Courses / Decentralized Systems / Week 9

Week 9 of 12 · Intermediate

Mechanism Design And Plural Governance

You learn how to design rules so that self-interested people produce good group outcomes. You study Sybil resistance, quadratic voting and funding, conviction voting, futarchy, and cosmo-local ideas.

3 hours18 slides6 min read

Open slides Read study notes

Mechanism Design In Plain Words Incentives And Sybil Resistance Quadratic Voting Quadratic Funding Conviction Voting Futarchy Plural Ideas And Cosmo-Local Design

You Will Be Able To

Explain mechanism design as designing rules for a goal.
Describe a Sybil attack and three ways to resist it.
Compute quadratic voting costs and quadratic funding matches.
Compare conviction voting and futarchy with simple token voting.
Apply cosmo-local thinking to a local community project.

Mechanism Design In Plain Words

Last week you saw how DAOs fail: plutocracy, low turnout, and attacks on the rules. This week asks a constructive question. If people follow their own interests, can we shape the rules so that those interests add up to something good?

That question is the subject of mechanism design, sometimes called reverse game theory. Ordinary game theory takes the rules as given and predicts behavior. Mechanism design starts from the outcome we want and works backward to find rules that make the wanted behavior the sensible choice for each person. The economists Leonid Hurwicz, Eric Maskin, and Roger Myerson received the Nobel Prize in 2007 for this field.

A decentralized system needs mechanism design more than most, because no manager can step in to repair bad incentives. If a rule can be gamed, someone will game it, and the code will let them. Designers therefore ask three questions about every rule. What behavior does it reward? Who can fake that behavior cheaply? What does an attacker gain?

Incentives And Sybil Resistance

Incentives are the rewards and costs that push people toward actions. Good incentives line up private benefit with group benefit. A staking system, for example, asks validators to lock tokens that can be taken away if they cheat. Honest work earns rewards and dishonest work costs money.

The hardest problem for any one person one vote system is the Sybil attack, named after a book about a woman with many personalities. An attacker creates many fake identities to gain extra votes or extra rewards. On a public chain, making a new address costs almost nothing, so counting addresses tells you nothing about counting people.

There are several defenses, and each has a price. Cost-based defenses make each identity expensive, as proof of work and proof of stake do. Identity-based defenses check that an account belongs to a unique human, using government ID, biometric devices, or video verification, though these raise privacy concerns that you studied in week 7. Social defenses use web of trust: existing members vouch for new ones, as in systems like BrightID and Proof of Humanity. Another approach is the proof of personhood party, where people meet in person and verify each other. Most real systems combine two or more defenses.

Cost: make every identity expensive.
Identity: check that an account maps to one human.
Social graph: members vouch for each other.
In person: verify people at gatherings.

Quadratic Voting

Simple majority voting treats all preferences as equal in strength. A person who cares a little and a person who cares a lot get one vote each. Quadratic voting, proposed by Glen Weyl and Steven Lalley and developed with Eric Posner, lets people show how strongly they care.

Each voter receives a budget of voice credits. Casting n votes on a single issue costs n squared credits. One vote costs 1, two votes cost 4, three cost 9, and ten cost 100. Because the cost climbs quickly, a voter spreads credits across the issues they care about and puts the most votes where they care most.

The mechanism has a clear goal: a voter with intense feelings can outweigh many mildly interested voters, but only at an increasing price. It works well only when each person really is one person. Without Sybil resistance, an attacker splits their credits across many fake accounts and pays far less, because many accounts each casting one vote are cheaper than one account casting many. Quadratic voting and identity checks therefore travel together. Collusion is a second risk, and week 7 showed how MACI reduces the ability to prove how you voted.

Quadratic Voting: Cost Of Votes

credits

Cost is the square of the votes, so strong feelings are possible but expensive. Source: Lalley and Weyl, Quadratic Voting, 2018

Quadratic Funding

Quadratic funding applies the same idea to public goods, which are things many people benefit from but nobody pays for alone, like open-source software. It comes from a 2018 paper by Vitalik Buterin, Zoe Hitzig, and Glen Weyl called Liberal Radicalism.

Donors give small amounts to projects. A matching pool then tops up each project. The matching formula rewards the number of distinct contributors more than the size of donations. A project's total is the square of the sum of the square roots of each donation. For two projects each raising 100, one from a single donor of 100 and one from 100 donors giving 1 each, the second receives far more matching, since it shows broad support.

Gitcoin Grants has used quadratic funding to distribute millions of dollars to open-source and public goods projects over many rounds. The weakness is the same as before: fake accounts can imitate broad support, so each round needs Sybil defenses and review.

Many small donors beat one large donor.
The matching pool amplifies broad support.
Sybil resistance protects the match.

Quadratic Funding Example

Project	Donors	Each gives	Matched total
A	4	$1	$16
B	1	$4	$4

Both raise $4 from donors. Matching is the square of the sum of square roots, so breadth beats one big gift. Source: Buterin, Hitzig, Weyl, Liberal Radicalism, 2018

Quadratic Funding Round

Many people donate small amounts
Take the square root of each gift
Add them up and square the sum
Pool tops up the gap
Projects receive donations plus match

The match is high when many different people give.

Conviction Voting

Standard voting happens in bursts: one vote, one deadline. Conviction voting works continuously. A member stakes tokens on a proposal, and the weight of their support grows the longer the stake stays in place. When the accumulated conviction crosses a threshold, the proposal passes and funds are released.

This design has three effects. Impatient or last-minute whales gain little, because conviction takes time to build. Members can change their minds by moving their stake, which resets that part of the conviction. Proposals do not compete in a single dramatic vote, so a community can fund many small things as support accumulates. The 1Hive community and Commons Stack projects have used it for small grants. Its weakness is slowness for urgent decisions, so many groups pair it with a fast track for emergencies.

Futarchy

Futarchy is a governance idea from economist Robin Hanson. Its motto is 'vote on values, bet on beliefs'. The community votes once to define a measurable goal, such as the price of a token or a health index. Then prediction markets decide which policy is best. Traders bet on what the goal metric would be if proposal A passes, and what it would be if A fails. The policy with the better predicted outcome is adopted.

The strength of futarchy is that it rewards accurate information. People with knowledge profit by trading, and wrong guesses cost money. The difficulties are practical. Markets need enough liquidity to resist manipulation. The goal metric must be hard to game. Ordinary members must trust a system that feels abstract. Experiments such as MetaDAO on Solana have tried small versions on real treasuries, and the results are early. Treat futarchy as a promising research direction with open questions, not a settled answer.

Plural Ideas And Cosmo-Local Design

Plural governance asks for systems that respect many groups and many kinds of knowledge instead of reducing everyone to one number. Glen Weyl and Audrey Tang describe this direction in the book Plurality. Tools such as Pol.is cluster opinions so that a community sees where groups agree, including across divides, instead of only who won. Taiwan's vTaiwan process used Pol.is to shape national policy discussions.

Cosmo-local production is an idea from the P2P Foundation and writer Michel Bauwens, summarized as 'light global, heavy local'. Designs, software, and knowledge are shared globally as open commons, while physical production and care stay close to home, using local materials and local skills. A village workshop can build a water filter from an open design published thousands of kilometers away.

For decentralized builders, the lesson is to match the mechanism to the scale. Use global tools for sharing knowledge and verification, and keep decisions about shared local resources among the people who live with the results.

Key Terms

Mechanism design: The craft of designing rules so that self-interested behavior produces a wanted group outcome.
Sybil attack: Creating many fake identities to gain extra influence or rewards.
Proof of personhood: A way to show that an account belongs to one unique living human.
Quadratic voting: A rule where casting n votes on one issue costs n squared credits.
Quadratic funding: A matching method that rewards the number of contributors to a public good more than the amount given.
Conviction voting: Continuous voting where support grows with the time tokens stay staked on a proposal.
Futarchy: Governance where prediction markets choose policies to reach a goal set by a vote.
Public good: A resource that many people can use and that is hard to charge for.
Cosmo-local production: Sharing designs globally as commons while building and caring locally.

Exercise: Compute Quadratic Costs And Matching In Python

Open a Python file and write a function vote_cost(n) that returns n squared.
Print the costs of 1 to 10 votes and notice how they grow. Write one sentence on what this does to a voter with 100 credits.
Write a function qf_total(donations) that returns the square of the sum of the square roots of the donations.
Compare qf_total([100]) with qf_total([1] * 100). Compute the matching amount for each as total minus raw sum.
Add a Sybil attacker: split one donation of 100 into 100 donations of 1 and show the result. Write two sentences on why identity checks are needed.
Share the output with a partner and agree on one rule to limit the attack, such as a cap per round.