Utopian governance
Hedonistic mechanism design
2020-05-04 — 2026-04-16
Wherein the mathematical impossibility of any fair voting system is noted, and alternatives including sortition, liquid democracy, and futarchy are surveyed as partial remedies for governance failure.
TODO: rewrite this as a flagrant technologist. Do some technohabermas. Talk about democracy as a UX problem, an incentive mechanism problem and talk about how we have better solutions now, involving more sophisticated online mechanisms and generative AI which can potentially smooth over details.
In the real world, the laws and policies we have are, in the best case, good-adjacent policies. My mental model is that the best we can hope for, the best attainable policy is the projection onto the manifold of expertise-informed policies subject to the constraints of pressure from vested interests allowed at the time we chose that policy and available attention. Ever since that high-tide mark, that inadequate best effort has become increasingly obsolete, fitting our needs today ever less and less as time goes on until eventually, maybe, it becomes intolerable enough that we revisit the policy, updating it with a revised set of constraints upon expertise, vested interests and attention that may or may not improve things. Then we repeat, unless it is so bad something breaks. Politics is the art of enacting rules that are just good enough that the system does not collapse. Some phrase this as Politics is the art of the possible, although FWIW it seems to me to be closer to Politics is the art of the contingently tolerable. Is that sufficient to govern modern societies?
If we imagine this on the smaller, more directly-actionable scale, we might consider community governance.
Let us set aside the constraints of power and practicality for a second, and ignore also all those complexities that technologists like to ignore and speculate, luxuriously, about what a better system for governing ourselves would look like if we simply got to choose it. Let us imagine we are founding fathers/mothers/etc of a new state, or the delegates to a constitutional convention, and we get to give the constitution a do-over based on the observations, experience and academic research of the last century or two. How would we do stuff differently? How could we better trade off the limited resources of attention, understanding and power to create and maintain a system that does better? What is the kindest system we can execute on our moral wetware?
Here I mean utopian in a particular way; not the one where I propose a system that is the best system, and my system gives us minute details of how we should each live our lives in my system, and I can promise that each of us would be happy in that system. If nothing else I am sceptical of such systems because many of the things that I would need to be happy are not the things that you would need. My ideas are not better than everyone else’s ideas. A better system would spend as little time as possible micromanaging the details based on any one person’s flawed and incomplete ideas of what to do, and let us choose our own details as best we can, working out our own local conventions and trade-offs and so on. Of course, we instantly run into the classic trade-off of liberties where good for one might harm another. And imperfect information — which of the ideas about what is good are most correct? And inequities in power — how does good even happen if it inconveniences someone with a great deal of clout. That is what we want a system to mediate. Of course, I’m not claiming any such systems are perfect, or that they are not hard. I am, however, arguing that just because they are hard does not mean we can’t do better than the current system, which with high likelihood will kill a great many people in a slow grinding ecological apocalypse and crush much human flourishing.
Better systems then. If a better system does not guarantee each of us our happiness, perhaps it would grant us our best chance for happiness. Rather than assuming I, the system designer, know how the world works, how can we, the collective system designers, leverage the greatest possible amount of collective wisdom, while causing the least conflict and friction? What systems have the least waste (in terms of observing belated good-adjacent policies that are ideal for few) and the greatest potential (in terms of discovering, and continuing to adapt, good policies for the most)? What systems are ones that the greatest number will want to live in because they offer the best chance of flourishing?
I do not have an answer here. But this notebook will collate some ideas and tools that spring out of recent history, especially about the areas in which we have made vast progress since the beginning of the modern era, and which legacy institutions do not make good use of. We know more now than we did when they drafted constitutions for almost all modern states. Indeed, we know more now than we did when even fresh and modern states like Timor Leste started. Science is moving fast. We know more about economic psychology — that humans are predictably irrational (Kahneman, Tversky), which means systems that assume rational actors will fail in predictable ways, and systems that account for cognitive biases can do better; algorithmic game theory — that we can now compute mechanism properties, not just prove theorems about them, making it possible to design and simulate institutions before we build them; market design — that markets can be engineered for specific purposes (matching markets for kidneys, spectrum auctions, school choice) rather than treated as a force of nature; common property regimes — that commons can be governed sustainably without privatisation or central control (Ostrom’s Nobel-winning work), if the governance rules are right; distributed algorithms — that consensus, coordination, and fault-tolerance can be achieved in systems with no central authority, and we understand the trade-offs (CAP theorem, Byzantine fault tolerance); collective knowledge — that crowds can be smarter than experts under the right conditions (Surowiecki), but also dumber under the wrong ones (groupthink, information cascades); collective action — that the barriers to cooperation are not just free-riding but also assurance, trust, and the cost of coordination itself (Olson, Ostrom again); community building — that small-group dynamics, psychological safety, and cultural norms are not soft variables but structural determinants of whether institutions work (see community governance); attention — that it is the binding constraint on democratic participation, and any system that demands too much of it will be captured by those with the most to gain; persuasion — that we now understand far more about how beliefs form and change, for better (deliberation design) and for worse (misinformation, dark patterns); voting systems — that every one of them is formally broken in specific, characterisable ways, and we can now choose our breakages deliberately; envy-free mechanism design — that fair division is a solvable engineering problem, not just a philosophical aspiration; reputation systems — that trust can be computed, delegated, and aggregated in ways that go far beyond personal acquaintance; risk management — that tail risks can be modelled and hedged, and that institutional failures are often failures to take known risks seriously; finance and portfolio theory — that diversification, option pricing, and insurance are general coordination tools, not just Wall Street furniture; forecasting — that calibrated prediction is a trainable skill and that prediction markets can aggregate beliefs more accurately than committees; statistics — that we can measure what works, with increasing rigour, through randomised trials, causal inference, and meta-analysis.
Oh, and we have better and faster communications than ever before. Not only do we have many advances in theory. We also have the biggest, most expensive laboratory of social behaviour and governance that the world has ever seen: the internet. That internet which could be the petri dish where we study exactly the things we need to know about ourselves. It could be enabling us to develop, understand and trial the mechanisms that might make the world better. In practice it is largely being used to persuade us to hate movies and enable authoritarianism, though, so that one is a mixed bag.
0.1 Why current systems provably fail
It’s not just that governance is hard in practice. Social choice theory tells us it’s provably limited. Arrow’s impossibility theorem shows that no voting system with three or more options can simultaneously satisfy a handful of seemingly reasonable fairness conditions. As Alex Tabarrok puts it, “the only leviathan that rationalises group choice has the preferences of a madman.” The Gibbard-Satterthwaite theorem adds that any non-trivial voting system is vulnerable to strategic manipulation — people will game it, and no design can prevent that in general. These aren’t practical difficulties to be engineered away. They’re mathematical constraints on what preference aggregation can do. Every voting system is broken; the question is which breakages we can live with, and whether there are mechanisms beyond voting that can do better.
I do not have any ideas about which utilitarian calculus is the right one to use here. I will table the ethics questions for later. I do not see this as a huge problem. Not wiping off 100 billion quality-adjusted life years from the face of the planet is bad in almost any moral system and that’s a great starting baseline.
And beyond the mathematics, there is the practical coordination problem. Robin Hanson gives a useful disambiguation:
Two ideological attitudes are common, but insensibly stupid:
- All government activity is bad, no matter what it does.
- The only reason to oppose a government program with a purported goal is because that goal is bad; program opponents must oppose its goal.
The key thing to understand is: governance is hard, especially in a democracy. Fundamentally, this is because coordination is hard.
It can be very hard for even a single owner to coordinate with a dozen subordinates that each coordinate with a dozen employees in an ordinary firm to achieve a simple clear goal like making and selling a simple product at a profit. Organizations fail at this task all the time, and for thousands of different reasons. Most new organizations attempting this fail, and most that are succeeding now will fail in a few decades. When they fail, they will fail so badly that it will not be worth trying to save them; better to throw them away and start anew.
Once one appreciates the difficulty of coordinating even small organizations, and that bigger coordination is harder, one can see why it can be extremely difficult to manage the vaster coordination required by government. How can ordinary citizens continue over centuries to coordinate to support interest groups that coordinate to support politicians who coordinate to approve and manage policies that empower agency heads to coordinate to manage thousands of agency employees to achieve the vague incoherent goals of many millions of citizens?
The Effective Institutions Project takes this seriously as a research programme:
The Effective Institutions Project is a new global working group dedicated to building a cross-disciplinary community of interest around the challenge of improving institutional decision-making (IIDM). We aim to increase the technical quality and altruistic intent of important decisions made by powerful institutions. We leverage the accumulated wisdom of researchers and practitioners across many fields in order to identify opportunities to make change that are likely to have the greatest positive impact on society.
Actions taken by powerful institutions—such as central governments, multinational corporations, influential media outlets, and wealthy philanthropic funders—shape our lives in myriad and often hard-to-perceive ways. Yet on issues from COVID to climate change, our leaders too often misjudge risks, make choices based on political expediency, and fail to imagine wiser alternatives when it counts the most.
It’s no surprise, then, that according to an analysis by 80,000 Hours, “improving the quality of decision-making in important institutions could improve our ability to solve almost all other problems.” We need our institutions to do better. Our lives may literally depend on it.
0.2 The design space
So: we have formal impossibility results telling us preference aggregation is inherently limited, practical coordination problems at every scale, and institutional inertia that selects for the contingently tolerable over the genuinely good. Given all that, the mechanisms in the rest of this notebook aren’t a shopping list of alternatives to pick from. They address different failure modes of governance, and the interesting question — one this notebook is circling towards — is how they compose.
The problems, roughly:
Representation and capture. Who gets to decide, and how do we prevent the deciders from entrenching their own power? Elections select for people who want power; incumbents shape rules to stay in power. → Sortition addresses this by making selection provably random.
Attention and expertise. Citizens can’t be informed about everything. Delegation to elected representatives or experts creates principal-agent problems; delegation to parties creates ideological bundling where you must buy the whole package. → Liquid democracy tackles this through revocable, per-issue delegation.
Belief aggregation. Which policies will actually work? Voting tells us what people want but not what will happen. We need mechanisms for discovering ground truth about policy consequences — this is governance as an epistemic problem. → Futarchy and prediction markets attack this by putting skin in the game — though the causal validity problem bites. See also the epistemic framing in the civic technology notebook.
Incentive alignment. How do we make it individually rational to pursue collective goods? Standard mechanism design gives us tools here. → Social impact bonds and attribution-based economics make it profitable to do good, or at least to be compensated for having done good.
Small-group consent. People who must work together need a way to make decisions that everyone can live with, without majority tyranny. → Sociocracy provides consent-based governance for groups with shared goals.
Equity and access. How do we prevent the powerful from opting out of shared systems? How do we structure property and markets to avoid runaway concentration? → Radical markets, commons governance, and the brute-force approach of simply making everyone use the same public systems.
These problems interact — attention constraints make capture easier; bad belief aggregation makes incentive alignment harder — and so the mechanisms need to compose. The Anuna case study below shows what that composition can look like in practice. For technology-enabled approaches to many of these problems — civic platforms, AI-mediated deliberation, blockchain governance — see civic technology.
1 Sortition
Sortition — government by random sampling of representatives from the population. What statistician could avoid at least toying with this idea? David Chaum makes the statistical argument that random-sample elections are “far lower cost, better quality, and more democratic” — see social choice for more on the formal case.
David Van Reybrouck’s Against Elections (2014) makes the case at book length. The Assembly Guide provides practical guidance for anyone wanting to try it:
The Assembling an Assembly Guide is a resource for any institution, organisation, city administration, or policy maker interested in running a Citizens’ Assembly. It is also a useful tool for citizens and activists wishing to learn more about what a Citizens’ Assembly is and how it works, in order to strengthen their advocacy efforts.
Hugo Sterin and the Anuna Research Cooperative are attempting cryptographic sortition at organisational scale: their board is selected by a verifiable random function, not elected — 6-month rotation, no campaigns. The aim is to structurally prevent capture: no one can position themselves for power when the selection is provably random. See below for more on the Anuna model.
TODO: how does sortition interact with liquid democracy? With futarchy? With the attention/expertise trade-off discussed above?
3 Futarchy
Robin Hanson’s futarchy probably generalises social impact bonds. It’s also a direct response to the impossibility results above: if voting can’t reliably aggregate preferences, maybe we should vote only on values (what do we want?) and use markets to aggregate beliefs (what will get us there?).
This short “manifesto” describes a new form of government. In futarchy we would vote on values, but bet on beliefs. Elected representatives would formally define and manage an after-the-fact measurement of national welfare, while market speculators would say which policies they expect to raise national welfare.
Note that the causal validity problem may bite us. See also the epistemic framing in the civic technology notebook — the distinction between preference aggregation and belief aggregation is central to why futarchy might work where voting doesn’t.
4 Liquid democracy
Liquid democracy is a hybrid of direct and representative democracy: voters can either vote directly on issues or delegate their vote to a trusted representative, and can revoke that delegation at any time. It addresses the attention problem head-on — we don’t need to have an opinion on everything, but we’re never locked into someone else’s opinion either.
TODO: connect to the attention/expertise trade-off in the essay above. How does this interact with sortition? With futarchy?
5 Sociocracy
Sociocracy is a system of governance that seeks to create psychologically safe environments and productive organizations. It draws on the use of consent, rather than majority voting, in discussion and decision-making by people who have a shared goal or work process
A consent-based rather than majority-vote-based system — decisions stand unless someone raises a reasoned objection. The Anuna Research Cooperative uses sociocratic circles (Portfolio, Operations, Partnerships) with consent-based decision-making and rotating roles, nested under a sortition-selected board. See below.
See also community governance for how these ideas play out at smaller scales.
6 Equity and inclusion
One of the most direct ways to improve governance is simply to ensure the powerful can’t opt out of public systems.
Hamilton Nolan, Everyone Into The Grinder:
One of the most direct ways to improve a flawed system is simply to end the ability of rich and powerful people to exclude themselves from it. If, for example, you outlawed private schools, the public schools would get better. They would get better not because every child deserves to have a quality education, but rather because it would be the only way for rich and powerful people to ensure that their children were going to good schools. The theory of “a rising tide lifts all boats” does not work when you allow the people with the most influence to buy their way out of the water. It would be nice if we fixed broken systems simply because they are broken. In practice, governments are generally happy to ignore broken things if they do not affect people with enough power to make the government listen. So the more people that we push into public systems, the better.
See also the Review of Joseph Heath, Enlightenment 2.0 on Habermas + Kahneman ideas in the public sphere.
7 Radical markets and mechanism design
The RadicalxChange movement explores mechanisms like quadratic voting and common ownership self-assessed tax (COST). See mechanism design and social choice for more on the theoretical underpinnings.
This economist wants to abolish private property using blockchain:
Weyl proposes to transform society into a continual auctioning process: people should set an ideal price for each of their possessions—houses, cars, pieces of clothes—and be ready to sell it to whoever comes forward bidding for it. Obviously, people keen to latch onto their stuff would set a very high price to keep bidders at bay. Here’s the rub: in Posner and Weyl’s blueprint, every person would pay a hefty tax on their overall wealth, which would of course be higher the higher the self-assessed value of their chattels. In other words, property will either be auctioned and decentralised across the society, or generate a high tax revenue, which could go on to fund a Universal Basic Income for the have-nots
8 Attribution-based economics
Social impact bonds and retroactive public goods funding (above) solve the problem of funding public goods. But there’s a related problem: how do we fairly compensate contributors to a commons when value is diffuse and delayed?
Hugo Sterin and the Anuna Research Cooperative propose Dialectical Inheritance Attribution (Kasivajhula): every contribution is a traceable identity in a provenance graph; value is revealed over time as downstream work inherits from upstream work; attribution is appraised collectively by all constituents; and revenue flows backward through the inheritance tree. This solves both the horizon problem (why invest in long-term work?) and the free rider problem (why invest at all?) — it’s compensation for traced work, not profit distribution.
TODO: connect to the literature on open-source sustainability, Elinor Ostrom’s commons governance, and the retroactive funding mechanisms above. How does this compare to certificates of impact?
9 Case study: Anuna Research Cooperative
Anuna Research is a research cooperative for free software and open science, founded by Hugo Sterin. It’s interesting here not because of what it builds but because of how it governs itself — it composes several of the mechanisms discussed above into a single working institutional design.
Their core thesis is that the institutional form is upstream of the innovation: “Only variety can absorb variety” (Ashby). Large organisations optimise for certainty; government labs game metrics; companies underinvest in long-term bets; philanthropy follows donor cycles. Siloed institutions select for more of the same. A cooperative of cross-disciplinary researchers, peer-governed and anti-fragile by design, has more response repertoire — and that variety is what lets it handle complex problems.
The governance model combines:
- Sortition board — directors selected by cryptographic sortition, 6-month rotation, preventing capture.
- Sociocratic circles — consent-based, open participation, rotating roles.
- Attribution-based economics — Dialectical Inheritance Attribution for internal revenue distribution.
- No external shareholders, no extraction pressure — the people who do the work own the work and set the direction.
- Self-learning organism — governance, coordination, and research infrastructure are not separate from the R&D; they are the R&D. The institution studies its own processes through the tools it builds.
Their commercial interface preserves the commons: output is open; commercial value comes from execution, integration, and trust — closer to Red Hat or Linux Foundation membership than a licensing model.
TODO: follow up on how this evolves in practice. The open questions they flag — transition from informal to formal governance, deferred compensation models, sustainable platform economics — are exactly the hard problems of institutional design this notebook cares about.
10 Incoming
Anuna Research — see case study above. Remaining threads to follow: their SPINdle defeasible-logic engine (policy-as-code for governance rules?), the transition from informal to formal governance at small scale, and how attribution-based economics plays out in practice.
The disastrous voyage of Satoshi, the world’s first cryptocurrency cruise ship
Sam Bowman, A neoliberal agenda for the 2020s
Zvi Mowshowitz, Announcing Balsa Research
Julian Agyeman and Duncan McLaren, Sharing Cities:
McLaren and Agyeman propose a new “sharing paradigm,” which goes beyond the faddish “sharing economy”—seen in such ventures as Uber and TaskRabbit—to envision models of sharing that are not always commercial but also communal, encouraging trust and collaboration. Detailed case studies of San Francisco, Seoul, Copenhagen, Medellín, Amsterdam, and Bengaluru (formerly Bangalore) contextualize the authors’ discussions of collaborative consumption and production; the shared public realm, both physical and virtual; the design of sharing to enhance equity and justice; and the prospects for scaling up the sharing paradigm through city governance. They show how sharing could shift values and norms, enable civic engagement and political activism, and rebuild a shared urban commons. Their case for sharing and solidarity offers a powerful alternative for urban futures to conventional “race-to-the-bottom” narratives of competition, enclosure, and division.
Now, how did that one pan out?
Ask a Neoliberal: An Interview with J. Bradford DeLong - Dissent Magazine
Participedia — theory, methods, and case studies for participatory democracy. Also relevant to civic tech.
Plurality: The Future of Collaborative Technology and Democracy — also relevant to civic tech.
2 Social impact bonds
Ronnie Horesh floats an interesting idea: social policy bonds
Scott Alexander observes Lewis Carroll Invented Retroactive Public Goods Funding In 1894: