Glen Weyl presents his ideas around Social Technology (tools and tech for cooperation and governance, that help the public to be a good public). He discusses when capitalism works, when it doesn’t, and what other methods might look like that can serve us well when existing capital systems don’t. He showcases Quadratic Funding, an experimental mechanism for allocating public resources and funding public goods developed by RadicalxChange. And he illuminates the open questions in how to develop political and economic systems that take more of the real social and political life of people into account, institutions that are really in support of how we live. This meeting is part of the Intelligent Cooperation Group and its accompanying book draft.


Presentation: Glen Weyl



What the inventive genius of mankind has bestowed upon us in the last hundred years could have made human life carefree and happy if the development of the organizing power of man had been able to keep up with his technical advances. As it is, the hardly bought achievements of the machine age in the hands of our generation are as dangerous as a razor in the hands of a 3-year-old child.

  • Albert Einstein, 1952

Over the last few decades, communication technology has advanced astoundingly, moving us toward richer and richer representations and higher and higher fidelity communication.

This same level of advancement should and can be achieved for our political and social technologies. There is an enormous amount of richness in our interpersonal lives, within our “Dunbar groups”, that is not present for larger scale and anonymous social interactions. Let’s change that. Enter SOCIAL TECH.



I. Why increasing returns undermine capitalist optimization

    • Prior to the 19th Century, the “labor theory of value” reigned: there was a belief in a linear relationship between labor inputs to a company and its outputs.
    • As the industrial age bloomed, this belief was updated. It was found that there was a sweet spot, a point beyond which adding another more inputs increases output less and less. Think of a gold mining operation, where the cost to mine more increases more and more as the easy to reach stuff has been mined.
    • We call this phenomenon “decreasing returns” and capitalism works when you in a context of decreasing returns! When everyone is paid out their fair share, there’s some left over as profit.
    • In contrast, there are economic contexts in which we find increasing returns. The whole is greater than rather than less than the sum of its part, and adding more inputs increases output more and more rather than less and less. This is what’s sometimes referred to “economies of scale.”
    • The world is full of these economies of scale, and yet the capitalist instutions we make use of are not designed to allocate resources efficiently in these contexts. If everyone were paid out their fair share, it would be more value than the enterprise is generating to begin with.
    • What happens when you try to have profit seeking when you have economies of scale? Either you put some kind of monopolistic structure in charge, concentrating power and returns in a small few, or you don’t do that and the business goes bankrupt. Think Big Microsoft in the 90s vs. an open-source software project with no money to pay contributors.
    • BUT! All the great examples of “Capitalism” are actually these economies of scale situations: this is fundamentally inconsistent with the claims as to what capitalism is good for. What to do?


II. Optimal institutions for increasing returns

  • Quadratic Funding
    • Quadratic Funding is a mechanism for resource allocation developed by Radical Exchange, what they call an optimal mechanism for the provision of public goods. It’s sort of like the idea of “matching funds” that’s quite common in charitable giving, with some different considerations.
    • Basically, the smaller a person’s possible contribution, the less likely they are to actually contribute because their contribution doesn’t matter much. Taking that into account, you can set up a system to weight the contributions by size, reversed. Technically, you want the amount received to be the square of the sum of the square roots of each contribution.
    • This flips the script from, “What if I contribute 1 dollar more? What’s the cost-benefit analysis of that?” to “What if I could get everyone to contribute based on how much they care?”
  • Well, where do the matching funds come from?
  • There’s an economic idea know as the Henry George Theorem.
    • An example: Everyone knows areas with really good schools have high house prices. Sometimes this is a 1:1, if you build a school, the house prices go up a predictable amount. If you can quantify the value increase made by the school, and you could somehow get access to those funds, then you could fund any school that costs less to build than that increase in property values that it drives.
    • A gernalization of this is the Henry George Theorem. The increasing returns activities will throw off value but not be able to absorb it while decreasing returns activities will soak up value.
    • If you can recycle the value accrued by the decreasing returns activities (through taxing perhaps) into the increasing returns activities then you get this self-fulfilling kind of superconduction.
    • But how can you do this without the tax impacting the economics negatively? Enter the SALSA: Self-Assessed Licenses Sold at Auction. Under this scheme, property owners name their price but are required to sell to any buyer who meets the price.
    • There are all sorts of impediments to property being repurposed for new uses, in both private market contexts and political contexts. We have regular elections so as to reallocate “political property”. What this sort of compelled sale idea, you can build guaranteed periodic reallocation into the economic system.
  • Uses in the Wild
    • Taiwan is using quadratic voting for many civic services and votes, including their Presidential Hackathon.
    • The State of Colorado is using quadratic funding to allocate it’s budget for the second year this year
    • Gitcoin is a crowdfunding service on the Ethereum blockchain that uses quadratic funding


Ill. Next steps for “optimal institutions with pervasive increasing returns

  • Problem: the examples really only work in theoretical contexts where there is one public good and the rest of the economy is private goods, a sort of generic context. But the world is specific, and the social realities of the world must be taken into account.
  • Many of the participants in a market that are vying for access to the same kinds of goods are in relation to each other in some ways. If the good is truly a public good, in that it may enjoyed by those even who didn’t fund it, that affects the allocation of resources. You or your spouse winning an auction are not unrelated outcomes.
  • How do you deal with the fact that people have different social relations, and these social relations will affect how cooperative and competitive those people behave?
  • The solution: a new framing. Rather than a list of people each with an amount of money, you have a graph of people with weighted relationships amongst them. You can then run political economies atop this graph. This non-linear weighted graph allows for blossoming complexity and rich social structures on top.
  • There are so many things missing from the description of social and political life optimized for in the design capitalist institutions: signalling, cultural undestanding, representation, non-linear ways in which we combine information, community structures around property ownership (not just individual property rights), democratic governance of collective entities, etc.
  • These are so clearly part of human life but not part of our institutions. As AI is reaching closer and closer to human-like cognition, and VR is reaching closer and closer to human-like sensory experiences, why can’t we do this with our political and social tech?


IV. Broader agenda

  • Let’s get these techniques and technologies into use with companies, collectives, activist organizations, everyone! Let’s kickstart the Social Tech scene evolve our ability to cooperate.


Doesn’t seem like employers actually do the decreasing returns calculation when they decide their employees’ salaries… is there data that this is actually happening?

  • Glen: Well, I mention the theoretical case only to show that since the world is full of increasing returns, and many companies are out there making profits, the returns of that economic activity must not be distributed inefficiently or the companies would all be bankrupt.



I work in optimal use of health records. We model self-sovereign communities and patients and doctors to vote on how data is used and the like. We’ve been struggling to figure out how to introduce a business model into this setup. How would a nonprofit encourage this non-linear social graph sort of business?

  • Glen: I don’t think these ideas are relevant to very small organizations that can rely on interpersonal relationships. Once the organizations scale up to where these tools are useful, you have to put really sharp boundaries around the community.
  • If this model works as well as we think it will, one of these orgs is going to swallow up the whole world because of its extreme productivity and growth.
  • As for nonprofits, the move is to bring together charitible or government funding with the management of the community, so you can use these funding mechanisms to allocate resources within the community



Degree of coupling graph is very powerful thing to show

  • There are life partners, close friends, aqaunintnces… what should we do with that information if we could track it all? If we start building things on top of that info, does it mess up the graph?
  • Glen: Well these are all still quite experimental. Quadratic Funding and the like I expect to see in use in ~15 years maybe. Harnessing the social graph iteslf, even further out. BUt I will speculate.
  • you want to find an eigenspace of that fgraph, and then set up an economic system where you’re doing some kidn of quadratic voting on that eigenspace rather than ont he actual social space
  • you want to do something where basical the whole setupof the problem is about the issue of difference
  • taxes and profits would be recycled to subsidize cooperation across distance in the graph, which would transform the graph
  • how we usually think about it are atomic indiividuals acting in a market, but that’s silly. We are in a big dyanimcs always evolving social graph of people with all h
  • what are acutallt the underlying factors out there hat determine cooperation? races? affilliations? These are driving ther est of the system
  • when people cooperate, see how similar they are
  • constant ecological evolution of relationship and cooperation
  • this graph of relationships replaces money as a the reputation system for the world



I’m concerned with the idea of this monitored, constantly updated social graph. There seems to be a pretense of knowledge, central planning fallacy here. The idea that theres any kind of external objective measure that not going to be gamed, that’s just not a workable idea.

  • Glen: Agreed! Any version of this should be decentralized and this information should be stored privately. The strenght of my connection to someone is like money. I’m incentivezed to have strong connections. But the weaker my connection, the more subsidy I get from the system to cooperate with this perons. So these forces may combine to generate incentive-compatible information.
  • Because it’s a graph of social forces and not individuals, we have to really play with what incentive-compatibility means in this context.



This is dynamic social graph thing seems like a system that requires information you’d never be able to get. I don’t see how you can model it in a way that doesn’t do violence to the underlying reality. How does this translate into policy suggestions or other actionables?

  • Glen: I have a range of horizons to think about Social Technology. I’m really in a theoretical and speculative place here. Some tools we’ve built and work today, but the rest is very much a speculative probing into the future. I see dearth of focus on these technologies, and I’m taking a crack at potential solutions.



Gitcoin uses Quadratic Funding: how is it working? Some people claim it allows for “invisible corruption”,

  • Glen: As Gitcoin has scaled, it’s run into THE problem of Quadratic Funding: the model of people as individuals that are making contribution decisions in isolation is just not how the world works. People have relationships, they work together. This isn’t “collusion” really, it’s just how people work.
  • How do you overcome that? There will need to be a culture around its use, tools and best practices and all sorts of supporting infrastrcuture like we’ve built around our current models for voting and funding.
  • When you scale up to huge size, you have to start putting in tighter and tighter controls to maintain community quality and prevent attacks and corruption.



How does AGI play into this?

  • Glen: First off: I don’t think that AGI is a useful concept that warrants analysis. I think that an alternative to reaching for what folks call “AGI” is the reach for some sort of system that allows for massively increasing our ability to cooperate across lines of differences and to allow for an ecology of increasingly differentiating and divergent agents to cooperate. I think that social technology could be a core part of that vision.
  • “There is a competition and contestation between ultra-optimizing agents and systems that want to promote pluralistic cooperation.”




What is the most exotic representation of social relationships that you’ve seen?

  • Glen: I really like Max Flow. It’s a computer science concept of mapping a fluid system of resources flowing through pipes that can be used to reason about social connections.




Seminar summary by James Risberg.