want to commit anymore I can sell that nft I can trade that instrument in a secondary Market the market will decide how much to pay for it but it will let you get out of that commitment with a maybe a small market-based penalty it's not the platform introducing any penalty on you it's what the market thinks those tokens are valued uh that you've locked up before what they think they valued now right so that gives you a sort of a way to get out of the commitment it's tradeable in a secondary market and an important thing about a secondary Market is that it carries information so when you look at treasury bills or bonds you look at the slope of the yield curve to think about what is the economy how is the economy performing so when the Yi curve turns negative you think that well maybe a recession is oncoming you can have that same sort of logic here once there are enough maturities of these boating bonds you see the sort of implicit return that you're getting on them the interest you're getting on them and if that is sort of downward sloping yield curve that you get it tells you that the market view is that the blockchain is not in good shape so there are informational properties the same way we get information from Bond trades you can get information from voting Bond trades as well so it taks a lot of these box and here I just had a quick um diagram not particularly useful I guess at this point but uh the green one is your V escow it's a straight line with respect to T I've got time commitment on my horizontal axis and here I've only had 18 periods but the red one is Bond voting and if you keep extending that it'll just escalate and exponentially go up so it starts slow but it really rewards long-term commitment right um so just to summarize the benefits of bond voting um it's linear in token so it should be civil resistant uh you've got time to fight the plutocracy problem time commitment solves the vote buying problem you've got you know t as an exponential function so the power of time commitment increases very rapidly you really reward long-term um you know commitment to that platform um we actually solve for the Dynamics of this as well and say that even after you lock up if a new proposal comes in and um you know you really want to vote for that one what would your reaction be so we sort of think about the Dynamics and justify why proposal by proposal voting should work in this scheme staking with flexibility you you you've got sort of this nft that you can get rid of whenever you think that commitment is no longer valuable you don't need convex and VM and all these things to come up on top of that and when you've got this happening for a large amount of time you will see that there are informational properties that you can get out of it just like we get information from bonds so uh that sort of summarizes what I would like to say and uh thank you again for coming today and if you have any questions please reach out or just want to chat please reach out to me thank you epic stuff VJ thanks very much uh so audience don't forget to scan the QR code and uh submit any questions that you might have or upvote the ones that exist uh so first question there uh are there any examples of voting mechanisms which uh involve could involve eth uh I mean it's a do by I'm not sure I understand the question exactly but yeah I mean if your token governance token is eth in any way um or your platform whatever you know your DA has as the governance token um your voting mechanism is based on that so yeah it could involve eth in fact there are actually very clever ways of doing it I don't want to get into too much detail now but you could sort of make eth a numer for your Dow and then instead of using your token use the eth token with the numer but then you got to worry about your token versus the e token exchange rates and stuff like that so actually that's a very good question there are answers to it but it's it's a bit involved yeah good well I'm I'm glad I'm glad the question hit at the end there uh okay so maybe just a maybe a clarification uh at one point one of the audience members thinks you describe collusion is crazy crazy unlikely uh no no it's it's sorry maybe crazy was bit extreme but it's undesirable right so we we if we trying to get rid of undesirable things like civil attacks and plutocracy we don't want to then create something which is subject to something else like collusion right because collusion is also a bad thing you can get actors uh combining together to hijack the voting and people with small token Holdings will then suffer from that so you want to avoid I mean we didn't explicitly think of collusion as a problem but when we when we ended up with the fact that civil resistance needs a linear function we thought of getting rid of collusion as a as one of the reasons why it has to be linear got it thank you uh in your opinion would incorporating a tool like World ID or ZK passport which offer Cil resistance allow you to run a fair vote using only a function like log Q sorry sorry can you repeat that yeah so that's the question's up on on the screen there so it is would in your opinion would incorporating something like World ID or ZK passport something which offers Cil resistance allow you to run a fair vote using a function like log Cube yeah absolutely so if you've got log Q it that is going to be plutocracy resistant in my opinion that's strictly con it is going to be but you do need something else and that could be something like proof of personhood right so yes if you've got a pop protocol or something in the background you need another instrument then to take care of the Cil resistance log Q itself will guarantee you plutocracy resistance but but then you're you know how how are you going to um guarantee the Civil resistance you've got to have another instrument in fact um proof of person Protocols are being investigated for that purpose how to use it to avoid a civil resistance but you've got to have that instrument and that's perfectly valid right so if you want to give log Q as your curvature to Q uh in that voting function you've got to have a second instrument that could be uh proof of person Hood proof of identity ZK whatever you're able to devise that second instrument has to be there got it yeah that makes sense okay couple more questions there then uh is there anyone using Bond voting in those uh no absolutely not uh at this stage this has come out of um our um head at at this point uh but if some da does want to use Bond voting we'd be thrilled and more than happy to you know figure out how it can be translated to the real world brilliant uh and I guess maybe we'll say this this the by one can the can this time lack be manipulated by an attacker tricking tokens into locking up for too long than using smaller amount of liquidity to maybe pass dangerous legislation um tricking tokens into locking up I'm not sure how you would trick and then using smaller liquidity yeah I mean but the whole point of um that can happen in in time lock in the sense of uh vote escrow but how would you trick when you can get so the the tokens that are being tricked in this case in this example if you can get out of that commitment you won't remain tricked right so if you can sell that token commitment uh and move on then how do you remain tricked so that's the whole point of having the secondary Market that you cannot have stuff like this if for any reason whatsoever you feel like you need to get out of those tokens maybe you're being tricked you get out of them got it yeah I guess there would be a discount though right there would be a pretty big there would be a discount so if you're going to if you've locked up Q tokens no one is ever going to pay more than Q for claiming Q at the end so there will be a discount but the discount is based on the amount of time left and uh what they think those Q tokens are going to be valued at the end when they get it so it's based on Market perception got it yeah okay now nice note to finish on well thanks again VJ Round of Applause uh great great talk thank you thank you very much thank you