{
  "metadata": {
    "transaction_key": null,
    "request_id": "metagov:revel-metagov-20230222",
    "sha256": null,
    "created": "2025-10-27T23:38:02.313423+00:00",
    "duration": null,
    "channels": 1,
    "models": [
      "metagov-manual"
    ],
    "model_info": {
      "metagov-manual": {
        "name": "metagov-manual",
        "version": "2025-10-01",
        "arch": "manual"
      }
    },
    "warnings": null,
    "summary_info": null
  },
  "results": {
    "channels": [],
    "utterances": [
      {
        "speaker": "Speaker 1",
        "start": 0.0,
        "end": 0.0,
        "transcript": "I'm really thrilled to introduce Manon Revel to speak with us today. I met Manon at the the plurality event that Glen Weil and Friends hosted in January in Berkeley and was just really thrilled to hear about what she's doing and just couldn't wait to to get her in front of the the Med Gov community. Manon Ravel is a PhD candidate in social and engineering systems and statistics at the MIT Institute for Data Systems and Society and is a democracy doctoral fellow at the Harvard Kennedy School Ash Center for Democratic Governance and Innovation. She works at the intersection of computational social choice and political theory to research decentralized government governance and the alignment of decision making processes with democratic principles. This is, really, really deep work that engages critically with some of the tech tools and techniques around online governance and beyond that a lot of us have been interested in. So I think this is a, you know, a great opportunity to build on build on stuff that we've been exploring in the past with much more depth and and criticality than than, you know, I've seen in a lot of context before. So, Manan, welcome. Take it away."
      },
      {
        "speaker": "Speaker 2",
        "start": 15.0,
        "end": 15.0,
        "transcript": "Thank you very much, Nathan, for the introduction. It's also really great to meet you all. I am definitely a new member in this community, and I'm really interested in trying to connect more in general with the community as as the next step for me of research unfolds. So I would also love to, of course, get your comments on anything I'm going to say, but connect and collaborate in the future. I'm going to share a few slides, which I think basically is a condensate of everything I feel I understand or don't understand about liquid democracy, which has been a topic I've been really interested in and focusing a lot of my time and work on on this concept. So I'm going to present a bunch of different pieces of work that I've done in collaboration with many amazing, people. So I just want to start acknowledging them. My co author, Danielle Halpern, with which I've worked on some of the pieces I'm going to present. We've also worked under the supervision of different professors, Ariel Procaccia, Elhan Woselle, Adam Berensky, Joe Halpern, Ali Jabavai, Eric Birum, and Argan Fung. And more generally, the ASH Center, the Democracy Center at the Harvard Kennedy School. So, you might have heard of Liquid Democracy. If not, let me, tell you a bit about how it works. So the first important thing about Liquid Democracy is that it is a voting mechanism or selection mechanism for people. So you can really think of it as a selection rule for governance. You want to select people that are going to make a decisions either in a direct setup where you just have to actually make some type of referendum, but you allow people to delegate their votes. Or you might want to compose an assembly of people that are going to represent your organization for different issues, and you can also use use Liquid Democracy in that sense. The kind of meta question I'm really interested in is the the extent to which liquid democracy would fit for some type of digital constitution, which I think is some of the topics MedecoV is interested in. And, and what I mean by this is I don't think there is one ultimate form of selection that would work for governance, but I was really interested in understanding the different underpinning of liquid democracy. A bit more precisely, let's say that you have some type of network, so you have people that know each other. And at the time of selecting the assembly or making a decision, everyone can either cast a vote or delegate their vote to someone they entrust. And so in red, now you have the delegation patterns, and you see that you have three people that self selected and four people that delegated either directly or transitively their votes to someone that represent them. You count the weight of all of the representatives in the decision by just counting the number of people that trusted them. So one person interested the green agent, two trusted the orange, themself, and this person, and four the blue because of the transitive pattern here. And so what I want to cover here are some thoughts on how Liquid Democracy perform on an epistemic and procedural dimension. And what I mean by this is there are a few things you care about when you think about selection rules for governance. And I guess two very traditional dimensions people study is what is the substantive outcome that this rule delivers, and what can you say? So this is the epistemic idea. And then there is procedurally speaking, what is there about this method that intrinsically or instrumentally is interesting? The procedural part is something that's much more embedded into political theory, which I'm going to go about in the second part. And the epistemic framework is actually something that has been heavily studied mathematically. And so you you can do this in different ways, but I'm going to show you how we can rezone about this making good decisions with the selection rule in a very specific context that you might have heard of before where so this is a modeling abstraction. It's a mathematical abstraction of reality. Doesn't mean that it's capturing exactly what's happening in people's mind, but it's a way to take away some aggregate trends on on how people behave. And precisely so this the objective at this point is to try to understand how liquid democracy performs in in term of substantive outcome it can deliver. So if you assume that you have n people that were on a binary issue, and you assume that one of the outcome is the ground truth, which you can typically do of course, no one knows the ground truth, though you have just the whole point of the game here is to find a correct outcome or a beneficial outcome for your community if it's something that you can define. And you can assume that every voter finds the correct outcome with some probability PI. So let's say that this is your society, you have three people, and their relative expertise here is 80%, 80%, and 90%. So they send you something about the information they have about the decision at stake. If you look at the probability that a majority of them is correct, which happen when at least two of them are correct. So either three, two, two, and two. You look at all of the possible possibilities. But you see that the probability that the majority is correct is higher than the probability that a few people are correct sorry, than than the top expert is correct. So the probability that the majority is correct is 92%. The probability that the top expert is correct is 90%. And there is some type of interesting, intriguing findings when you look at these three examples in that, in fact when you aggregate signals that are not necessarily as good as that expert one, you can actually beat what the expert performance would have been on finding correct outcomes. And this in fact is because of this underlying theory that was developed in the eighteenth century by the Marquis de Colerseil, who said that if you have a bunch of independent votes and the average expertise of your society is above a half, then the probability that the majority is correct is going to go towards one as the population increases. And this result has kind of coined and triggered a bunch of different lines of research around these ideas of collective intelligence and wisdom of crowds. This idea that in fact you could beat a fixed number of experts by aggregating independent votes as long as the average expertise is above a half. Note that this is not an ode to with Democrats. If you look at the complimentary scenario where the average expertise is below a half, then the probability that the majority is great goes to zero and type of a dystopian confusion of the multitude scenario. The reason why I show you this here is to kind of build what the impression we can have on liquid democracy. Because the idea of liquid democracy is that, you know, if you were to try to get substantive outcome, you would assume that people that are less confident about what they know or just know less would tend to delegate their vote upwards. And so you could increase the probability of correctness by having delegation going from people that you're less to people than you more if if you had some reason to believe that this is how people would behave. But even in that scenario where people would delegate their votes to, you know, people better informed, you could still end up in this, suboptimal situation because if everyone delegates to the same person and you just have one person representing the entire group, you bring this entire idea of having this collective intelligence because basically you collapse the entire potential people that could make a decision on a few people. So you have kind of in if you want to think epistemically or substantively about liquid democracy, you have a trade off between how many people vote and what is their expertise. And this interplay of do you have enough competent experts and you really need to have enough of them and that they are, you know, competent enough, but how how can you actually characterize this interplay? And in fact, this is something you can study completely mathematically from a theoretical perspective and an empirical perspective. And I want to give you a snapshot of some of the things we understood there. So then I'm going to kind of merge the empirical and theoretical expertise ideas here, where you can look into our experiments. So we ran a bunch of experiments where we asked people to make decisions and either delegate or or vote themselves. And what we did is that after the delegation phase we asked everyone to vote on all of the issues they had delegated. So we could reconstruct the direct way, the direct democracy way where everyone vote, and the liquid democracy way where people actually delegate. And so we could compare these two worlds. And some other thing we saw, if you look at the performance of liquid in pink versus direct in blue across different questions and across different experiments, you see that overall there is this interesting trend that the liquid vote ends up beating the direct vote. And the cases that are maybe most interesting are the one where liquid democracy is above a half and direct democracy is below a half. We have here these five data points where direct democracy is wrong. In two of them, liquid democracy is even more wrong, but the outcome of these vote would just be that the the population made a wrong decision overall. But then if you look at the cases where the average for the liquid democracy is above a hat, then you're in a situation where the outcome of the vote would have been the correct outcome with liquid and not with direct democracy. You can do some aggregate statistics trying to compare the average impact of liquid versus direct and find that there is the the exact percentage of the increase is statistically significant, but it's small. So it's it's 4% in in these experiments. Now what I was telling you in the beginning showing you these different graphs is that basically you need to have some interplay between the average expertise of the group and the people that get delegations. But delegations are a, local mechanism. You know, there is not someone, a central planner that says you should delegate like this so that enough people get represented because otherwise, you probably break the whole point of self governance. So you need to really rely on these local mechanisms of people deciding to delegate. And there is a very important information that we need to understand in terms of how people actually delegate. And because I mean this epistemic realm, the thing that's really important is trying to understand how the delegation behaviors relate to the expertise of our agents here. And so this is something that we studied theoretically. And the way we did this is we we looked into two ways of parameterizing how people can behave. The first one is what are their probability of delegating? So you you have yourself arrow when you vote yourself, which happens with probability q. And you have with probability one minus q, you delegate. So this is, you know, one way to capture some type of trends. And then when you delegate, you're gonna choose someone with a specific weight. And we assume that this weight depends on your expertise and their expertise. Though this is just a way to again, it's a mathematical mathematical abstraction of how people delegate, and we're trying to understand how the delegation patterns relate to expertise. And, very important thing also, we're not saying that there is any, you know, micro cognitive mechanisms through which people actually make these decisions. We assume that it's some kind of, you know, abstract mathematical trends. There must be many other things that go into the actual decisions of the people, but if we see some significant trends, it can actually give us really interesting insights on the performance of liquid democracy. And what we prove theoretically is that mechanisms should these delegation mechanisms should belong to certain classes for liquid democracy to work well, quote unquote. And in fact, we find in practice that the the mechanisms we observe match, one of the classes that we find theoretically. And I'm just gonna jump to tell you what is this class. It's overall pretty intuitive. So here you're looking at a graph where you see the the expertise of agents and as a function of sorry. You see the probability of delegating as a function of people's expertise. In that case, the person delegated. And you see that the least expert people are, the more likely they are to delegate. So you have this interesting decreasing trend where basically it it does what you would, you know, expect. I don't know. Or otherwise, you actually depends on what type of models you have for, you know, people. But it's interesting that you have you see this type of trend that on aggregate people that no more are are less likely to delegate. On the other hand, if you look at the weight that agent puts on others, you see that it increases with their weight, with their expertise. So you're more likely to choose someone who's higher expert than than than not. And a few interesting things within these results is that first of all, as I said, it matched our theoretical understanding of when should liquid democracy work. But also we ask questions, we work with groups such as classes and students and companies, and we ask questions about finding constellations in the sky or predicting a soccer game result or a basketball game, which I think that were not really related to the expertise and we didn't exactly know what they knew about each other. There was a real reason to know that, to think that they knew a lot about whatever the other knew about star constellations in the sky but we still saw these trends that that seem to suggest that there is a relation between delegation patterns and expertise, which might just be what we need to think that liquid democracy can in fact deliver substantial outcome. A very important note here is that the results that we have are for very well connected graphs and non polarizing issues. So we really think of this as working in a context where you have a group of people that know each other and that work towards the same goal, but I don't think we can say much about how these results would carry into a democracy. Another aspect of this work is that we find overall that there is not too much consolidation of power. Other experiments, in particular in political setups, found that you had a lot of consolidation of power. And so our understanding is not that liquid democracy would always lead to one or the other, but you have contexts in which one is more likely than the other. And what we're really trying to do is to understand in which context liquid democracy seems to work, in which context it doesn't seem to work, to try to reason about where we can actually install it with some type of guarantees that, you know, over time, it should more often than not work well versus not. So a last few thoughts about why liquid democracy versus why not liquid democracy. Not only thinking about this very mathematical approach of when things are wrong or exploits this apparent legitimacy paradox. What I mean by the legitimacy paradox is the idea that when you have decisions that are made by a group of people on behalf of the larger groups you always have this really tricky legitimacy question that relates to the plurality type of ideas that Nathan mentioned we were part of a group that was thinking about all of this, where if you have a bunch of people that disagree with the outcome, why should they still obey to the the consequence of their decision that was made by a smaller group representing the bigger population? It's something that often occurs when you have some some type of democratic government with a plurality of opinions. And something that's always or sometimes is presented as paradoxical in this legitimacy question is whether you can have the same time decision based on competent decision makers and democratic processes. And Kurose first hinted at the idea that this was a false distinction and that you could have both, you know, majority voting is in a sense democratic, but also in some cases a guarantee that you can reach good outcomes. So you have this idea that and which is something that we need liquid democracy further exploits that you could actually have an indigenous way of finding experts in a way that respects everyone freedoms and chance of actually participating in the process as well. Something that's also interesting is how liquid democracy leverages the underlying social connection of your, of your group with the decision, which is something we never really do when we make decisions. We kind of only care about the information agents have about the outcome, but we don't care about the interpersonal relation that people have about one another. So it's an interesting extra piece of information that we get into this. In a sense also liquid democracy extends what we typically think of as inclusiveness and equality when we think about the electoral theory of representation. But note that when we think about elections, you're included in the process if you're part of the demos, if you have the right to vote. And we think of equality as one person, one vote. But this is really inclusiveness and equality at the selection selection stage. So you have an equal chance of influencing the selection process of the people, but you don't have inclusiveness or equality defined at the decisive stage. In In liquid democracy, basically, you completely abolish entrance barriers, anyone that yourself, like, will be part of the decision process. So you have a very different take of inclusiveness and equality that in a sense just strengthen what we currently understand in the theory of political representation within the electoral framework. Similarly, it's strengthened authorization, which is the the way to which you authorize the people to give consent to the people to rule on your behalf. The way it typically work in The US when you have a multi voter voting with one district, single district member winner is that basically you're going to authorize the winner if a majority of the people voted for them. And this is a, you know, rather weak sense of consent because you might have not voted for them but still you obey to the outcome because of this idea that the majority endorsed them. In liquid democracy, you basically have a chain that leads you to the people the person that represents you. And so you have a much more direct understanding of authorization through through the transitive patterns of liquid democracy where you basically always have someone that represents you and even your weight is being carried because your weight count in the India decision. And it's probably better used offline than online than offline because of the the because of how hard it would be to count if you had to actually track the transitive chains if you had paper ballots. So these might be set setups where liquid democracy would would make sense. Where liquid democracy, in my understanding, and this is really up to debate, it's something I've been thinking a lot about, but, you know, still under construction is why not liquid democracy? And before I go into any of these considerations, a few things you can already see with liquid democracy is that you have a risk of potential cycles. If I delegate to Nathan and then if it delegates to me, we are on a cycle. And let's say that a lot of people delegated to Nathan, so now we broke all of these people within an an indecisive outcome where we don't really know what to do with this. And you have ways to resolve the cycles, but it asks a bunch of more questions. The other thing is that you can have a lot of delegations going to someone that does not participate, which also create issues in terms of how do you account for these people, how do you tell them that they are not being represented. And then you have different versions of liquid democracy. One way you can have instant recall. So you could constantly say, I want this person to represent you or not, which gives a very strong account of accountability, which has very good properties, but also can create instability. So on the implementation side of liquid democracy, you can have issues. And the last one, of course, is that you can have a concentration of power. And it's not and even, you know, when we prove that it's unlikely in the scenarios we're studying, it doesn't mean that it cannot happen. And then what is the cost that you incur even even if it's something that happens every two hundred years? So all of these practical questions are really related to, you know, when you read the Federalist Papers and when you see how constitution was was thought, the common understanding is that there is no perfect rule and that you need to have some type of gatekeepers that are here to safeguard when things go wrong. And these discussions are super important for liquid democracy as well because it has a a list of shortcomings as any other system. On a more philosophical perspective, liquid democracy does not guarantee a quality of influence. Some people might have much more weight. And so I was saying it could lead to someone, you know, eventually even having if you have more than half of the weight, you basically are a dictator within the assembly. And so this can create, you know, issues. And so some would want to argue that it's okay because people actually decided to like to dedicate their vote to these people. But now if you think of an environment where you have captured by special interests and you have manipulation being done by people that have concentration of wealth or media, information or stuff like this, then you might be really worried about how this inequality of influence comes about. And so really following up on that on that thread, we are right now in my understanding a very interesting moment where we have been operating with the electoral framework for a lot of the governance representation theory, at least at the political level. And this electoral framework is being questioned. It's it's not really clear to to know whether there is an intrinsic problem with elections being really prone to capture by special interest, whether this is a contingent observation that we are in a situation where this happens, but regardless, we there is a lot of convergence toward the idea that this is becoming a a problem that, that we have in certain scenarios. So if you're worried about capture by special interest, currently the theories of representation with elections don't provide you necessarily good answers. And so they are emerging alternative theories of representation. For instance, relying on drawing people at random, drawing citizen assemblies with a lot of amazing work done by both Ariel Pukesh's group, philosophers like Elena Damour, Alex Guerrero. And there you have other ideas of inclusiveness, equality, and diversity that are, you know, another theory. And because liquid democracy really fit within the more intellectual framework, it's not something that actually, liquid democracy would address if we were thinking that we had to go that route. It may require really large assholes as a lot of people's self select. So you might also want to think about whether you need deliberation, how to accommodate deliberation with large groups. And okay. So the last few things I wanted to say are that our understandings is that delegation behavior seem to relate the expertise of the agents with well, there is a relation between the delegation behavior and the expertise, and that the trends we have seen seems to seem to indicate that in the well connected non political context we study, there is a really interesting case to be made about how epistemically liquid democracy can perform. And I really highlight that this is for a very specific context. And, we know that conservation of power, for instance, can occur. And we're really, you know, in this kind of cornerstone mindset, try to discriminate between cases where things go well and things don't go well just for us to better understand the systems and better understand systems and better understand how to use them. In terms of of limitations, the again, you know, deploying a system really depends on the context in which you are. And it may be really different when you're trying to do a change at the level of a country versus in a company, whether it's in an online community. And so our understanding is that equity democracy would probably work better technically, mechanically in online communities. That, you need to want to think about whether your online community has, you know, a very polarized community, with special interest, not special interest. All of these are things that really come into play when you want to think about using Liquid Democracy. And that we have kind of this decision to understanding of when Liquid Democracy from a personal perspective would make more or less sense. That's it. I hope I was on overtime. If so, I apologize, but I'm opening the floor for questions."
      },
      {
        "speaker": "Speaker 1",
        "start": 30.0,
        "end": 30.0,
        "transcript": "That's perfect. Thank you so much. What a great start to a discussion. Ofer, you had a comment early on, and then Steve will turn to you next. Ofer, do you wanna get us started? Maybe not. Maybe we'll start with Steve then."
      },
      {
        "speaker": "Speaker 3",
        "start": 45.0,
        "end": 45.0,
        "transcript": "Hey. I'm sorry. Can you hear me?"
      },
      {
        "speaker": "Speaker 2",
        "start": 60.0,
        "end": 60.0,
        "transcript": "Yes."
      },
      {
        "speaker": "Speaker 1",
        "start": 75.0,
        "end": 75.0,
        "transcript": "Yeah. Go ahead,"
      },
      {
        "speaker": "Speaker 3",
        "start": 90.0,
        "end": 90.0,
        "transcript": "Oleg. Yeah. Sorry. I was just driving. So I put a little new paper in the comments from a group at Columbia. There's some economists that experimented with with democracy, and it actually underperformed because people delegated too much. I wonder if you were aware of this paper and if there are ways around it, for example, by incentivizing people maybe to delegate less."
      },
      {
        "speaker": "Speaker 2",
        "start": 105.0,
        "end": 105.0,
        "transcript": "Yeah. So I we are aware of this paper, and we got in touch with them, and we are trying to understand exactly, you know, what is the driver of the different scenarios we observe. What what we know is that you will have constitution of power in some scenario. And we're really trying to understand what conditions do you have constitution of power and what conditions you don't have constitution of power. And so this kind of theory versus empirical work we've been doing has been identifying the scope where constitution of power does not occur, which, you know, leads to these good results. We have intuitions for why you have constitution of power in some cases because our theory just also suggests some you can think of mechanisms within the scope of our theory that would lead to constitution of power. I don't exactly know what is the mechanism at stake within within their paper. So I I'm not I'm not sure about exactly why it happens. But really that's something that's super important that I always try to emphasize is that we don't think of or I'm gonna talk about myself. I don't think of any solution rules as a silver bullet for anything. And I think a lot of the privilege we have within this academic environment is trying to, you know, have the time to think about the context in which it work and the context in which it don't work. And we, you know, always acknowledge that both are possible, and we're trying to understand as much as possible what's the rationale for one world versus the other. Does that make sense?"
      },
      {
        "speaker": "Speaker 1",
        "start": 120.0,
        "end": 120.0,
        "transcript": "Thank you. Steve, do you wanna go ahead? Or yeah."
      },
      {
        "speaker": "Speaker 4",
        "start": 135.0,
        "end": 135.0,
        "transcript": "Yeah. I I thought your critiques were pretty good for their for liquid democracy. My main critique of it is that I don't think that using it for governmental democracy is the way to go. Like you said, small groups is is a, I think, initial good scenario to use them in. But, also, I think shareholders, you shareholders essentially replacing the board of directors. So all that functionality of the board of directors should be taken over by shareholders. And, well, I actually I believe that I wrote the first modern paper on liquid democracy. So in the intervening twenty years, I've come up with a lot of augments to solve a lot of the problems you identified too. So I'm and I'm glad you guys are studying this. Finally, I was I was afraid that Dunning Kruger, you familiar with that whole business, where basically people who have the least expertise, most overestimate their own expertise? I was afraid that that might undermine the very premise of liquid democracy, but you guys have proved it out, and I am so happy."
      },
      {
        "speaker": "Speaker 2",
        "start": 150.0,
        "end": 150.0,
        "transcript": "Always very, you know, careful with any any two broad conclusion that we could read from this. We definitely see this happening as well. We definitely see that you have we have metrics of confidence and we see that you have some, you know, overconfidence by people who have expertise. The thing that's interesting is whether all of these trends will happen in reality and we're trying to understand what are the kind of, like, you know, probabilistic macro trend. And as long as you have enough people with lower expertise that think about delegating, you you end up within the scenarios we identify. But again, you know, we are super aware and cautious about the fact that these work in very particular scenarios, and we we've only tested this with trial groups and in very controlled environments. We think it's a really interesting understanding, but but other things could come out."
      },
      {
        "speaker": "Speaker 4",
        "start": 165.0,
        "end": 165.0,
        "transcript": "Just just to break back in here, I think one thing you have to consider as well is that, at least in my original conception of it, that there would be sort of expertise built up over time in groups. So it was only when there's demonstrated expertise, not sort of like everyone just kind of fishing around to try and figure out who's the expert in just one vote. So I think that the potential as it iterates is really where you're gonna and there's all sorts of things with float whatever. I'll drop my original paper, which I think is I think as far as I discovered, it's still the most complete, sort of, elaboration of the idea. So"
      },
      {
        "speaker": "Speaker 2",
        "start": 180.0,
        "end": 180.0,
        "transcript": "Yeah. I I'd love to read it. Thank you for sharing it."
      },
      {
        "speaker": "Speaker 4",
        "start": 195.0,
        "end": 195.0,
        "transcript": "Sure. I mean, hopefully, it's not nonsense. I haven't read it in ten years myself. Okay. Alright."
      },
      {
        "speaker": "Speaker 2",
        "start": 210.0,
        "end": 210.0,
        "transcript": "Just a a few more comments on this. Other things I'm I'm worried about is is, you know, we also know that you have a lot of biases one would want to avoid in terms of how people self select and how people you know, you could see also the other way around of people with high expertise for some biased reasons seeing thinking that they don't have. And we know that they are, you know, reasons to think that there are ways through which this pattern exists that depends on features that you, you know, you won't want to not, encourage within a liquid democracy system. So all of these things, I think, are thing that we need to, you know, think about in general. And and, yeah. That's again, I think it's yeah. The other thing I would say is that we I'm sorry that I maybe put too much of the emphasis in the end within the political context. And I think that for me, the representation literature is really interesting to think about governance either offline and online, and it's why I kind of build on it. But I also take its insight to think about online governance. And so this idea of you were talking about stakeholders. I know of hedge funds that are working with some version of liquid democracy or proxy democracy where you're trying to actually delegate your, you know, your outcome to people that aren't motivated to have the time to work at the moment. And so for me this three d matrix of context versus implementation rule is super important to explore and is not fully captured by of course a mathematical"
      },
      {
        "speaker": "Speaker 1",
        "start": 225.0,
        "end": 225.0,
        "transcript": "theory. Thank you. I also really wanna just underscore that concept that Steve raised about the development of expertise and the comparison between this kind of logic and, for instance, a citizen assembly logic or a jury where cultivation of expertise is explicit and and intentional rather than kind of epiphenomenal. And yeah. So next, we have Yanis. Do you wanna go ahead? And then Anna."
      },
      {
        "speaker": "Speaker 3",
        "start": 240.0,
        "end": 240.0,
        "transcript": "Yes. Thank you. Thank you for the for the discussion. And, Steve, yeah, I'm looking forward to to reading your paper. So I'm looking forward that you can put it in the in the chat. So so my question, it lies on the how does liquid democracy address the asymmetry of information that naturally happens in in social systems and which is the main, I think, culprits for the imbalance of representation, power, and influence. And then the corollary to that would be a question of scale. Echo a little bit what Steve also asked, whether liquid democracy can only practically function in small scale systems, and how would you envision, you know, organically growing that scale bigger and bigger to kind of, like, reach the the the entire, let's say, quality? Thank you."
      },
      {
        "speaker": "Speaker 2",
        "start": 255.0,
        "end": 255.0,
        "transcript": "Thank you for for the for the questions. To your first point, I don't think liquid democracy can solve the asymmetry of information. And so I guess, like, I have a more neutral versus skeptical take on on that. The all the governance problems are multilayer case. You have how people are, you know, enter the policy, how they are informed. You have, I don't know, all of these different layers. And then the ones you have close to the top is how do you select the people that make decisions. But the selection rule can have a feedback impact on everything that comes before. But in fact, they are not necessarily meant to influence the information stage. So the more neutral stake is that, you know, it will take whatever structure of information it has and then deal with it. Maybe a more skeptical tick could be that you would have maybe a tendency for those less informed to just delegate and not actually try to fill that gap, which you can make arguments for why this is not good. I don't exactly know how this would play yet. There is for sure there's gonna be some type of feedback loop, but it's not necessarily the first thing. And this problem would be, you know, pretty constant across electoral rules, blah blah blah. So I think it's kind of the first element of answers would be would be that it's a selection rule. It does not directly address the asymmetry of information problem. Now I was talking about the captured by special interest as as a problem that philosophers have been, for instance, looking into. The asymmetry of information is another problem. Education about political issues is another problem. And so the way I really try to think about about these selection rules is overall from a substantive perspective, how does it change, you know, our lives in the end? And of course there are so many parameters, but I think trying to to think about what you're trying to solve is in this super important. And if you're trying to solve capture of interest, it's not the same thing as if you're trying to to solve asymmetry of information. And maybe the intervention for these different problems is not how you select the representatives, but you have all the things that come into governance that might be better suited to fit with asymmetry of information. On the small committee sides, all of our experiments are in fact on groups that are small. The the largest experiment we have has a 127 people, I think, and the smallest is 12 people. And most of them are have less than probably 50 or 70 people. We only have one, you know, large scale for our purposes. So we do think that it has power to work in these scenarios, especially when you have a bunch of decisions that are really in very different topics, and you don't want to have to mobilize everyone, the same group of people to have to think about all of this. But when you want to kind of divide labor and have a way of dividing labor that leverages the information people have about one another. And this could also, in a sense, be, you know, even maybe I don't know what better it works, but it can definitely can work in smaller setups. I'm working with companies now that are trying to implement this internally, and these are for small scale implementations."
      },
      {
        "speaker": "Speaker 3",
        "start": 270.0,
        "end": 270.0,
        "transcript": "Did did you notice a small color color, Larry? Sorry. Did you notice difference in the implementation between groups that have face to face, like the group meets face to face, and that's kind of like a community building versus the one that only see each other through digital platforms or from far. I would be very curious if that has has an impact on the effectiveness with which the implementation."
      },
      {
        "speaker": "Speaker 2",
        "start": 285.0,
        "end": 285.0,
        "transcript": "Yeah. That's a really good In fact, in the the places where I'm working with, one of them is full anonymous online environment, and the others are in person people that know each other. So, indeed, it will be super interesting. And I don't know exactly how it's gonna play out, but it will be super interesting to see how it changes in these different places."
      },
      {
        "speaker": "Speaker 3",
        "start": 300.0,
        "end": 300.0,
        "transcript": "Thank you."
      },
      {
        "speaker": "Speaker 2",
        "start": 315.0,
        "end": 315.0,
        "transcript": "Thanks."
      },
      {
        "speaker": "Speaker 1",
        "start": 330.0,
        "end": 330.0,
        "transcript": "Alright. Anna, go ahead."
      },
      {
        "speaker": "Speaker 5",
        "start": 345.0,
        "end": 345.0,
        "transcript": "Thanks. This is really interesting. My background is much more focused on citizen panels and citizen juries, so very different types of approaches. So my biggest question is around gender. And I don't know if you did mention this, and if that's controlled for or how that's thought about just in terms of, thinking about expertise. There just there are gender norms in the way people perceive their own expertise. And just curious kind of in other identities as well. I just gender is the one that I'm most familiar with."
      },
      {
        "speaker": "Speaker 2",
        "start": 360.0,
        "end": 360.0,
        "transcript": "Yeah. So it's it's exactly what I was hinting hinting at answering Steve's question, but we don't account for this at all in in our model. It doesn't mean that you don't care about this when you work on equity democracy, but in our scenarios, we, you know, we ask people we have a group and we ask them to delegate and we don't control for or we don't nudge them. We don't you know? So we don't try to account for that. But what I was exactly thinking about when I was telling that one of my worries thinking about implementation is, you know, what if some people in a very biased manner decide to self select to opt out and delegate their vote for, you know, systematic wrong understanding of their own competent that is actually inaccurate. I think this is indeed a very important problem."
      },
      {
        "speaker": "Speaker 5",
        "start": 375.0,
        "end": 375.0,
        "transcript": "Yeah. In your experience just with all of the the groups that you're monitoring and looking at, I mean, have you just sort of, like, qualitatively do you have a hunch that it's impacting? I don't know if you wanna answer that on the record, but I just I've just for my background, I've been on congress for the last three years, and, absolutely, there are certain groups of people who think they're more experts than others. So just that's my qualitative experience."
      },
      {
        "speaker": "Speaker 2",
        "start": 390.0,
        "end": 390.0,
        "transcript": "No. That that's super interesting. In fact, I think we have some you know, at the end of surveys, we asked demographic information. I did not look into this, but I think it would be super interesting indeed to and I so I don't know the answer because I so I can say this on record because I don't know whether it would be validated or not. But I guess I wouldn't be surprised if we were to find biases as the ones you were alluding to."
      },
      {
        "speaker": "Speaker 1",
        "start": 405.0,
        "end": 405.0,
        "transcript": "Amel, would you like to go next?"
      },
      {
        "speaker": "Speaker 6",
        "start": 420.0,
        "end": 420.0,
        "transcript": "Yeah. Hi. Thank you. That was that was really just, like, inspiring on where where this stuff is at. I I I've been kind of grinding on some of this just from the perspective of a, an enjoyer of these things, and trying to solve, like, real problems out of DAO too. But, let me just kinda ask this question, which is, kinda what is the state of of, the technology in terms of being able to forward one's, vote? I found what I would call sort of an extreme limitation in the platform that we were using at Index Scoop, which is is Snapshot. And so you could only forward kinda one hop. Right? Meaning, if I wanted to forward my vote, I could forward it to Nathan. But Nathan if if Nathan didn't show up to vote that day and Nathan had forwarded his vote on to someone else, well, too bad. That voter, you know, it didn't it didn't pass through enough to register that. I came up with and I and I the frustration around this is I came up with an idea to, like, what I thought would solve a big problem, which is, forced constrained delegation and liquid democracy bolted together could register a 100% participation rate on every vote either through the the, you know, the act of delegation or even laziness. Right? You've delegated previously and you left that delegation open, but you could then just measure who who voted after every vote right and get to the place of how much power was used. You know expressed through you know the that sort of aggregation, you can almost even find the loops and things like that, But I guess getting to a place of, have you found limitations in the tech yourself? And do you where do you see it kinda going? Because I I see that being, at least on the forefront of, like, the least possible amount of information I need to install in a human, to get them to do a very tiny action or very tiny ongoing actions regardless of context, and then have something that where I can glean repeatable, you know, data out of the system. This just seems like the some of the greatest promise in in the space in terms of terms of really getting down to, how do I put it, p to p repeatability and and things like that of of governance systems like the nanotech of decisions and and as the way I think of it as a DAO worker. So, yeah, I would love to know kind of where you're at with that because this has been just an impressive survey of, like, something that I've been just knocking around for a little while. So thank you."
      },
      {
        "speaker": "Speaker 2",
        "start": 435.0,
        "end": 435.0,
        "transcript": "Yeah. It's it's a really great question. Thank you. And I don't know that I'm gonna have the most, satisfying answer for your purposes. But, yeah, I'll I'll I'll tell you what are my thoughts. And I guess the kind of prior information I want to give you about this is that I'm working with my own, you know, kind of cheating, working with a small scale smaller scale tools that work for research purposes, but that I would not recommend for at scale deployment. So when I'm working with companies, I'm working to build in house things with them because we probably don't want to use any other things we're using. The only one I'm aware of that's deployed is, as you said, Snapshot. I didn't know how they were exactly handling I don't know how they handle cycle. I don't know how they handle this the pipe that the chain doesn't go anywhere. I might sense something really wrong, but my under my understanding technically would be that it would not be that, you know, hard to say if your vote ended up somewhere, you get a notification saying you can't change your vote. Now something I'm more also working on these days is that this has really big implications when you think on the theoretical perspective because of if you you know, now if you wanna think about manipulation. Let's say that I receive a notification because I was in a cycle, then maybe if I'm a malicious agent that want to force you to delegate your vote to me, I can delegate my vote to you, and then I wait to see whether we get the notification saying we're in a cycle. And if we get this notification, I could check that you actually delegated your vote to me, and then I can go onwards and, you know, undo it and then do whatever I want. So the my understanding but, again, I've I've not worked at all close to people working on the software development of this type of thing, so might not have the most informed opinion here. But my understanding is that, technically, these things are things that we can completely implement, But they they come around with new sets of manipulability questions that you need to add on your context checklist when you think about how you deploy it, whether you think you're in a high manipulable scenarios and whether that is the case and what do you do. And if you're not, then maybe it's okay to put them on the features. But if you are, then how do you have some type of strategy proof, mechanisms, or incentives around this new feature? It's kind of how I think about the more technical implementation. It's what new problems added features"
      },
      {
        "speaker": "Speaker 1",
        "start": 450.0,
        "end": 450.0,
        "transcript": "open. Thank you. We got some of those up there in the videos. We've time for a few more questions, maybe a couple more depending on how much they elicit. I have one I'd love to share. A group of us in the MediGov community are developing a research project on attention economies, you know, use the language of in governance specifically, use the language of epistemic burden as, you know, as one vector that you're looking at and thinking about. And and, you know, a question for us here is just how you know, what what kinds of patterns are emerging and what kinds of best practices might we see around the organization of attention in in governance, especially when one is engaged in multiple governance services. You know, say you're a member of, you know, a couple dozen DAOs or or whatever they are, and you have to navigate your participation in them simultaneously. Do you have any thoughts from your work on the liquid democracy across spaces where you might be holding, you know, spaces in parallel? Or do you have any kind of thoughts or questions or guideline or or encouragement for, a research project around these kinds of questions? What kinds of things we might, especially try to look for, dilemmas that you've run into in thinking about, epistemic burdens in in the context of these these practices?"
      },
      {
        "speaker": "Speaker 2",
        "start": 465.0,
        "end": 465.0,
        "transcript": "Yeah. It's also a really good point. I had never really thought about the multitasking across different scenarios. I would say that my intuition is is that the liquid democracy in itself, the theoretical definition is this three layer cake, which is transitive delegation, instant recall, and per issue delegation. And what's really baked in the third layer, the per issue delegation, is this idea that you should be able to divide labor and you should be able to say, well, in, you know, that moment, I want to be really active in that DAO and not in that one. And so that really this division of labor ideas is really inherent. And because it's so granular and you can you could actually create committees that are dedicated to very specific issues and choose one representative for issue, you could decide when you want to be active in something or not. And you could also say, you know, I want to be active in multiple DAOs, but I'm interested in different aspects of them. And then I can be part of different committees and I can also select based on, you know, what is the workload on it and in anything. So I think in that regard, liquid demand, which is really great. And it's it's, you know, again, one of these moments where context tells you two different things. In that in a sense, this division of labor idea brings about less epistemic burden on the one hand, but also the downside of it is that, as Anna was pointing out, you might have some weird self selection biases. So it's you know, the the very sad thing I feel I'm learning, researching governance is that every single solution has positive externalities and negative externalities. And to answer your question specifically, I think that that's one of the feature in the design of liquid democracy is precisely this division of labor granularity so that you can multitask. But it comes with downsides. And, you know, again, I think I I because I don't I'm trying to be really in the real world now, but I had the bias of coming from a research perspective where I can, you know, try to think very abstractly. And my easy answer was always, okay. I need to understand whether in this scenario, we have we need more of these or we need more of that, and what is the cost for you to pay against this? And so but in a sense, all of this benchmarking is kind of clear in my head in terms of what you gain and what you lose. But now going on making a decision in a very specific context on, you know, what is the risk you're gonna take and what is the cost is something I found much harder. I also understand that it's how things actually change, and I'm, you know, getting to this, but definitely need some help. So I would love your helps on this more impactful kind of questions."
      },
      {
        "speaker": "Speaker 1",
        "start": 480.0,
        "end": 480.0,
        "transcript": "Just quickly, I I'd love to ask a follow-up. You you mentioned earlier that you are involved in implementing some liquid democracy processes and organizations, smaller scales. Can you just reflect a little bit? You've you've mentioned a few times about that that privilege of research. Can you reflect a little bit on the process of translating this research into practice and what kinds of new challenges have emerged and and and how you're how you're thinking about that that kind of transfer?"
      },
      {
        "speaker": "Speaker 2",
        "start": 495.0,
        "end": 495.0,
        "transcript": "I think, overall, I am feeling terrified with these conversations because I'm really you know, I think even when the status goes bad, it has good things and we're used to it and there is inertia. And I'm always terrified that bringing anything new would even if it brings something better, it would also bring something worse. So at the personal level, I think it's been my overall kind of approach to all of these things. And and then when I try to think about how we can actually make design and if I try to get guarantees on this design, then I still realize that, as I said, you know, any decision you make come at a cost. And and we've not deployed anything. It's still in a rather early stage in these discussions. So I'm feeling that the hard part's gonna come when things is live and I will, you know, see the mistakes we made in the design or or not and what we can prevent for and what we can't prevent for. So I guess what I'm learning is that the research is really interesting, and I think it frames the the brain in a way that's very self critical of the time, which I think is really important in scenarios, and I'm really, you know, grateful for it. And there is social choice in particular, and the persons I work I work with at Harvard in this topic have are doing amazing job that I I found so useful for me to try to think about governance at large. I I realized there's a gap when you now you're going to practice, and I found the gap terrifying, I guess, is my honest answer to your question."
      },
      {
        "speaker": "Speaker 1",
        "start": 510.0,
        "end": 510.0,
        "transcript": "Thank you. That sounds like a great place to end, sheer terror. Is is there a a good place that people can find out more about your work and contact you?"
      },
      {
        "speaker": "Speaker 2",
        "start": 525.0,
        "end": 525.0,
        "transcript": "Yeah. I'll put it in the chat. And sorry. I don't want to end on the word terrifying. I want to end on the word exciting. So as much as I find it terrifying, I also find it extremely exciting because I think we are at a time where there are so many transformations and there are so much things to do, and we need to do something. And, you know, people would do something. And I I'm thrilled to try to be engaged in this as I kinda making these changes. And it's the flip side of the terrifying thing. It's how exciting it is. And it's not like, we need you know, I'm I'm so happy to also join this group and and learn from all of your initiatives. It's a really exciting time, and I guess it's good to be terrified because then you're double guessing yourself all the time. And maybe you're you know, you can do more on the cautious side, but exciting, though, would be the word of the of the and I'm putting my email in the chat."
      },
      {
        "speaker": "Speaker 1",
        "start": 540.0,
        "end": 540.0,
        "transcript": "Great. Thank you. I'm ravel@MIT.eduforthefor the video. So at the end of the we're at the end of the session. Everyone, please prepare to unmute and show Manon our our appreciation for all that she shared with us today with applause. Three, two, one, go."
      },
      {
        "speaker": "Speaker 3",
        "start": 555.0,
        "end": 555.0,
        "transcript": "Great. Great. Great."
      },
      {
        "speaker": "Speaker 1",
        "start": 570.0,
        "end": 570.0,
        "transcript": "I think that's the most you you got your enthusiasm. Alright. Thank you all so much. We've got much more to come. Please, follow the the schedule at medigov.org/seminar, and hope to see you next week. Take care, everyone."
      },
      {
        "speaker": "Speaker 2",
        "start": 585.0,
        "end": 585.0,
        "transcript": "Thank you so much for everything."
      }
    ],
    "summary": null
  }
}