--:-- Robert Hackett: 2024 is an important year for democracy. This year, over 60 countries have and will head to the polls, representing just about half of the global population. But governance extends past the leaders of nation-states. In fact, as the revenue of some companies dwarfed the economies of some nations, the way these entities govern themselves is both more interesting and more important than ever. Listen in to this episode, originally published on our sister podcast, Web3 with A16Z, as they explore topics ranging from content moderation to incentivizing participation to new governance experiments from DAOs to oversight boards. I hope you enjoy.
--:-- Robert Hackett: Hello and welcome to Web3 with A16Z. I'm Robert Hackett, and today we have a special episode about governance in many forms, from nation states to corporate boards to internet services and beyond. Our special guests are Noah Feldman, constitutional law professor at Harvard, who also architected the Meta Oversight Board, among many other things. He is also the author of several books. And our other special guest is Andy Hall, professor of political science at Stanford, who is an advisor to A16Z Crypto Research, and who also co-authored several papers and posts about Web3 as a laboratory for designing and testing new political systems, including new work we'll link to in the show notes. Our hallway-style conversation covers technologies and approaches to governance from constitutions to crypto, blockchains, and DAOs. As such, we also discuss content moderation and community standards, best practices for citizens' assemblies, courts versus legislatures, and much more where governance comes up. Throughout, we reference the history and evolution of democracy, from ancient Greece to the present day, as well as examples of governance from big companies like Meta to startups like Anthropic. As a reminder, none of the following should be taken as tax, business, legal, or investment advice. See a16zcrypto.com slash disclosures for more important information, including a link to a list of our investments.
--:-- Robert Hackett: So I want to start with a really broad question. Maybe it's too broad, but what is the right model for internet governance? We have all of these companies that host platforms that people participate in, that they build on top of, that they connect and communicate on. So a question is, what is the right way to govern those platforms? Right now, the most popular ones, at least, are owned by corporations, and they sort of get to call the shots there. Is that the way that it should be, idealistically, in a world that we all live and work in and play in?
--:-- Noah Feldman: May I start with a little bit on your formulation? Corporations don't exist in the ether, much as we sometimes like to fantasize either positively or negatively that they do. They're products of law, usually Delaware law, because of the weirdness of how the American federalist system works. And they're governed by a plethora of obligations and regulations that are not only in their place of incorporation, but are also state law everywhere and federal law. A lot of the duties that they have, they can't get rid of, even if they tried. So if you make and sell a product, you're liable for the bad stuff that happens to people as a result of that product, even if you pretend you're not, and even if you don't want to be. That's just to give an example of a duty you can't get away from that means that all corporations are intensely regulated all the time already, even before we get to the specific regulations that are relevant, say, to a user-generated content social media platform.
--:-- Noah Feldman: So the first thing is, you're going to be governed by state and national laws. And in a more tenuous way, you're going to be governed by those international laws that your government bothers to apply to you. And which, by the way, there are plenty of governments that are democratically formulated. Some bunch of these laws are presumably coming from the will of the people. And some of them are undemocratic and lousy, but they still affect you if you want to do business in a country. So all of these things are in place. That's the first thing I would say. So there already is a lot of governance that comes from governments, the good old fashioned kind of governance.
--:-- Noah Feldman: Then there's a further and really important question of whether there should be more governance coming from governments, and if so, of what kind. And that's a very hot topic right now. The U.S. Supreme Court this year is deciding three different sets of issues all related to what are the rights of platforms? What are the rights of users? What is the role of government with respect to social media? Until now, there's been exactly zero Supreme Court doctrine on social media companies. And now we're going to have a whole bunch of it. So this is a major transformational era. Unfortunately, we don't know the answers to any of those things yet definitively. We have sort of tea leaves from the oral arguments. So that's another really huge issue. How far should it go? How much regulation? Should the states of Florida and Texas be able to weigh in on content moderation?
--:-- Robert Hackett: So you're referring to the NetChoice case that the Supreme Court is reviewing. What's at issue there is these states would like these social media platforms not to censor or moderate certain types of speech. I mean, that's how the states would put it. The companies say that the states want to force them to carry content that they may not want to carry or want to limit the ways that they can moderate the content on what they call their platforms. So that's, like any Supreme Court case, you've got two sides and those are the two main perspectives.
--:-- Noah Feldman: And then beyond that, there's a super fascinating question. I know Andy's thought very deeply about this too, of modes of regulation that platforms can choose, first of all, for themselves and use internally. Then there's potential collaborations between different platforms on collective self-regulation, which is a form of governance. And then there's the weird kind of hybrid structure that the Facebook Oversight Board represents and then in a different way the Anthropic Long-Term Benefit Trust represents. Those are methods whereby a company creates a kind of hybrid structure where there's some independent actors whom the company agrees to be governed by for some limited set of purposes. And that's another really interesting and, to my mind, kind of innovative and cutting edge way of doing this, which like any innovative cutting edge thing has its pros and its cons and is also very much in an experimental phase.
--:-- Andrew Hall: I think that we're already getting into something very deep and interesting, which is like, where are the boundaries? Where do we think the real world governance ends and when or why would any type of organization want to go beyond it? It seems like there's at least a few reasons. There's obviously just a sense as a company or as an organization that you want to do more than government has told you to do. That's one possible motivation. It seems like there's also something about the global reach of some of these platforms. We know that global coordination around regulation is very challenging. So in the absence of that coordination, you might feel the obligation or the strategic need to do something to coordinate that yourself.
--:-- Andrew Hall: And then there's also something, this hasn't come up yet, but I think a third explanation has something to do with competition. Going back to your original question, Robert, in a world where we had perfect competition in a very classical sense between all different types of internet services, I think a lot of the governance beyond traditional regulation would be done by users voting with their feet. The challenge we get into is the scale of many internet services and the network effects put them in this kind of interesting position where they're at a very, very large scale, which makes it in some ways hard for users to vote with their feet because they like being where their friends are. But at the same time, there's still a lot of economic competition across services. So it's not obvious that these companies are really monopolies in the traditional antitrust sense. It's put them in this weird zone where there is a sense that users are unhappy being locked in to any particular platform, but at the same time, traditional antitrust tools don't really seem to apply. In that weird gray area, it might make sense for large platforms to play with additional modes of giving their users the ability to decide together how the platform's going to work in a world where they can't freely individually move between platforms for the same service.
--:-- Noah Feldman: And Andy, I don't think you would disagree with this. There's also the advertisers who, in the perfect economic picture, would just go where the people were, where the users were. So in that sense, they seem less important. But as we've seen in what you might call the X-Files, the advertisers have turned out to be major players because they don't wait to see what the users will do. They also do that, but they take preemptive steps based on their perceptions of what might happen and the reputational costs and all of those sorts of things. So they're big players, too. And they're another reason why a company, a platform might want to have its own self-regulatory mechanisms.
--:-- Noah Feldman: And here I just want to add, in this super polarized world, there are no neutral decisions that you can say, "Oh, well, everybody should leave me alone because I made a neutral decision." Whether you take content down or leave it up, you're non-neutral. Whether you amplify it or don't amplify it, you're non-neutral. Depending on how much you amplify it, you're non-neutral. Depending on your algorithm, you're non-neutral. Everyone has sort of realized by this point, certainly in this ecosystem, that there's nothing neutral. And where a lot of people are mad at you and there is no neutral position, you might have an incentive to offload some of the decisions just so people will be mad at someone else. Also because users might not trust you as the platform, and you might think it's better off that people will trust somebody else more than they trust me, even though they may not trust them fully either.
--:-- Andrew Hall: I totally agree. I think it's an exercise in trying to rebuild trust. And I think something Noah said is really important: trust in tech companies is falling in a lot of parts of the world, not in all parts of the world, but in many parts of the world. But at the same time, faith in government is falling in a lot of those same places. So there is no obvious actor to make some of these very hard calls in a neutral, procedurally fair way. The key is that as a user, you'd like to be in a position where you know a platform decides on a piece of content or what app is allowed or what transactions are allowed in a way such that even if you disagree with the particular decision, you still buy into the process by which the decision was taken.
--:-- Andrew Hall: Today, I think there's a lot of people who question both a company's ability to build such a process, but also a government's ability to build such a process. In the absence of trust in either of those processes, it makes sense from a business perspective to try to improve your trust in society as a company by finding a third option for a fairer way to make that decision.
--:-- Robert Hackett: This sounds like a really good segue into the oversight board, because that is an attempt to satisfy these conditions that you're talking about, to offload some of this responsibility onto a third party to enhance transparency and trust, hopefully, and create a process that is reasonable and rule bound for moderation and all sorts of decision making behind the scenes. Let's talk about the formation of that and the needs that led to its creation.
--:-- Noah Feldman: I think the core background situation that Facebook was facing, now Meta, when Mark Zuckerberg decided to create the oversight board was the recognition that there were some very, very hard content moderation questions. They were still pretty simple at the time. They were sort of, "Do you leave up content? Do you take it down?" They were not as fine-grained as they now have become. But in which he might have a strong intuition about what the right answer was, but reasonable people could differ. I think the insight that he was able to see is that actually there might be a right and a wrong decision. I'm not a relativist. I don't think that there are always two decisions. No, there might be a right and a wrong decision on this, but reasonable people could differ about what the right decision is. And it probably didn't matter that much from the standpoint of platform governance which of those two decisions were taken. But if he made it, it was going to take him a lot of time, a lot of effort, he had no special expertise in any of the underlying questions, and people were going to be really angry at him and distrust him. So no matter what he decided, it would decrease the legitimacy of Facebook as it then was.
--:-- Noah Feldman: So the idea that I proposed to him and that I think resonated for him was, maybe this decision doesn't have to be made by you. In the first instance, the company is going to make a decision. They're going to have a policy and they're going to try to file that policy. But if it's controversial, why not create a group of independent experts and bring those questions, the hardest questions, to them? Have them answer the question in light of Facebook's principles, international principles, common sense, good judgment, have them explain their decision and see if it worked.
--:-- Noah Feldman: I would say that the secret sauce there is the explanation of their decision. The secret sauce is that if you're answering a really hard question where you admit there are different possible answers, you should explain why you're doing it. That explanation, that practice of reason giving, has the capacity to confer a lot of legitimacy when it comes to hard decisions. Because I think it reassures people that thoughtful people gave thought to this, that they had some degree of expertise, that they deployed that expertise, and that they tried their best. It's not a magic solvent. Some people will still be really mad about the decision, but everyone will be able to say, I hope, this is the closest thing to a fair decision-making process that we could have come up with.
--:-- Noah Feldman: My last point on this is, obviously, this idea of legitimacy and fairness did not come out of the blue. It's borrowed from real-world institutions, especially courts, that often are not elected. Often their decisions aren't appealable. If it's like a Supreme Court or a High Court, they're the last word. They're not always right. They're often controversial. People get mad about them. But the institutions on the whole have a fair amount of legitimacy. I mean, it's a funny moment to be saying it now when the Supreme Court has done a lot of self-harm on the legitimacy front. But the Supreme Court still has much greater legitimacy in the eyes of Americans. And that's even after a range of other things that the Supreme Court has done that the majority doesn't care that it's undermined their legitimacy. So the idea that you can get legitimacy from reasoned decision-making was borrowed from that realm and applied experimentally to the oversight board. And I think on the whole, it's working. Not perfectly, but it is working. But that's a topic for further discussion.
--:-- Robert Hackett: You mentioned that this is a funny moment to be talking about the legitimacy of the courts. Now, of course, it's an institution that's hundreds of years old, but based on the way things have shaken out, is there anything you would have done differently in your conception of this sort of analog body to a court system?
--:-- Noah Feldman: If I can answer in the reverse way, there is something that we did that I'm actually kind of proud of which differentiates us from the Supreme Court. The oversight board is a pretty big body of people who decide cases, sometimes all of them but sometimes in subsets, and it isn't partisan. So it's not set up so that when you change the composition by a vote or two, everything flips on you, which is what happens at the U.S. Supreme Court. And then retirements and deaths. So that kind of, it can all change in a second. It's highly partisan and it's very responsive some of the time to political changes. Those are all features that we intentionally did not put in the oversight board to lower the temperature around its decision-making process.
--:-- Robert Hackett: Is this something that the government should take into account? Like if you were setting up a new government and a new court system, would you make changes of this sort?
--:-- Noah Feldman: Yes, absolutely. I mean, it's crazy that the Supreme Court... You have to go back to 1787. People died a lot younger. Being on the Supreme Court was not that good a job. A bunch of people who had that job would quit and go back to private law practice because they couldn't stand the job and it didn't pay enough and you had to get on a horse and ride around and hear cases all that. It just wasn't that pleasant. Now, people live longer. They get into that job and they never leave. They leave the court when they die, a lot of them, which is unhealthy, it's not good. And we also have a much more polarized politics than they certainly did at the beginning, although there have been periods of intense polarization in U.S. history.
--:-- Noah Feldman: So yeah, if you were designing it from scratch now, you would have staggered terms with time bounds. It wouldn't be the luck of, you know, Justice Scalia died, Barack Obama was president, but Mitch McConnell was able to block the confirmation hearings. That kind of crazy town, hardball politics should not affect the composition of the Supreme Court. Since one of my jobs is that I watch the court closely, part of my job shouldn't be having to have a pretty good amateur's knowledge of the health of the individual justices. If they're diagnosed with a certain cancer, what are the probabilities that they'll live and how long? That's distasteful and also absurd, that that should be part of our way we decide things. We shouldn't be thinking in those terms.
--:-- Andrew Hall: I want to raise one thing that I want to make sure we spend more time on: how the board is doing. I think there's a lot of interesting, different aspects of that to unpack. One thing I just want to raise that Noah and I have talked a lot about in the past and lots of other people have thought about that I think is important is one important difference between, say, the Supreme Court or U.S. courts or courts in general and the oversight board is sort of whose laws they're interpreting. So Noah referenced this earlier, that they're basically considering cases on the basis of Facebook's principles or Meta's principles. And I think that's the logical way to start. But one thing people obviously then point out is, well, the Supreme Court isn't referencing a single corporation's rules. It's referencing democratically written rules that the legislature writes. Nevertheless, it's an interesting thing to consider in the long run. Can you get a sufficient amount of legitimacy, as Noah called it, from a court online that's interpreting the rules of the company when part of the concern is whether the company is setting the rules in the right place or not? One thing people have talked quite a bit about, but I don't think we've cracked, is how would you democratize the legislative process as well so that something like the Oversight Board would be making decisions with respect to democratically written policies?
--:-- Noah Feldman: Yeah, this is super interesting. Let me add one more twist to the way things have been playing out, because this is something that has not been broadly discussed or broadly covered, except in a very narrow group of specialists, mostly academics and NGOs who follow the oversight board. The oversight board has been really worried about the thought that they're following Meta's rules. They want to be more ambitious than that. They want to follow something more democratic, but they can't use the laws of any particular country because there isn't one particular country that's binding. So what they've been doing is claiming to apply international law, the international law of free speech, which is not anywhere near as democratic as the laws of a democratic country, but it is sort of second order democratic because these are treaties that were enacted by countries, some of which or many of which were democratic and which all countries recognize as legally binding.
--:-- Noah Feldman: The upside of that is that it sounds a lot more legitimate than "we're deciding these important questions based on Meta's rules." And it's also a little more democratic. The downside of it is international law actually isn't that well set up to answer these questions because it's not designed to govern what a platform should do. It's designed to govern what countries should do. That's why it's international. It's the law of stuff between countries. So the Oversight Board has had to be really creative. They've basically, relying on some work by scholars, created a whole new body of what they call international law that they say binds Meta and should, in principle, bind other platforms. They're out there making this law, and there's no one to tell them no. I just supervised a student who wrote a dissertation saying this is unjustified. But my response to that was always, that might be true, but who's going to stop them? They're trying to do that. So that's a bid that is out there, and it's interesting to watch. As I say, it hasn't been very well covered.
--:-- Noah Feldman: But I also want to get back to the point that Andy's talking about, which is what are the experimental, creative things that a big platform like Meta could use to bring democratic input to bear on some of the hardest decisions that it makes.
--:-- Andrew Hall: Well, it might be good to start with the history, which is that this is something apparently Mark Zuckerberg has thought about for a long time. Back in the 2000s, when the platform was still pretty young, they had a sort of a disagreement among users about what the new community standards should be. And Mark held a global referendum on the platform. It's super interesting. You can actually, on the internet archive, go and see the video he posted encouraging all the users on Facebook, there were about 300 million at the time, to vote. What they learned through that process was, actually, it's really hard to get people to vote. Out of the 300 million users, I think something like 60,000 of them voted. The reason for that is probably because most people are not logging on to any part of the internet to do the hard work of governance. They're logging on to have fun or see their friends or whatever. In addition, the decisions being considered were pretty abstruse. You had to read these 100-page documents and stuff. So I think it was a really innovative idea that suggested early on that there's a way to potentially make decisions about a platform democratically, but it turned out to be challenging to make work in practice.
--:-- Andrew Hall: So now if we fast forward to today, I think lots of people are interested in ways to set at least some kind of broad, simple policies where the values that users hold are relevant to the decision through some kind of democratic process. No one has really figured out how to solve this core problem of participation, as well as informed participation, as well as if it's a platform that's global in scope, how do you actually get people in different countries who speak different languages to even work together to figure it out? What if they have radically contrasting values or goals? There's a lot of really hard questions, but I think there's at least two veins of experiments worth going deep on.
--:-- Noah Feldman: One, which I think Noah... By the way, voter participation rates are a problem not just in online voting but in real life voting too.
--:-- Andrew Hall: That's true.
--:-- Noah Feldman: Although the participation thing, you could make it mandatory. You open Instagram and you can't use Instagram unless you vote on these issues. But then you're not going to get highly informed voting. These problems that he's describing, I just want to make them concrete because I think they're two kind of good examples. One is the problem that I don't know if people who are listening have all heard of this or not because it may be a little dated, but they call it the Boaty McBoatface problem.
--:-- Andrew Hall: Oh yeah, this was Reddit, right?
--:-- Noah Feldman: It was the British government. The British government decided they were going to use this new amazing thing, the internet, to gather votes on what they should name a new battleship. The winning vote getter was Boaty McBoatface.
--:-- Andrew Hall: Which I think we can all agree is a fantastic name for a battleship. I don't see the problem.
--:-- Noah Feldman: Yeah. Her Majesty's Ship, I guess it was then, Her Majesty's Ship Boaty McBoatface. Somehow it didn't land for the Royal Navy.
--:-- Noah Feldman: And then the other example that I like to use, and Andy mentioned this, is let's imagine you're setting a nudity policy. There's one policy that works in the south of France on a beach, which is more permissive than what we have in the U.S. And then there's Saudi Arabia, which is a lot less permissive than we have in the U.S. It's not obvious that either of those should control the entirety of the platform. It's not obvious that our model should follow the platform either. It's mostly a default that a lot of these platforms are U.S.-based companies, and so U.S.-based standards end up controlling. Americans think, "Oh, those are the reasonable standards." But of course, there's no inherent reasonableness answer to the question of how much skin you should show. That's a classic example of something that different cultures are equally confident that their way of doing it is the only right way to do it. So that's a genuine challenge.
--:-- Noah Feldman: And then that leads to thinking, well, maybe we should have regions with different rules in different regions. And it turns out it's harder to do that on a global platform than you'd think. Most of the global platforms have tried as much as they can to avoid that because they don't want this kind of extremely balkanized platform where it all depends on where you logged in from. If you have a VPN, you can get different standards. You can imagine what a mess it can quickly become. So those are part of the challenges.
--:-- Noah Feldman: And then the last part of the challenge is all of our examples of democracy, and you can correct me if this is wrong because you're the political scientist, but at least the ones that come to my mind, all of our examples of democratic governance assume a denominator that is the relevant group of people. It's the people who live in your town are going to vote on town policy or your state, your country. We don't really have true democracy at the global level. So that raises the question of who's the denominator? Who's the relevant community to vote on these issues? Maybe it's the users, but a lot of issues affect non-users. So is it everybody, including non-users? Is it a subset of the users? That problem is, I think, the very first problem to grapple with. And in a lot of ways, it's the most challenging.
--:-- Andrew Hall: Let's unpack the Boaty McBoatface thing for a second, because I think it's really important. It's a great example of two, or really three, deeply related problems in the use of voting to make collective decisions. The first is just the lack of participation. Very few people actually voted on the Boaty McBoatface decision because why would most people pay attention or even know about it? Second, which is closely related to the lack of participation, is who then are the weirdos who choose to vote when not very many other people are voting? It's people who have very extreme, weird preferences, like the kind of people who think it would be hilarious if there is a vote named Boaty McBoatface. So those two things go together quite tightly: low participation, and then selection for unusual, unrepresentative, extreme views or preferences or trolling. So I think that's an important thing to understand.
--:-- Andrew Hall: And then the third, which is also deeply related, is just the general problem that you're asking people to vote on decisions and they have no skin in the game. That, again, lowers the incentive to participate and lowers the incentive for people with representative views to participate.
--:-- Noah Feldman: Yeah, what if you had to serve on that ship?
--:-- Noah Feldman: You might care a little bit more. Well, that kind of gets to Noah's denominator point. I think there's a very broad challenge. It's much broader than just democracy, which is how do we govern things that are not economic in nature in ways that are smart when the people we're asking to do the governance don't have skin in the game. And that was the problem with OpenAI's governance. It's been a challenge with university governance. The trustees in some sense don't have skin in the game. You measure the skin by how much money they put in. They would say, "I'm not saying I'm always sympathetic to them, but the donors, we're the only ones with skin in the game. We're donating the money. All you guys are doing is taking a salary." That's a good point. That's what they say. I'm not saying I'm on board with it. Anyway, go on.
--:-- Andrew Hall: Yeah. And when we look at all these voting problems, if you're asking people to vote on things that don't clearly affect them, that's a reason you might not get very informed participation. Furthermore, when you ask large groups of people to vote, you have this problem we call the paradox of voting, which is that even if you do care a lot, even if you do feel like you have skin in the game, if lots of other people are voting, then your vote doesn't matter. It's not going to swing the outcome. So you could care infinitely much about the decision, and yet your incentive to actually pay attention and vote is very, very small.
--:-- So we put those things together and we ask, okay, how are you going to do something democratic to make a big decision on a big online platform? I think you have to start with this fundamental set of problems: you want to get people to participate, you want them to pay attention, you want them to feel like it matters. How on earth do you do that?
--:-- I think there have been several directions of experiments that are pretty fascinating. One, which Noah alluded to earlier, is what Anthropic, as well as Meta and others, have been playing with, where you don't use voting, you use these things called citizens' assemblies. The other is what's going on in crypto, where you're experimenting with different types of voting that try to draw in only the people who have some sense of skin in the game through the tokens that they hold. So those are two sets of experiments we can learn a lot from.
--:-- I would just add one more framing thing here. Most forms of democratic government in the history of the world have tried to address these questions by saying, we're not going to have direct governance where every single person has to vote on every issue, because that's time-consuming. One way to solve both the time commitment and the skin in the game is to have representative decision-making. This is the move from ancient Athens, where every free man could show up and vote in the assembly, to a system where you elect people and your representatives are professionals or quasi-professionals who do the decision-making and the voting. They have skin in the game. They have the job of caring about getting reelected based on whatever incentives they have. And that gives them the incentive to try to guess or figure out what you're thinking.
--:-- One of the really interesting things about the internet, and I think it's important background when we dive into both crypto and citizens' assemblies, is that from its early days, the internet has been broadly skeptical of the idea of elected representatives. There's this impulse, and it's really interesting to explore why. Is it because of hacker culture? Is it because of coder culture? Is it the personality type of the people who first got really good at using the internet before everyone else? There's a lot of possible explanations. But there's been a deep impulse to get around the idea of elected representatives. Both of these experiments—the citizens' assemblies and the crypto DAOs or organizations—are attempts to say, we are going to reinvent the wheel. We're not going to do the thing that all democracies have done, namely rely on elected representatives. We're going to do something different and better.
--:-- That's kind of appealing, especially if you have some skepticism of our elected representatives. It's a good time to try to reinvent what they're doing. But there's also reason to be modest when all of the countries in the world that use a certain technology—democracy—have converged to a first approximation on one solution to this problem: elected representatives. And you're like, "Nah, they don't know what they're talking about. We can do better." Maybe you can. It's worth trying, definitely. But you want to have some degree of modesty about the probabilities of coming up with something that in 2,500 years of thinking about this, no one has really managed to solve in any other way.
--:-- Noah Feldman: I want to build on that quickly because I think it's really important. I completely agree with Andy a thousand percent. There's something about technologists and people online that makes them really crave direct democracy. One of the key claims they make is that technology is going to allow for direct democracy, that it didn't work in the physical world because it was too burdensome.
--:-- Robert Hackett: This is going to be my question. Is it a step change? Have we entered a new era where, based on the tech that's available now, this is conceivable?
--:-- Andrew Hall: There are no certainties in the social sciences, but this is as close to a certainty as I'm willing to go: technology is going to make no difference to this problem. Human nature is just the way it is. It goes back to the paradox of voting I was talking about. It's very burdensome to be asked to understand and become informed on every possible decision that a group is going to make. You might think—and this is the fallacy that a lot of people in tech have committed—that having more democracy by putting more votes to the people is going to get you a more empowered user base. But the opposite is in some sense true. The more things you ask them to do, the more burdensome it is for them to do it, the less that they'll choose to do it. That then creates a really important and dangerous vacuum because now you have very few people voting, and interest groups can come in and capture the decision-making process at a relatively low cost because there aren't that many other votes out there for them to compete with.
--:-- This is fundamentally why no successful society on earth has stuck with direct democracy for very long. And it's why technology is not going to solve the problem, because it's actually not a problem about voting in person being too difficult. It's a problem of information acquisition and analysis. People have also claimed that we'll be able to provide voters with all the information they need to make every decision through some kind of app. I've had a lot of undergrads in my office pitching me on the next app that's going to inform us all about politics so we can vote on everything ourselves. The problem is that information or data is not enough. You actually need to analyze it and then decide what to do. That's quite difficult. Being buried in data on what your community is voting on is just not enough to help you do this. It's a completely fundamental problem that goes way beyond technology that has to be solved through other models besides direct democracy.
--:-- I'll just note, almost all of the big voting DAOs in crypto now have some form of representative democracy as a result of this.
--:-- Noah Feldman: If you're listening to this and you think, "No, you're wrong, technology will fix it," what I would say is, if you're an aspiring technologist, ask yourself, what's your vision for the company you're going to found? In 99.9% of cases, people are like, "Oh, I want to found the company and then I want to have founder shares and I want to control the decision-making." And you say, "Why?" And they say, "Because otherwise my VCs will get involved, then there'll be shareholders. They're going to dilute my votes and take the company and do all kinds of stuff I don't want to do with it. And I'm the one who knows because it's my company. I spend the most time on it. I care about it." That's exactly the problem that Andy's describing.
--:-- I always find it amazing when those students come in and say that. I say to them, "Are you imagining a company where you will have no special input into how the company runs? Every decision will be made collectively?" And they're like, "Are you crazy? Of course not." And I'm like, "Well, then there's a contradiction in what you're proposing." The intuition that technology will fix this can be easily fixed by just asking yourself, when you get rich with your own company, do you want to be in charge of it? In my experience, almost everybody thinks the answer to that is yes.
--:-- Although there are some tweaks. Another thing that Anthropic did—and I know about this because I was involved in advising them on it—is they've created a Long-Term Benefit Trust. This is a trust that appoints at first just a few, but eventually will appoint a majority of members of the board of directors of Anthropic. The people on the trust who are going to appoint those board members are not themselves shareholders. So what Anthropic is doing there is not the full OpenAI thing where the parent board was completely made up of people who had no stake in the company, which blew up. This is a more modest in-between version where there will still be shareholders on the board of directors. The idea is to have some external check. It's a little bit better. But that's very rare.
--:-- Robert Hackett: So let's talk about the example of Anthropic. You're saying that in this case, there are actual shareholders that are able to serve on the board and that this is different from Meta's oversight board. And it's also different from OpenAI, where people did not have a stake in the company. So maybe there will be a more representative collection of views for keeping the concern going and also mitigating risks that could arise.
--:-- Noah Feldman: Yeah, it's designed to avoid problems on both sides. Take the OpenAI problem. You want to have some break-glass measure. The break-glass measure was they created this nonprofit entity, and that was the overarching body that the board of directors belonged to. Underneath that board of directors was the for-profit entity that we think of as OpenAI, the actual company that does stuff. If the people on the board of directors of the nonprofit—who had no financial stake at all—believed the company was going the wrong way, they could break the glass and fire the management of the company. And they did that. They believed that things were going terribly awry, that this was bad for the world. And they broke the glass and they fired the management of the company.
--:-- Robert Hackett: Until it backfired and the glass flew back in their face.
--:-- Noah Feldman: Exactly. They didn't realize that what would then happen would be that Sam Altman would say, "I'll just go to Microsoft and I'll take every single employee in the company with me." And then the employees all went online and signed up that they would go with him, which is maybe not so shocking because of course you're going to agree to jump ship. So then it backfired. The members of that board of directors had to resign.
--:-- What you need in the real world is something that doesn't go that far. You need people who are not financially incentivized who serve on the board, but also people on the board who understand the real world and who do have financial incentives. They develop a relationship with each other so that it is possible for the people who are worried about a problem and want to break the glass to do so in a thoughtful and responsible way where it won't blow up in their faces. That's what the Long-Term Benefit Trust is trying to do. It does give some real reasonable responsibility to people who don't work for the company. In that sense, it's a kind of cousin of the oversight board. But it's different from the oversight board because those folks don't have day-to-day decision-making responsibility. They're there at the level of oversight and break glass in an emergency. Whereas the oversight board is meant to be an everyday body that weighs in on hard problems Meta is dealing with. I would call them cousins. They have some conceptual similarities, but they're not siblings. They're definitely not twins.
--:-- Andrew Hall: It seems a little bit interesting that you have a certain set of people who have this opportunity working side by side with people who are, I assume, just getting a regular salary. And then how you get those people to really care about the decisions that are being made. They have to care based on their reputation. This was relevant at the oversight board, too. So how do you get the oversight board members actually to care that they do a good job? It's not so simple.
--:-- Noah Feldman: The short answer is, it's nobody's full-time job. The jobs that they have are as activists, scholars, or people who care about free speech. The idea was that they will care a lot about their own reputations and that will give them the incentive to do a good job. It's not perfect, but it was the best we could come up with. If it was their full-time job, their incentives might be too closely allied with the power of their institution. They still probably want to do a good job, but it wouldn't be that they had a reputation to preserve elsewhere. That was a complicated, compromised decision, because it turns out it's a hard problem. How do you get people to care about their jobs? Usually, they get paid more if they do it well, and they get fired if they do it badly. On an independent body, you can't really do either of those things. You can't reward them for making good decisions. And if you can punish them by firing them for making the wrong decisions, they're not independent anymore. So you need something somewhere in between. Reputation, I think, is the most powerful motivator.
--:-- You might think, "Oh, no one's motivated by reputation." But it turns out there are a whole bunch of people in the world who take the jobs that they have—a lot of them are academics—rather than some job that would pay a lot better, because they care about the thing they're doing. They like doing it. When I got tenure, someone said to me, "This is so great. You'll never have to do a day's work again." And I thought to myself, I never thought of it that way. I'm so much of a neurotic, I'm going to keep on working hard. But rationally, if I were a real rational actor, maybe I would stop. The main reason not to is then my reputation would be in tatters. People would say, as they do say about some academics, "Oh, there goes Feldman. He's one of those people who the day he got tenure, he never did anything ever again." And that hurts. And I care about it not hurting.
--:-- Robert Hackett: It's impossible to picture you, Noah, stopping working after tenure. Such a layabout. So Anthropic is also one of the examples of these citizens' assemblies. So it might be...
--:-- Noah Feldman: Yeah, describe the citizens' assemblies, Andy. Let's dive into that.
--:-- Andrew Hall: The general idea here is you have cases where you're either worried that too many people won't participate if you hold a vote, or that they won't actually know enough about the issue prior to having some kind of deliberation together or briefings on the issue to make an informed decision. So instead of holding a vote, you try to randomly sample a representative group of your citizens—in the case of a citizens' assembly—or users—in the case of one of these online user assemblies. Other people call them different things. You bring those people together. Because you've randomly sampled, they'll be representative of the user base of your platform. You get them together, you have them debate the issue, you have them hear briefings from experts so that they're informed. And then you have them make a decision through voting or some other collective process on whatever the difficult issue is.
--:-- This has had some history in the real world going way back. It's closely related to this idea from ancient Greece of sortition.
--:-- Robert Hackett: What is sortition?
--:-- Andrew Hall: Sortition was this idea that you would randomly choose people to be in charge of various issues. Election by lottery. Some people call them lotocracies. The idea is basically we need someone to have the full-time job of running some part of the government. We don't want to vote for them because that creates all kinds of weird incentives. So every year or twice a year we have a lottery and we pick one or two people who are going to have this job. They only do it for one term. They don't have to worry about getting reelected. But for reputational reasons, they try to do a good job because they'll look bad if they don't. So you pick them randomly and then you do it again the next year. It's sort of like jury duty, but heightened to the absolute utmost level.
--:-- Noah Feldman: Well, the Athenians had very aggressive jury duty. They had much larger juries and they paid them quite well, which is interesting too.
--:-- Andrew Hall: I think the main justification for sortition was this idea—I'm going to paraphrase it—but there's this great quote from Douglas Adams, the author of *Hitchhiker's Guide* and several other great books, that's basically like, "Any person who wants to be in charge is ipso facto unqualified to be in charge." I think that was the idea of sortition. We don't want these untrustworthy, overly ambitious, potentially corrupt or sociopathic people who desire power to be the ones in charge. We want the average, well-meaning, patriotic citizen to make decisions instead. The way we're going to make sure we don't allow that sociopath to insert himself into power is we're going to randomly choose who's in charge.
--:-- It's an interesting solution to that problem. It also raises a bunch of its own challenges. If you think that a big part of what makes democracy work is what we call accountability—the idea that the person whose job it is to pay attention, learn the issues, figure out what maps to your preferences, and make decisions on that basis, and the fact that they're worried about whether they're going to get to keep their job is what makes them do all those things—well, then sortition is a terrible method. I have no incentive to do anything. I've been randomly chosen to do this job and now I just have to do it no matter what. I don't have those strong incentives that are coupled with my desire to win reelection. So there's no accountability in some sense, unless considerations like reputation, patriotism, or genuinely just caring about the issue are dominant. That's one of the key trade-offs.
--:-- Noah Feldman: But if you do a bad job, maybe you'll get ostracized and exiled. They used to do that. They would actually do that too.
--:-- Andrew Hall: But there's another key trade-off. The background assumption of sortition is that you don't need very much expertise. That all the expertise you need to do the job, you already have or can pick up really fast. Imagine you're running your startup and you're going to choose someone to be your CTO. And you're like, "Oh, I'll do it by sortition." Bad luck for you if your lots fall on me, because I don't code well enough to be able to do a good job of reviewing the other people's code that's necessary for the CTO. There's no chance that in the six-month or one-year period I could get up to speed fast enough. Maybe if you gave me a year, I'd make some progress. But I don't have the relevant expertise.
--:-- So it's also partly based on a very amateur—in a positive sense—idea of government. You can imagine 19th-century British aristocrats who think that anyone in their club, if called upon, could sit in government and figure it all out. But life is very complicated today. The things that government does today are infinitely more complex than they were in an antique society in Athens. Even if you assume that people could have done a good job then, it's not so clear they could do it now. Even the Athenians didn't do it for certain jobs. They didn't pick their generals by sortition. They weren't dumb enough to think that they could go out and win wars by picking a random person to be the general. That required some expertise.
--:-- Then a third problem is, is this job the kind of job you get better at over time? Because if you get better at it over time, you're not going to get a chance to because you're in for six months or a year and then you're out. If it's the kind of thing where there are upsides to doing it for longer, you're going to lose those upsides. Obviously, there are also downsides to having someone in the job for too long. But it's sort of like term limits on steroids.
--:-- Noah Feldman: I completely agree with all that. Where this has gone—I was one of a number of people who helped design this user assembly for Meta. It's called the Community Forum. Based on these concerns, but also knowing that having straight-up user voting would be very challenging, we really wanted to focus on really value-laden decisions where it's not really a matter so much of expertise, but really about trying to capture in this community what your core values are that affect this decision. That might be a place where this is a more workable model. But I do think it's quite limited.
--:-- Andrew Hall: Going forward, my personal view—I don't know what Noah thinks—but my personal view is I think these things have to be expanded in some way to bring in more expertise and to allow for delegation, as we call it in crypto, because of the things Noah was just saying. I think it's particularly hard that someone serves on one of these things, learns a bunch, and then disappears. Developing ongoing expertise in the types of decisions that are socially fraught—going back to the beginning of this whole conversation, the kinds of decisions a platform might want to give over to its users—we're going to want people to develop ongoing expertise and a sense of accountability for that.
--:-- Noah Feldman: As Andy says, there couldn't be a lot of upside. Take a hard values-based, ethical or moral question. Let's go back to what should the nudity policy be on a social media platform? It's a hard question. You could imagine that if you take a thousand people and they're genuinely representative of the users—which is a tricky thing, as we've talked about, but imagine that they are. Maybe they all have an impulse one way or another when they start, but then you give them enough information for them to have some thoughtful conversations and then see where they land after, say, three longish conversations over three days. They may all land in a place that's a lot more thoughtful. In fact, they will land in a place that's a lot more thoughtful than where they started. They may have a different perspective than the people who work inside the company or the outsiders who start off very committed to one point of view or another, either because they're pro-sex and want to have as few clothes as possible, or they're deeply religious and want to have as many clothes as possible. You could imagine that you get—and I think it's plausible—and I think in some of the experiments that Andy's run with Meta, you get a more thoughtful, nuanced, balanced answer. For those purposes, it's great. It's better than a focus group because there's maybe a little less control. In focus grouping, the problem is if you're good at running a focus group, you can make the focus group say almost anything. Here, there are more people, more space, more protocols to stop the people who are presenting the question from driving the answer in a particular way. It just makes sense that we would get a better result. Not every time. There are some scenarios where deliberation produces perverse results, where people get into a cascade—this is like how people burn witches. Groupthink and herd mentality. You get into that and everyone goes towards some crazy extreme thing. That can happen. But there are a lot of techniques that Andy's design builds in that are designed to stop that from happening.
--:-- Robert Hackett: Part of this Community Forum is you're trying to provide information so people can make an educated decision about something and maybe change their mind so they arrive at a workable solution. But my question is, how do you ensure that the information and education that people are receiving is non-biased or objective? Whoever's presenting you the menu of options has a great amount of control in steering you toward a certain outcome. Good question for Andy. And then the other tricky problem is how do you make decisions about what's the right information? Imagine what we're discussing is precisely the misinformation problem. What's our baseline? Where do we even start on what is reliable?
--:-- Andrew Hall: It's a super challenging problem. I think there's a couple different answers, none of which are perfect. And this is a broad problem in all democracy, not just the online world and not just these assemblies, though it's perhaps sharpest in these assemblies since the entity organizing the assembly makes decisions about what information to present.
--:-- I would say a couple of things. One is you try not to have a monopoly on information. A big part of these assemblies is that people get to discuss and deliberate, and they're completely free to bring in any information they want gathered from whatever source they want. That's the first most important thing probably.
--:-- Then of course you try to do as good a job as you can of bringing in experts that are credible to represent the different sides of the debate. I think it's become in vogue a little bit on a certain part of the Twitter community, or X community I guess it would be these days, to say that there's lots of fraught debates in society where it's obvious that one side is factually correct. And so we shouldn't show both sides, that both-siderism is itself a problem. Of course, there are instances and places where that might well be true. But I don't think you can get people to make an informed decision about any issue if you don't let them hear both sides or more than two sides of the issue. Even if you think the preponderance of evidence rests on one side than the other, you really can't get to an informed decision unless you've heard all the arguments.
--:-- I think that's a big effort in this: get a diverse set of experts to present all sides. But long run, I think you want to be in a position, if you really can build out a democratic system that goes beyond these assemblies, you'd want to foster a competitive ecosystem in which different information providers crop up and provide their different analyses and views. That's something I've talked quite a bit about with DAOs because I think that's an issue in crypto too.
--:-- Noah Feldman: On that point, there's also a body of literature about people in brainstorming settings. When there are bad ideas proposed, it actually enables the group to arrive at better decisions over time.
--:-- Andrew Hall: Yeah, that idea, the classic modern formulation of that is John Stuart Mill. His argument for free speech was we actually need the bad arguments because working through why they're bad and wrong and false will help us get the right arguments. It sounds kind of like a trite argument, but if you really think about it, it's pretty deep because it's not obvious.
--:-- Noah Feldman: I think another really interesting aspect of this whole challenge is jumping out to the bigger political world of democracy. You were asking before, Robert, about it being a funny time to be talking about the U.S. Supreme Court as a model. I think the same is kind of true about these ideas of democratizing. But one of the aspirations of the community assemblies or forums that Andy's talking about is to get real disagreement that is thoughtful, that doesn't turn into polarized yelling at each other. One of the ways to do that is by narrowing down the issue.
--:-- Let me point to one weird feature of our polarization. People turn out to have really strong views about things they don't know anything about, and they have them instantaneously. This happens to me sometimes. I'm talking to somebody and they're like, "Oh, I'm sure I think X." And I'm like, "I've known you for 30 years. You don't know anything about that. Why do you have this strong position?" The answer is, as a time-saving device in a world where there's so many issues and they're so complicated, once we've picked a team—blue, red, libertarian, anarchist, whatever—there's now a list of positions associated with that team. We just default as a defensible, time-saving heuristic to siding with our team on those issues because we kind of think that someone else has thought about it, or a whole bunch of people have thought about it. And probably this is where I'm going to end up because I generally agree with this group of people on various things.
--:-- One of the things about doing a citizen's assembly or a community discussion about a narrower topic is you can sometimes avoid this habit of defaulting to a polarized position as a time-saving device because you're going to be told about it. You're going to get the information in front of you. It's not only that we're closed-minded and that we don't want to listen to the other side, although that does happen. We're also partly closed-minded because we're barraged with so much information that we don't have time to consider every perspective on every issue. We just don't have the cognitive bandwidth to do it.
--:-- I actually think that's one of the problems with our current polarization. It's not the only cause of it. There have been polarizations in many societies that do not have anywhere near as much information as we're getting. So I'm not making some claim that this is the main causal factor. But it is one reason that we do better in a well-designed citizens assembly often than we do out in the wild of politics.
--:-- Andrew Hall: I think that's been one of the most interesting and, to me, surprising learnings from them, both in the physical world and online. They're generally organized around the most fraught, culture wars type issues. When they've been run, one of the big learnings has been that if you get a relatively small group of people together and you structure the conversation well, people generally are pretty reasonable. It turns out they don't want to be so partisan or so hostile when they're put into the right environment.
--:-- That connects to other things about our polarization. It's much larger than it seems because we consume so much of it through these online platforms where we hear the loudest voices, but we don't see the in-person interactions that turn out to be usually less polarized.
--:-- Noah Feldman: I had a funny example of that from my own life. I was moderating a panel at Harvard and some students decided they were going to leaflet against me. They were handing out leaflets to people as they came in. The leaflet was like a four-page leaflet, but its text was entirely a download from a Twitter thread that someone had created attacking something that I had written. It was just so weird because it didn't translate well to the format of the leaflet. But also I just thought to myself, the extremity of the formulations that this person was using against me online didn't really translate to the forum where we were having a thoughtful conversation.
--:-- I sort of felt bad for the people who were leafleting because it's one thing to call what I said "vapid horseshit," which is what they were describing me as saying. In context, it seemed weird and rude to be saying that because everyone in the room was having a rational, reasonable conversation. Whereas on Twitter, I'm sure it seemed great. The punchline was there was a substantive issue where I'd made an argument in the thing I'd written that made them angry. Possibly wrong, but I made an argument and I gave evidence for it and I supported it. Their response to this argument was, "It's obvious that this is wrong, so we're not even going to entertain it." That just sounded so absurd in the context of a university. On Twitter, it sounds great, right? "Feldman is full of vapid horseshit, so why should we even bother to refute him?" All you need to know is that we think he's stupid, and then that's enough to reach the conclusion. But in personal conversation, very hard to carry that off without looking silly.
--:-- Robert Hackett: Since you brought up Harvard, are you feeling this intense tension between the administration and the student body? It just seems like such a powder keg over there.
--:-- Noah Feldman: It was a very, very, very intense fall. As intense as I have experienced on the university campus, and I've been on and off of this campus since 1988. So it was extreme. Now things are sort of slowly getting back to maybe normal is too strong a word, but towards a calmer way of being. I will say during all of this, our classes were normal. School went on. People learned stuff. They studied for exams. We did have a normal campus life, but all of the intensity that was being felt was felt. It was an experience, and for most people, I think it's fair to say they don't think it was a good experience.
--:-- Robert Hackett: Have things subsided a little bit?
--:-- Noah Feldman: Yeah, I think so. Universities are really bad at making fast decisions about anything. They don't like to do it. And if they do it, they make bad decisions. If your personality were that you could really react in real time, you'd be a trader. You'd have some job where you could really take advantage of that. Of course, there are some people in university who are really intellectually quick. But the best of them use that intellectual quickness to fill in their depth. And then you make the decision over time.
--:-- So universities are bad at the fast stuff. The fall was all fast stuff—responding to a news cycle, issuing statements and declarations. Now we're entering a phase where the university has two task forces already set up to study these issues. There are two more at least coming. This is going to be the part the universities do pretty well: take a deep breath, figure out what we've done wrong, figure out what we can do better, give reasoned explanations of how we should do better in the future. Our nervous system of the university will return closer to what it's best at.
--:-- I want to be clear. There are some people whose nervous system is always go, go, go. That's really valuable in a whole bunch of dimensions of life. But a university is not really one of them. We do our best when we breathe a little bit. That's now what's happening. And that's better. So we're heading directionally in the right way, though by no means are we there yet.
--:-- Andrew Hall: I think it's a really good and important example of just the general topic we've been discussing. A lot of it has to do with governance of these super important, very old, long-running institutions that are not straightforward businesses and therefore face quite complex governance issues. Certainly some of the ones I've observed over the last 10 years that make it so challenging is you have the exact same uneven participation problem. Not every faculty member leans in and participates in the governance of their university evenly. The ones who choose to do that may have different views than the ones who don't.
--:-- On top of that, you have the mission creep of the university, which is a lot less focused on protecting academic freedom and developing the very best research in the world and a lot more focused on a lot of other issues that extend way beyond research. It's hard to figure out how to fix those governance problems because the institutions are so big, entrenched in so many different areas, and trying to spread their attention across so many different things.
--:-- I worry that by making declarations, the institutional part of the university contributes to our understanding of the truth and of knowledge. I don't think that's true. If the deans get together and hold a meeting and announce that the second law of thermodynamics is true, I don't think that gives us much special knowledge into whether the second law of thermodynamics is in fact true. However, when the people who are at the cutting edge of some area of inquiry publish their research in a peer-reviewed journal, there's reason to take that seriously. It's not always correct. It's always subject to revision. But I would stop and listen closely to what they had to say and give it some epistemological benefit of the doubt because they are expert and they're speaking on their area of expertise in a thoughtful way.
--:-- I think one of the things that's happened is that our universities have kind of fallen into this part of the mission creep that I'm describing, of thinking that they have to express a public statement on every matter of public importance.
--:-- Noah Feldman: I understand the moral impulse to say the right thing, but that has to be balanced against, are you good at that? And are you contributing to the university's mission, which is to pursue the truth in the broadest sense through reading, writing, and teaching? That's different from being in the announcement and declaration game. I don't think universities are very good at that. Stepping back from that is by no means an overall solution, but it's a first step.
--:-- Andrew Hall: This is happening all across corporate America and beyond. These calls for activism and for institutions to take a position on various matters. I have clients coming to me, corporate clients every day, saying, "We're in so much trouble about this." I do try to tell them there's no such thing as being neutral. Genuine neutrality is not possible in the world. But within the framework of realizing that you can't be neutral, you can sometimes step back and say, "Look, we're not going to take a view on this or that important thing."
--:-- Companies find themselves in these positions because they're lobbied. They're lobbied by people who are trying to affect consumer behavior often. Sometimes they're able to pull that off. Meta has been subject to a boycott. Other big companies are also subject to various boycotts. So you are living in a real world. If you have customers, you have to worry about your customers. But you also have to be aware that your customers could be all over the place. That's one of the reasons to have step-back policies, including policies of referring something to some other group of people and saying, "This is a really hard one. We're not qualified to decide this. We're handing it off to somebody who is qualified to weigh in on it."
--:-- Noah Feldman: I agree. I think the key value is tying your hands. You have to make what we would call in the social sciences a credible commitment. You have to get to the point where the people who would try to force you to take a position on this issue that's irrelevant to your core mission, yet important to them, believe that there's nothing they can do to make you take the position.
--:-- I think the critical mistake that a lot of the most important universities made, as well as a lot of corporations made, was to give in to those demands and then create common knowledge that they can be forced to make those statements and that they aren't committed to not making those statements. People often point to the University of Chicago as one of the few universities that had less of a problem with this stuff because they had made a preexisting written commitment to not take such positions. That has turned out to be a huge luxury for them that other universities haven't afforded themselves because they hadn't made that commitment or the commitment wasn't credible.
--:-- The central problem is that over a 10-plus year period, a lot of the top universities demonstrated that they did not have a credible commitment to not taking those statements. Now they're trying to walk that back, but it's hard to create that commitment out of nothing.
--:-- Robert Hackett: I'd be remiss not to bring the conversation back to some of the subjects we were discussing at the very outset around internet governance. You're mentioning this concept of credible commitments in the context of the University of Chicago. But when it comes to making commitments with internet services and binding people to rules, I wonder what the path forward is there. When you look at something like the Meta Oversight Board, they have made some decisions that Meta doesn't necessarily have to adhere to. It adds a lot of transparency to the process, but I wonder if there are other systems that would bind services and corporations and applications to some set of agreed-upon circumstances. Of course, the thing that comes to mind is DAOs, distributed autonomous organizations, and how you can create these commitments in blockchain code that bind all participants.
--:-- Andrew Hall: I'll just say, my view on this, I think there's two things that are interesting about blockchain and DAOs with respect to this topic. The first is this trust problem and being able to write and commit to a process. One of the first things projects do when they start in crypto is essentially write a constitution. They write down in code how they're going to make decisions over different issues. Exactly how binding those are is up for debate. It's a little bit complicated, but in the long run with blockchain, I think it really is true that the constitution you commit to through that process in code is a lot harder to change without resort to some kind of democratic process than if it's not committed in that way, especially if it goes way beyond what can be protected through normal real world legal processes. I think there is something really interesting about being able to build an online community where you can make a sort of promise into the future that anytime this type of decision comes up, this is the process by which we're going to make that decision.
--:-- The second is more economic in nature. It has to do with one of the most important types of trust for online platforms: the trust between the different sides of the platform's market. Almost all online platforms have this characteristic that they're trying to bring together two sides of a market—drivers and riders, or developers and users for the app store, and so forth. You really need the producer side of that two-sided platform to believe that into the future, they're going to have an economically beneficial relationship with the platform.
--:-- One of the biggest challenges we're seeing in the space right now—look at the huge dispute between Epic Games and Apple, for example—is that developers are starting to feel like it's still a tremendous opportunity to develop on top of these extraordinarily good global platforms. But at the same time, the taxes they pay to the platforms are going up. The decisions made around their services are changing in unpredictable ways. They would like to have a much longer term promise about how the platform is going to treat them. That's another place where I think being able to make a long term, in some sense immutable promise over the economic relationship between a platform and its producers could be important.
--:-- Noah Feldman: In a broad sense, that's what a constitution is. The best metaphor for it is Odysseus in the Odyssey when he's on his boat and he's going to the island of the Sirens. He knows they're going to be so appealing to him, so beautiful, their song's so beautiful, that he's going to jump ship. So he ties himself to the mast and tells his crew members, "Don't let me go." That's one of the leading metaphors for what a constitution is. The idea is, we know that when the chips are down and we're in a panic, we're going to take away people's rights. We're going to silence them. We're going to take away their right against arbitrary arrest. So we try to bind ourselves from doing that. We try to create institutional mechanisms to bind ourselves.
--:-- Notably in government, there are two parts to that. There's the written rules, and then there's the human ability to interpret and maybe even override those rules. There's a productive tension between those things. It goes all the way back to a debate—I hate to be so Ancient Greek, but we've done a lot of Ancient Greeks today—a debate between Plato and Aristotle about which is better: whether it's better to have the rules in the end in charge because no one's perfect, which is the Aristotle view, or whether it's better to have the wisest person you can get your hands on in charge because the rules won't give you the best outcome all the time, which is broadly the Plato view. They're both right. You need some back and forth, some productive tension between those things.
--:-- That's true in DAOs. You don't want the thing to lead to a total spiral down, so you need to have some break glass measure. It's also going to be true in something like the Oversight Board, where the company sometimes has committed itself absolutely to following their rulings. Other cases, it's asked for an advisory opinion. But in the end, if Meta chooses to stop listening to the Oversight Board altogether, it could. They would just have to pay a reputational cost for doing it.
--:-- I think this really is in the realm of art rather than science. You always need to have some of each. You need to have some meaningful constraint, and then you have to have some capacity for flexibility. It'd be nice to say there's a magic solution to this, but there isn't. There's no purely technological solution, but there's no solution that has no technology because rules and the following of rules are a technology. You need both. That may not be the most thrilling conclusion—"Rules always win" or "People always win"—but actually, sometimes some wins and sometimes the other wins. That's real life. So maybe that's not a terrible place to end what I have to say on the topic.
--:-- Andrew Hall: I want to build on that in one particular way. One of the things we study a lot in the history of democracy is this idea of constitutions and what we call in political science and political economy, self-enforcing constitutions. There's no external authority that can bind a country to its constitution. The constitution only has power in the long run to force people to follow the rules if there's some track record of everyone having agreed to it for long enough, that it has some special power that it accrues over time. There's nothing to stop someone from tearing it up and ignoring it. That's why we call them self-enforcing. They only really bind into the long run when everyone has a long enough track record of agreeing to be bound by it.
--:-- This goes back to the Magna Carta where King John was forced to sign it. Then he was sort of like, "Well, actually, I'm not going to pay attention to it that much." Then there was a battle over the legitimacy of the crown and all that stuff. Tons of history of this. The U.S. Constitution is somewhat unusual in the length of time for which it's proven to be somewhat self-enforcing.
--:-- That same problem exists online. I think Noah is right. He was hinting at this, that blockchain or ways of writing immutable agreements into code do not, in a comprehensive way, solve this problem of self-enforcement because you still always have people with the option to fork and to leave and to do whatever they want, or to change the code as long as they agree to it. There's still something deep in what he was saying about art rather than science about it.
--:-- However, I do think for the online world, blockchain provides something pretty fascinating that I was really blown away by as a political scientist when I discovered it. I think it's best highlighted through an example. There's a DAO called Lido, and Lido has been discussing publicly for a while its desire to build in a veto for certain types of decisions. What they want to do is essentially update their constitution to say, when there's this one set of decisions, maybe the DAO just makes the decision. But when it's this other type of decision, it goes to an external veto.
--:-- When I was talking to them and to other people about how that would work in practice, one of my biggest concerns was that in the real world, if you try to set some procedural rule that only applies to one set of legislative votes and not another, then it's just obvious that strategic actors will channel it into whichever category is favorable to them. They'll just claim, if I don't want the veto to apply, they'll just claim this is one of the votes that the veto doesn't count on. Because legislatures get to make all their own rules, you really can't stop that. We've seen that in the Senate. Occasionally, the Senate parliamentarian tries to stop the Senate from doing stuff. There's a weird equilibrium where sometimes the Senate defers to the parliamentarian, but in the long run, the Senate has no real deep obligation to do that and can always change its rules if it wants to, if enough people want to.
--:-- So I kind of thought this veto is not going to work because you can just redefine Lido's votes into one category or the other. But I think the thing that turns out that I learned that's really interesting is no, because these votes, the topic of the vote is defined by the smart contracts that the vote actually touches or doesn't touch. You can define in a very deep and immutable way whether this is one of the votes that the veto is going to apply to or not. That allows for a form of commitment that I've never seen before in a legislature. I don't think it's a panacea. I think there are still these broader issues of self-enforcement that matter for all the reasons Noah's saying. But I do think it offers something pretty fascinating in the vein of binding commitments to legislative procedures.
--:-- Robert Hackett: Isn't this the miracle of the U.S. Constitution, Noah?
--:-- Robert Hackett: I mean, you've written a biography of James Madison, the fact that you can set these rules up front and yet make them flexible enough that they can endure.
--:-- Noah Feldman: Well, if we had Madison here, he would agree that it was his perfect design. You have to ignore the Civil War and some other blips. But yeah, I mean, the aspiration was to create something that would be able to be both rule-based and also responsive to change over time. Some of the design elements didn't work well. I mean, the amendment provision, we use it very, very rarely. He probably thought that was generally okay, but there have been some circumstances where we really need reforms and we can't really get them.
--:-- More fundamentally, it's a system that's designed to enable compromise and to enable compromise in the middle and to push politicians back towards the center. And that's why, right now, if you ask about the big question, the grand questions plaguing politics and political science, I think it's fair to say that if you have some confidence that the laws of political science are still true, then you think that we will migrate back towards the middle because our system is designed to push us back towards the middle.
--:-- And if you, on the other hand, are really panicked about where things are going and the possibility for breakdown, you think that it might take too long to operate. And then in the short term, you could get a breakdown in overall faith in the system's capacity. And the other part of it is looking at the world around us and seeing that for all of our polarization, lots of elements of our society are still functioning and functioning really reasonably well.
--:-- Robert Hackett: Amazing. This has been a fantastic conversation. Thank you both so much for all your time.
--:-- Noah Feldman: Thank you so much. Really fun.
--:-- Andrew Hall: Thank you. Yeah, this is great.
--:-- Robert Hackett: We sort of took it by fiat that direct democracy has problems. We mentioned Boaty McBoatface, which is a pretty humorous example. But are there more serious examples from history, from the physical world, of direct democracy in action and going awry?
--:-- Andrew Hall: The most classic example people point to, and which the America's founders pointed to, was Athens. Athens had a pretty robust direct democracy. The reality of how it functioned was quite complicated. It wasn't as simple as everyone showed up and made every decision. There were actually pretty complex layers of decision makers. But there was a significant component of direct democracy among landowning men.
--:-- One way that people say it went awry is that during the war with Sparta, the people got swept away with passion and were uninformed about the strategic decisions that had to be made. They essentially forced Athens to open a second front in the war by invading Sicily. That turned into Athens' Vietnam. It was a disaster. It massively sapped Athens' power. It was super costly. And that is often held up as an example of mob rule.
--:-- Robert Hackett: You said there are complex layers of decision making, but in this respect it was kind of a bottoms-up mob-like decision?
--:-- Andrew Hall: The claim is that the mob was manipulated by ego-driven generals who wanted to burnish their reputations by opening a new invasion, and they manipulated the mob into supporting them. If you look in the classical era, and you could question the motives of all these authors, that's the stereotype that arises. Post-Athens, a lot of Roman writing refers to mob rule and the ability of demagogues to manipulate the mob as reasons to be very skeptical of direct democracy.
--:-- There's a really famous passage in Virgil's Aeneid where he goes on at length about how cynical leaders can whip up the mob and what we really need are seasoned statesmen who will make decisions carefully on behalf of people.
--:-- More modern examples are hard to say. It's so dysfunctional that it's really not even tried. People don't even attempt it. You can point to Switzerland, where there's a pretty aggressive local referendum system. I think there's general agreement that it's kind of crazy, but it doesn't work as badly as some other things. At the Canton level in Switzerland, you vote on who receives passports. It's really quite wild.
--:-- Robert Hackett: Bizarre.
--:-- Andrew Hall: In the US, people point to California as pretty extreme on the end towards direct democracy. That can go both ways. In some ways, I think it's an important and potentially valuable institution because it allows voters to surface issues that the legislature might want to ignore. If you think your legislators aren't doing a good job or are captured by special interests, giving voters this alternative mechanism to force issues could be quite valuable.
--:-- The downside is that you end up with tremendous voter fatigue because you have lots of these votes. The California ballot is ridiculous. It's many, many pages long. Most people, including myself, can't understand most of the issues being voted on. The second problem, and this is often a problem in direct democracy, is interest group capture of the agenda. It's actually not that hard for a well-resourced, committed interest group to force carve-outs for themselves onto the ballot.
--:-- One thing that happens in California is every cycle, we have to vote on this extremely abstruse ballot initiative about whether doctors should be required to sit in on all dialysis. You might think it's doctors who are pushing this because they make money from it. But no, the doctors do not want to do this. They think it's crazy because there's no medical reason why a doctor needs to be present for the administration of dialysis. It sounds like a ton of time wasted.
--:-- Robert Hackett: It's crazy.
--:-- Andrew Hall: It's being pushed by certain interest groups that are in a conflict with doctors. They have stated publicly that the reason they put it on the ballot every two years is to force the doctors' interest groups to spend money convincing everyone not to vote for it. It's a complete waste of everyone's time and resources. That is a great example of how these processes get messed up.
--:-- Robert Hackett: By the way, the Irish referendum that just happened, my brother-in-law was just visiting from Ireland, so I was interested in this going on. It was interesting to see people reject so strongly these proposed changes to the Constitution. I don't know if you've been tracking it.
--:-- Andrew Hall: No, I haven't followed that very carefully. But I do think that's a good example where referendums, if they're embedded into a broader process in a healthy way, can be a really good way to get more signal from voters. This happened in California in 2018. We had a bunch of culture war related referendums or ballot initiatives. Quite consistently, the voters signaled through them that they were in a more centrist position than California's elected officials. I think that had a big impact on how the elected officials then proceeded after they lost these big ones.
--:-- So I do think it can be valuable, but it's complicated how you enact it in practice. I think that's where DAO governance is heading. There will still be some direct token holder voting, but there will also be a lot more delegation to professional experts on these issues.
--:-- Robert Hackett: You need to protect against mob rule, but also plutocracy.
--:-- Andrew Hall: Exactly. Find some sort of middle way. I think the delegation stuff is very valuable, but it's not going to solve the participation problem fundamentally. You're still going to need the token holders or the other voters to pay attention and make sure their delegates don't go rogue. That's going to require some pretty careful planning because there's definitely a temptation to set it and forget it in terms of delegating your tokens. But we have lots of reasons to suspect that if that behavior manifests regularly, the delegates won't have those incentives we want them to have to do a good job.
--:-- So I think a lot of the most interesting work right now in DAO governance is around how you build delegation programs that recruit and give good incentives to delegates, while at the same time still encouraging token holders or other voters to pay attention and to think about re-delegating their votes on a regular basis so that delegates feel like they're being watched.
--:-- Robert Hackett: What are people putting in place? What sorts of new rules or experiments are happening?
--:-- Andrew Hall: First thing that's happened, which I think is really interesting, is a bunch of the largest DAOs have instituted what I would call delegate programs. It's a combination of often paying the delegates, sometimes as a function of how many votes they accrue, and creating an online web interface that makes it easy for token holders to delegate to different delegates. You're creating incentives for there to be delegates, and you're helping token holders find delegates.
--:-- We've actually done some research. There's some pretty interesting evidence that rolling out those programs actually does increase token holder voting participation, probably because you're asking the token holder to do a much easier task. Instead of voting on literal changes to the underlying protocol's code, you're just asking them to find one of these delegates who they like and give them their voting power. Part of these programs is also having the delegates write platform statements or post videos about what they want to accomplish as a delegate. That helps the token holders find a delegate that shares their views.
--:-- Robert Hackett: Fascinating. So this is a paper that you're working on currently?
--:-- Andrew Hall: Yeah, I'm hoping to have it done in the next month or so. We'll see.
--:-- Robert Hackett: Very cool. So basically getting people to delegate their votes more, making it low friction, making the information available for people to make reasonable decisions about who they want to delegate to and the directions that they'll take a DAO.
--:-- Andrew Hall: Exactly. I think it's going to be a really important model. This came up in the conversation with Noah. I think a lot of people in other parts of online governance are coming to the same conclusion that we need representatives because we need this kind of expertise and accountability and we can't rely on direct democracy. But Web3 is years ahead, literally years ahead, in experimenting with how you actually set up representative democracy. So I think it's going to be a really important development.
--:-- Robert Hackett: We've got this intense Darwinian combat going on, all these experiments just let loose and seeing which ones will flourish. How far away are we from actually determining what works, what doesn't work, and seeing the kind of best practices shake out from all this?
--:-- Andrew Hall: I don't know. I think there's probably two aspects to it that we need to have happen before we'll know for sure. One is to get DAOs to a place where they have broader killer use cases for society, which will in turn make the governance decisions higher stakes for society. Then we'll see which governance structures can stand up to that pressure. In some ways, they're already involved in high stakes decisions in these DeFi protocols. But they don't have the same kind of global public pressure on them that other online platforms have faced because those platforms are much more mature and have many more regular users. I think that's going to be a really big, interesting shift as it occurs that's going to bring DAO governance more together with what we talked about with Noah in terms of Web 2.0 governance.
--:-- The second is we need, over time, to have more DAOs with broader distributions of voting power, which is happening over time. A big criticism of DAOs historically has been that they're clothed in the rhetoric of democracy, but the voting power is very unevenly distributed. That may make some of this delegation experiment unrepresentative of what would happen in a broader democratic system where there's more conflict among users with roughly equal amounts of power. That is the trend, I think, in token holding over time. As these become larger, as they touch on more killer use cases for society, I think we'll see that broader distribution. That will be another kind of pressure test for these systems. Those are the two things I'm keeping my eye on.
