Better DAO Decisions & Aligned Incentives: Research on Carroll Mechanisms

Proposal Type: Governance

Summary

Governance is critical for Scroll's success. Good governance drives growth and resource allocation, while poor governance leads to ecosystem collapse despite large treasuries. Currently, DAO governance suffers from two critical challenges:

• Misaligned incentives: DAOs suffer from voter apathy and farming of incentives (GPT delegates). Current manual approaches to rewarding delegates to do better work are biased and conflict-prone.
• Collective decision making is hard: DAOs suffer from haphazard and biased decision making, and frequent back-channelling limits collective learning and discussion.

Dr. Philip Brown is a recognised leader in mechanism design, having received extensive funding from NSF, NASA, and the Air Force and multiple accolades in this field. The Network Goods Institute, in collaboration with him and RnDAO, we have identified “Carroll Mechanisms” — an evolution of Futarchy and Prediction Markets — as a potential solution for governance systems where participants are incentivised to:

• share information honestly,
• evaluate proposals on merit rather than politics,
• and collaboratively discover solutions without relying on pre-defined choices.

The Scroll Foundation also recognises this direction as promising.

Scroll has an opportunity to implement this cutting edge governance mechanism to improve its performance as a DAO, reward delegates fairly, and build the L2 chain where DAOs go to thrive.

This proposal takes Scroll to the forefront of governance innovation by fast-tracking R&D in Carroll Mechanisms from theory into implementation. Via a 6-month research and implementation program from August 2025 through July 2026. The outcome will be better decision-making in Scroll and fair rewards for delegates.

The program is divided into three milestones. M2 and M3 are conditional on successful M1:

1. Base research on mechanism feasibility ($45k),
2. Applied research to identify the ideal mechanism to implement and calibrate the parameters ($45k),
3. Integration into the Negation Game product so Scroll can immediately benefit from this mechanism to upgrade its governance and delegate rewards system ($55k).

Rather than a traditional grant, we propose an investment structure with shared upside: a $145,000 USD-equivalent SAFE+Token Warrant investment at $10M cap that aligns Scroll's interests with the project's success through potential ROI and ecosystem growth.

Motivation

Problem:

In Web3, we've seen our share of governance failures. Multiple ecosystems have collapsed despite starting with massive treasuries. Too many died not because of competition but through grifting and self-sabotage.

Good governance leads to good decisions and growth. Bad governance leads to an ecosystem of misallocated resources.

Current DAO governance models suffer from multiple challenges to sustain participation, align incentives, reward governance work, and otherwise make good decisions. For DAOs to survive, improving upon the current delegates' models is an existential necessity. Challenges include:

• Conflicts of interest between management and shareholders
• Executive compensation misalignment with company performance
• Groupthink, confirmation bias, authority bias, HIPPO, etc.
• Limited access to expertise leading to risk management and oversight deficiencies
• etc.

Designing incentives for good governance is hard. Finding a better model could unlock massive value for all humans.

The current practice in DAOs is to appoint delegates who then vote on behalf of token holders. However, DAO delegate models suffer from their own set of issues and multiple problems of traditional board governance:

• delegates themselves are seldom kept accountable by token holders,
• can be apathetic (limited engagement with governance),
• and often show behaviours (e.g. GPT replies to proposals, always voting in favour to avoid relationship strain, etc.) coming from incentive misalignment.

Current approaches:

Manual approaches to incentivise governance work (i.e. reward delegates) are prone to biases and conflicts of interest as those who determine the incentives are beholden to the delegates they evaluate. Manual allocation of rewards is also prone to conflict due to the inherent subjectivity in the approach.

Rethinking the solution space from the ground up

Voting systems are incredibly popular governance models, whether taking the form of one-person-one-vote, token/share voting, or representative democratic models. However, voting systems have well-known shortcomings, both theoretically (cf. Arrow’s impossibility theoremˇ) as well as practically. Among the most salient shortcomings in the context of delegate-based DAO governance are the problems of:

• incentive misaligned factions (including all forms of bribery and corruption),

• whales (large power holders that strangle promising investments),

• and problems of low voter participation, capacity, and knowledge.

There are alternatives to voting-based governance systems, among them is futarchy. Futarchy is a governance approach that uses prediction markets—such as those found on Polymarket and Metaculus—to guide DAO decision-making based on forecasted metric outcomes. This has been implemented in straightforward cases like Uniswap, where token price provides a clear and mutually accepted success metric. Futarchy offers a more robust and credibly neutral solution, as it incentivises those with valuable information to engage in governance, while aligning incentives towards identifying the correct answer and long-term accrual of value. However, futarchy has severe limitations:

• Constrains deliberation to predetermined options, preventing groups from discovering novel solutions through collaborative reasoning. The system's effectiveness depends entirely on how well the initial questions and choices are framed, making it vulnerable to poor problem formulation.
• Selecting appropriate resolution metrics proves challenging, as chosen measures may inadequately capture true success, become targets for gaming, or suffer from Goodhart's Law effects where optimization distorts the underlying objective. Many meaningful outcomes are also difficult to quantify or emerge only over extended timeframes.
• Obscures the reasoning behind decisions, revealing only traders' final positions without the underlying rationale. This opacity limits knowledge transfer, reduces organizational learning, breeds distrust, and prevents the alignment-building that comes from understanding why decisions were made.

What if we could rethink the use of voting mechanisms altogether? Could we move away from simply voting on predefined options and invent a game that allows participants to deliberate and co-create new solutions?

Solution:

A promising solution is a set of mechanisms we call “Carroll Mechanisms”. Carroll Mechanisms improve upon the limitations of futarchy by helping define the resolution criteria for a market — not just determining what the outcome of a question is, but also how we determine that outcome.

For example:

• Instead of simply asking “Who will win the next presidential election?”, a Carroll Mechanism clarifies what counts as a legitimate source for declaring the winner (e.g. certified results by a national election commission).
• Or for a more complex question like “Does vaccine X cause neurological disorders?”, it helps define what we mean by “neurological disorders,” which studies or data are valid, what thresholds of causality apply, and how uncertainty or conflicting evidence should be handled.

Carroll Mechanisms work by building on prediction markets, which already let people bet on whether statements are true or false (like "Who will become president in 2026?"). But Carroll Mechanisms add something new: they let participants challenge whether a piece of information is actually relevant to answering the original question.

The key innovation is that these mechanisms can automatically figure out what information should count as relevant. Instead of just asking "Is this true?" Carroll Mechanisms also ask "Does this even matter?" — and automatically figure out the answer.

Carroll Mechanisms do this through a built-in system that identifies which participants are genuinely neutral and unbiased, as demonstrated by a history of changing their minds, then uses their input to determine what's actually useful for resolving the question. For more about Carroll Mechanisms, read this post.

This approach could enable governance participants to engage more collaboratively and productively, and be rewarded accordingly. Carroll Mechanism could hold the missing key to solving incentiv

... please visit link below to view full proposal

https://gov.scroll.io/proposals/55373797422300173363986777521147073392704200977193756403817105312614658692283