A RadicalxChange between Vitalik Buterin and Glen Weyl

The following is a series emails Vitalik Buterin and I exchanged over the last day about RadicalxChange ideas. We thought the discussion might be useful to some as a) it covers a number of issues not discussed elsewhere that we consider important, b) it represents some of our latest thinking about these issues and c) it shows a bit of “the sausage being made” that some may find interesting. However, be aware that this is an internal communication and thus is at a pretty high level of specialization; there will be many parts that those not already well steeped in some combination of RadicalxChange ideas, economics, sociology, intellectual history, philosophy and cryptography may find hard to follow.

Vitalik Buterin:
I thought I would try my best to put into words the issues that I think still need to be addressed better. Not all of these are fully-formed thoughts, but they are things that I think are worth thinking about, that I feel like we haven’t been thinking about enough.
* 1. Collusion resistance, and more broadly not privileging people with better ability to coordinate (I’ve mentioned it before, but want to expand on what concerns me). For example, if QV/LR is used online, if you have a family of 4 people, you can get everyone together to open their laptops and each person see everyone else’s contributions, and so if everyone agrees to contribute minimum of what everyone is willing to contribute, every person in the group would be willing to contribute 4 times more. Even if contributions are done in a voting booth-type setup, groups of people where everyone cares about some issue but one person cares a lot could see that one person ensuring that everyone gets a lot of money going into the booth, and then everyone in the group would feel honor-bound to contribute more to the given cause. 1P1V does have this problem, but it’s not a big deal because there’s a natural ceiling, where everyone in a given group votes, and so we can try to get *everyone* closer to the ceiling by making voting maximally easy (the extreme is making it mandatory, but you can get most of the benefit by dropping costs, eg. by having more and well-trained voting stations, and having a holiday on voting day). With a QV/LR regime, groups with a 2x better ability to use the kinds of techniques I describe above would be 2x more effective at achieving their goals inside the system; that too me feels like a potentially scary level of distortion.
* 2. The role of social motivations — one of the bigger critiques of economics is that humans have many motivations that specifically directly have to do with the form of relationships that they have with other people, including differentiation, belonging, fairness, domination, avoiding submission, etc etc. These kinds of motivations make economics problematic because they mean people have plenty of opportunities to impose negative externalities on each other even without violence or fraud (eg. I buy a [fancier suit | well-designed website | college degree], I look more respectable and people pay more attention to me, but attention is zero-sum so people pay less attention to you….), and they peek into the structure of mechanisms in cases where the mathematical proofs of their optimality implicitly assume no one does (eg. people getting upset when they learn that they are getting lower salaries than their coworkers). A grand economic ideology that seeks to be useful in guiding society should have some conception of how the economics interacts with all of these other factors. In the case of radical markets ideas, how does it? Do they interfere, can we sidestep most of the issues by making voting/contribution private and non-provable, can social motivations be leveraged to make radical markets ideas work better? Is this something that can only really be understood once we have actual real live experiments?
* 3. Rational ignorance | irrationality — basically Bryan Caplan’s argument that people’s vote has a negligible impact on society, so they have basically no incentive to learn how to vote in the way that actually maximizes what they care about. Believing this does not require believing in atomistic selfishness; the thing with social motivations is that even though they exist, they are often wildly uncorrelated with consistent moral systems that make sense at large scales. If the farmer in Kansas believes apples are flat, he’ll have a hard time farming; but the farmer’s motivation to believe vs disbelieve anthropogenic global warming is decided more by his human desire to fit in with his friends than any direct connection with the truth. I understand that mathematically speaking, this problem is smaller in QV/LR than in regular voting, because people can specialize, and people can more easily see the results of specific changes that they played a significant role in causing to happen. However, is this improvement large enough to counteract the increased scale of the problem that would arise from voting-like mechanisms having a much larger role in society as you advocate?
I think looking at governance proposals with more elements of “skin in the game” would be interesting. This is part of why I liked your immigration proposal that included the idea that sponsors would be fined if the immigrants they sponsor commit any crimes.
* 4. Effect on centralization of physical power — one thing that scares me about more complex systems of property rights is that they would require more complex centralized infrastructure, including surveillance into people’s private activities, to be able to correctly enforce. Taxes already have this problem (you may recall Adam Smith believing that income taxes would be impossible because they would require an unacceptable level of intrusion into people’s private lives to enforce), and I wonder if the various proposals that we have for changing them would make things better or worse in this regard. I like Harberger taxes because they don’t require infrastructure to police whether or not undeclared transactions took place, though I worry in other cases, eg. your comment that your immigration proposal would require stronger enforcement of immigration rules, which realistically means stronger efforts to find and kick out people who overstay, which requires more surveillance of various kinds. All in all, I don’t think the radical markets ideas altogether fare that bad, but I guess my comment would be that non-panopticon-dependence should be an explicit desideratum to a greater degree than it is now.
Glen Weyl:
These are great points. Some thoughts below. Overall, my perspective is these are good issues, and I will add some of my own, but I don’t really view any as impediments to the experimentation with the RxC paradigm in the near-term…almost all seem like good seems for learning and basis for improvement.
  1. I think these issues around collusion and coordination are the critical issue, not just for QV/LR but kind of for everything. To the extent there are “laws of nature” (like people can coordinate in groups of no greater than X people) I think we can largely solve these issues with adjustments to the function. The problem is that people’s ability to coordinate is highly variable and contingent, and uses a range of mechanisms from kin ties to religions. It seems to me that QV/LR has the advantage of at least making it clear what optimality is and what to try to avoid in terms of extra-system coordination. In our present mix of capitalism and 1p1v it is not as if coordination is not critical, nor that it is at one scale or others. Everything from the ability to inspire corporations, build networks, create class solidarity etc. plays roles like this and is the primary source, I would argue, of power. But the system is very unclear and there is no reason to think that anywhere we are even approximating optimality nor is it clear even how to push towards this. Having a system that leads to optimality modulo outside coordination seems like a step in the right direction, but it will then require generations of work on breaking down extra-system coordination or understanding it well enough to balance it.

2. This is an absolutely critical point and the focus of some work I am doing, though not one I have told you about. In relation to QV/LR, I really do not think this this is a major issue. Fundamentally I just don’t see why this would be worse or not dealt with in the QV/LR context by secret ballot. In fact, current preference intensity things, like protests, are far worse along this dimensions. Having a private way to do this seems a massive improvement. However, returning to the research I am doing, I think much though not all of this can be captured by signaling theory. And the basic prescription of signaling theory, almost always ignored, is that you should tax the signaling action. Tax SAT scores, tax using standard white English etc. This policy prescription seems to have somewhat more challenges in implementing (as you have to measure the positional externalities, though we have made some significant progress in how to do this in a pretty algorithmic way), but if feasible seems exactly right and very radical. It would be efficiency enhancing and redistribute what many people call “privilege” (as in “white privilege”) in just the way that the somewhat incoherent leftist and rightist critics of various cultural paradigms seem to be asking for. In sociology they call the ability to navigate systems of meaning making to your strategic advantage “cultural capital”. I am working with Kadeem Noray both an academic paper related to this called “A Signalling Theory of Cultural Difference” and on a white paper on “Efficient Cultural Capital Taxation”. If it interests you to join it, let me know.

3. This is another critical issue not covered in our work, but I definitely do not think anything like status quo institutions will handle it better. Basically, because of this issue, you cannot have anything like direct democracy, or direct democratic control of “corporations” etc. You need some more concentrated pivotal group. The optimal degree of concentration depends on the elasticity of information acquisition v. the heterogeneity in information aggregation and I do not think there is a uniform solution. I therefore strongly prefer some sort of “liquid democracy” type (though not liquid democracy itself) system that allows the optimal organizations to emerge organically from the population. I do think this is roughly what will happen from LR, and this rational ignorance is precisely a reason for economies of scale in organization and why I oppose fully “decentralized” schemes for e.g. data management. This is not to say that there isn’t a lot more research to be done incorporating information acquisition and thinking about how this should structure things, but I doubt there is a “master algorithm” here and instead think it will depend on cases and we should resist a once-and-for all structure like a parliament or president, but also resist clearly inefficient capitalist mechanisms that put some tyrant who can grab resources in some scramble from dominating. I think LR does this pretty well but we’ll have to see.

4. I strongly agree with this last point (though I think exactly what is meant by physical power is subtle, more on this soon). Again, I actually think overall RM improves on this, but not by enough. This is why I increasingly have come to believe in what philosopher Danielle Allen has just suggested the name (for something we may be working on together) of Polypolitanism. This is naturally suggested by LR. Imagine we have LR as some sort of a global principle, but we have a wide diversity of communities at different levels with different degrees of strength and funding that enforce various things. People may contribute to/be members of many of these and not all are geographically organized. While each has some enforcement, a fair bit of enforcement is also social as the ultimate rules of the system are widely viewed as legitimate and thus when people are seen as violating them many around them want to turn them in or even (as under Harberger) have an incentive to. This might even be true in the immigration system; we would have to think about it, but in this Polypolitan world the right immigration system might be quite a bit richer than what I describe in the book as it would involve movement into these different forms of governance/collective organization. Precisely what this looks like and how it works require a great deal more work, but I think fundamentally what it points to is that the notion of “central authority” is fundamentally broken. There is no central authority in the world, and it is not a reasonable approximation to think of the nation state as that especially in this world. The problem is we have lots of legacy institutions that are ill-adaptive and we need a more fluid way to construct and deconstruct these at a wide range of levels of social organization, which is what I hope LR affords us. But in any case, “physical power” is a very complex and not clearly meaningful concept. If Hitler tells someone to do something, they do this because of their expectations of what others will do if they fail to obey, not because of any physical power Hitler has. Every social organization is in this sense a bubble, which is what makes blockchain and the debates it engenders about legitimacy so relevant and fascinating.

Let me add one more concern that I think is absolutely critical and that I am still working on and will require much more research. The idea of distributing the funds raised by the COST as a social dividend is ill-defined and a complete kludge and I increasingly thing a dangerous one. For example, suppose that a community can influence individual valuations and that this community has greater (less) share of the value than the probability that that community or someone within it wants to purchase the asset from the current possessor. This will lead the community to want the citizen possessor to over (understate) the value. That will raise some of the potential collusion issues like those you raised above. In general, it seems to me, ownership shares need to be allocated as much as possible across communities and individuals in proportion to the probability that those communities or individuals are (in discounted present value terms) the efficient owner of the asset eventually. This would naturally lead to some combination of various uneven income flows to individuals and funding of public goods at the level of and within communities, avoiding Jaron’s concerns about central homogeneous flows which I increasingly share. It would also be much more efficient, less subject to risks of totalitarianism, and just more robust and stable. How to figure this all out and make it work, though, is tricky. It feels like there should be some elegant approximations floating around, but I haven’t closed the loop yet. In any case, I think this is a critical area to develop further.
Finally, a broader point that has really come home to me over the past couple of weeks talking to a wide range of academics about this stuff. I have come recently to feel that, especially with Liberal Radicalism, I’ve now made a breakthrough that lets me really reorganize thinking into a truly coherent political economy framework that integrates all the children fields but also really makes clear their fundamental failings at present. Basically, LR (as well as the signaling issues) highlight how bankrupt current economic discourse with its extreme individualist focus is. This is just not how social life is every organized; we live in societies where almost everything is produced by some at least partially increasing returns technology (from family through churches, networks, governments, corporations, etc) and economics takes as the basic starting point a totally individualist perspective that only works for decreasing returns stuff and when confronted with increasing returns has to treat it with market power or other awfully inefficient kludges. This makes no sense and once you look at things this way you can really cut through most of the standard problems that come up in economics. The last couple of weeks talking to economists, sociologists and philosophers I have felt like they are hacking through a forest with pen knife and this perspective enables me to look from above (things still fuzzy) and have a crew of chainsaws at my command. And the great thing is, you don’t need to exit the formalism of economics too much, just take it to its logical conclusions, and you end up in a substantive place that sociologists, anthropologist and political scientists have been for a long time, but without the ability to quantify, measure or precisely invent or prescribe because they lacked the formalism. Even if the particular RM ideas have weaknesses and need to be adjusted, I do truly believe we are push towards a powerful and integrated view of political economy as distinct from the neoliberal paradigm as the marginal revolution was from what came before, and in very parallel ways. Fundamentally the pre-marginal revolution paradigm depended on linear production functions and didn’t allow for the fact that marginal>average, so couldn’t handle profits, etc.. But somehow the marginal<average (public goods) got lost as a central problem we could actually address, and I think this is precisely what George and Walras were after. I think we now can really grapple with it.
Vitalik Buterin:
Definitely a great point that traditional liberalism doesn’t do a good job of handling scenarios with increasing returns to scale. I would also argue more broadly that economics based on equilibria and local optima and derivatives doesn’t do a good job of capturing how to deal with scenarios where there are multiple equilibria, as well as coordination problems. Whether or not to start engaging in an activity that has increasing returns to scale is just one example of this broader pattern. Adding or removing common holidays (as free time is more valuable if your friends are also free during the same time), switching time zones, regularizing irregular verbs in some language, etc etc. Having stronger social gadgets for making those kinds of agreements is something that seems quite valuable, and it seems also valuable to have those gadgets be constructed without requiring points of central control, both because central control can be abused and because sometimes (eg. think protocol standards that are trans-institutional and trans-national) there just is no central authority that would be up to the task.