Decentralized Governance at Scale

Transcript

okay so this is me and i want to speak a little bit about governance but let me tell you a little bit about myself so i'm a professor in bengal university in israel which basically means they have a research team and we write papers and we apply for grant money if you want to see some paper you can look at an immortal one at dblp it's like imdb for papers my main research is in computational social choice and slower sure sure because i drink coke then so my main research is in computational social choice which means that i do game theory mechanism design and also some kind of algorithm analysis or combinatorial optimization

i also help this nice company named daostat that maybe some of you know about so let me tell you a little bit about my my root or what is computational social choice which is my main research field so basically in computational social church we try to understand algorithms and protocols that will help a community reach decisions basically so stuff like preference aggregation like choosing a committee or how to divide resources like joint resources of some pool of resources or do matching with preferences and we do this by mathematical modeling and some game theoretical analysis and we try to do it in a principled way that will allow us to develop good algorithms for community

decisions and then some time ago i was reading more and more about those and my feeling is that somehow the ecosystem is still largely immature in terms of the governance capabilities of those and especially this is i mean i could observe it by the limited scale of current dows or like the decision making process of kubernetes and also by the nice introduction that hillary right so hillary gave before about the low participation rates in a lot of dows and and i think that one way to to solve this and basically to be able to scale those even

larger and to a better extent is to scale the dow governance that we use and so what what are the like the most standard dow governance systems that people use today this is what they call a reality where reality aware proposal based direct democracy basically means that you have the status quo or the reality this is like the current set of rules and now people can give proposals and each proposal is basically a binary proposal so either to keep the status quo if the proposal is rejected or to accept the proposal and then the status quo changes roughly like brexit so if brexit passes then the status quo changes so the reality changes and if

brexit does not pass then it does not pass and the status will remain which is a nice system but it's very simple and it has some um inherent scalability issue because if you want like a very big doubt so this is what i think i'm thinking about i'm thinking about a really huge dao that wants to be flexible in his in its governance capabilities so there are going to be many many proposals all the time and basically down members are now faced with this um increased train of proposals all the time and maybe i mean this is one of the reason i think that the participation rates are quite low because it's simply hard to do this

so in a way what we need to deal with is somehow this conflict between the desire to make many proposals to make the dao alive and flexible in its governance capabilities and this this is conflicting with the limited time that actually people have to um devote to participation in such decision making processes so there are some other ideas that at least from my perspective they try to solve it at least partially so companies this liquid democracy this concept of delegation which is also used in other settings and this in a way i mean if i am a member in some community and there are many many proposals that are coming so many many decisions that i have to make

but if i can name one delegate or several delegates that will make the decision on my behalf then this basically saves a lot of time and attention on my behalf there's also this nice idea of optimistic voting and also graphic consensus by the australia and there are many many others if you know of others i will be happy to hear about them later um but but still i think that the current state of the art uh at least in the dow ecosystems that i mean there are some solutions but they are not really sufficiently well understood and this is partly because the way they are developed is not in a principled way so and this also causes the current situation which now governance tools are not optimal in a way

and what i'm trying to imagine is whether there are ways to develop better governance systems for those that will somehow solve this conflict between flexibility and limited attention or limited time and better and this somehow relates to this field of study computational social choice which i'm coming from because there we basically try to view such settings in a very abstract and formal way and try to develop um properties that we want those systems to to have and and and really find like better ways to analyze the properties of this system and optimize certain

behaviors that we want so this is the way i view in this kind of problem so the problem of doing governance at scale so with really huge communities and really flexible decision making and what i suggest and what i try to do is basically the following blueprint so i think that it should start with some everything is okay okay good so so what i suggest is something like this so we should start with some some mathematical modeling of the problem very very abstract then coupled with some data analysis and this will allow to really have a unified way of presenting certain solution approaches and then when we have these solution approaches

formalized in a unified way due to our mathematical modeling and using the data analysis that we used then we can do some simulations and then we can write prototype and see really like the properties of this system so not simply use whatever tools we could find of the shelf and so let me let me delve a little bit into each of these things right so mathematical modeling looks something like this so you you look at the system you you [Music] you lay down on the sofa you you close your eyes and you think what the system should look like and then you write something like this um which is actually interesting also from the scientific point of view because as i told you in computational social church we do this kind of things so this

is just okay this this is that castle in germany and this is lawrence center in university of london where i just attended to conferences of this community so i mean this kind of people they know how to do this mathematical modeling but still some ingredients that are used for our setting um are still missing like with the i mean this problem is much involved because the communities are dynamic so you want the solution to be robust even to the setting in which community members come and go and also the decision making is should happen in a continuous way and usually works in this field are rather discreet or even like one time so you imagine a community

and you want to make money like one decision and you want to find good algorithms for this one decision so there's some things to be done like on the purely scientific level okay but this is the like the first part of mathematical modelling after we have this we can do oh actually in power we can do some data analysis so uh before i mean some some nice stuff that deep down was doing was also presented so they basically i mean they they have a lot of data about dows and i think what should happen is basically to take this data and analyze really what is the causes of these low participation rates and whether we can really improve it so if we have data from the dow or snapshot or whatever then we are able to do some kind of

[Music] comparison between different algorithms right because we have some data and some articles some mathematical modeling to to to help it so and then after this i think that what we should do is basically to chart like all of the solution approaches that we can find and unify them in a way so i try to do this i think that basically what are the general solutions that should help for this uh centralized governance problem so there is a so in the middle you can see this delegation based methods like liquid democracy that company is using in crypto economic method i think that the best example is a logarithmic consensus maybe also optimistic voting basically the idea of optimistic is that everything passes unless somebody

somebody objects and then there is some nice economical game of who objects and how and how she or he objects although holographic consensus the general idea is to use some kind of a prediction market to to help the community understand which proposal should pass or not in a very very vague way and and something else which is at least as far as i know is not really used in the house is sampling based method this is used in the i think in the us and uk for selecting jurors basically i mean you have the whole population and you sample a really small um sample from the population and you make the decision based on this sample so if you can do this sampling in the controlled way then maybe we can really allow for

having more decisions basically what i didn't put it is machinery which also i think could help but but i'm not sure yet okay so after we have this basically what i really think that we should strive for is something like like a big complicated solution which is parameterized so you can configure it to specific dials so then imagine that you have this weird machine that is composed of all of these ingredients that i tried to sketch before and it has some parameters so you can tune it or so you i mean like a dao can tune it um on the run whenever they go and then what i think should happen so a

dao could start with some something like this which combined this sampling based metal method is some kind of crypto economic ideas for helping this uh great so for helping the decision making and also some liquid democracy kind of solution all together and then as the community evolves these parameters should change and basically this general solution should adapt to the different size and the different type of the dow so this is somehow the the overall sketch that i that i propose okay so this is basically what i wanted to say i think that it's important to scale up basically the governance um system or the governance algorithms

and protocols that daos are using but this should happen properly when what i mean properly is like in a principled way and this will allow to scale up down to have like really massive collaborations happening in dallas thanks [Applause] if there are questions i will be happy if not then it's also fine yes was it slow enough relatively long i have no problem with uh being a native speaker or understanding an israeli accent so you'd have to ask somebody else okay um so my question is about multiple

proposals right now you're just talking about as far as i can tell yes no like there's a proposal yes no but how do you anticipate dowse could work better towards hey we have a problem here four different proposals um which seems like a better method of governance yeah i see this is a good question so basically as far as i know most of the dials today use these binary decisions basically there is the status quo and some uh but but i think that there are some other things so if you go in snapshot you see some other stuff and i mean i mean there are some algorithms for for making good decisions for let's say more complicated types of decisions so you can have like a set of proposals and

you want to rank them i mean you you can let people rank them you can let people approve substance of them and then you can run some algorithms what i'm still i'm not sure is how to somehow plug it into like the governance system of the dao because i mean a simple way would will be that okay everything we do in this way but this is not flexible enough and so i mean i don't have a really good answer to how to to really solve it one maybe slightly abstract ways basically you can view any decision as so i mean if you have like a space of all of the proposals then you can try to develop like a unified very generic kind of

um proposal system which is not only like binary but could encompass everything in like the same ui or same idea but it's still it's not clear how to really decide each decision what is the like the biotype that it would apply to i cse i mean uh i view it basically as two orthogonal components i mean you can scale up a binary proposal system and you can scale up ranking systems and you can scale up unified system but i think that i mean if you want to then integrate those two advances you will have to do some work but i don't think that it will be too much work so

so this is why i'm fairly safe i mean i am still i feel safe in scaling binary purposes okay thanks [Applause] you