Fixing DAO governance

For the past decade, I have helped govern self-owned communities, researched organization design, designed tech solutions for governance, and advised numerous ecosystems on their path to decentralization. But after all this time, I feel I have largely failed to have any meaningful impact in stopping the train wreck that is DAO governance today. 

Still, these challenges have gradually helped me get to a coherent and detailed vision of what a solution can look like for effective governance in decentralized ecosystems. This post is an attempt to share this solution.

Framing The Challenge

Memetics lead to copying poor practices

For anyone who’s seriously looked into the topic of decentralized governance, you know this is NOT an easy challenge. Fixing governance won’t be quick. For starters, organizations copy “best practices” not based on good governance but based on the economic success of the business model. For example, the Compound governance model has been copied a lot but, governance-wise, Compound is a dumpster on fire.

Governance is multi-stakeholder

Importantly, governance lies at the intersection of all stakeholders and their interests. Governance is by definition the system that different stakeholders (i.e. interest groups) will use to negotiate and ideally collaborate with each other. 

Although the exact mix of stakeholders varies in each organization and can evolve, the most common stakeholder groups are: 

• investors: investment capital providers 
• contributors: labor providers 
• customers: revenue providers 

In most cases, all of these groups are not only important but critical. When stakeholders are happy, they’ll attract others and create a flywheel for the organization to grow.

However, when governance is poorly designed, one stakeholder exploits the others, and even worse, they all suffer together. Governance failure leads to conflict, disillusionment, and ultimately the failure of the organization as the interest groups disengage and deploy their resources somewhere else.

Different forms of governance are based on different arrangements between these groups:

• Corporate governance gives power to Investors, assuming that when investors act as rational economic actors they provide valuable products and hire the right contributors at competitive market rates. This model works relatively well but is also prone to degenerate into exploitation when an organization becomes successful and the business shifts from value creation to over-optimizing for extraction by shareholders. And extraction ultimately leaves the business vulnerable to being disrupted by competitors. In another common mode of failure, corporate governance is captured by senior management. Executives accumulate higher and higher pay, sometimes receiving golden parachutes even when shareholders lose millions.
• Cooperatives usually give power to Contributors. This results in relatively good arrangements for workers but the model is not optimized for investors, which means most Cooperatives struggle to access financing to grow faster and hence why most tech startups are built as limited companies and not cooperatives.
• Other cooperatives are run by Customers. These are in essence buyer associations and are occasionally successful at representing the interest of Customers but are also prone to be captured by senior management, suffer from the tyranny of the minority issues (capture by overly vocal members), and can make members feel a loss of agency where they have to lobby instead of being able to directly purchase. Additionally, little innovation tends to come from customer cooperatives as it is Investors providing risk-oriented capital and entrepreneurs providing risk-oriented labor who innovate.

So my basic theory for governance is that we need governance that balances the interests of these groups and allows for win-win solutions instead of a single stakeholder group being in full control i.e. we want to avoid capture by any single actor. And even beyond that, we want governance that enables the different stakeholders to collaborate, that is, to discover win-win solutions and execute them effectively. Good governance is not just about dividing the pie fairly but about discovering how to grow the pie together.

Importantly, in DAOs and occasionally in other organizations, the different stakeholders overlap i.e. the same person or agent could simultaneously be an Investor, Contributor, and/or Customer. This means a good governance design needs to account for these overlaps.

Governance is a complex process—more than just voting on proposals

Through governance, the different stakeholders try to 

1. make sense of their situation,
2. identify and prioritize problems and opportunities,
3. ideate, prioritize, and refine solutions,
4. approve the usage of resources, 
5. execute,
6. and learn (looking back to step 1).

Governance should not be reduced to voting but understood as a process with multiple phases, where the different stakeholders engage. It’s a messy process where going back and forth is normal and multiple phases can happen concurrently and within the limits and complications of dealing with humans. The multiple phases are critical for allowing the different stakeholders to not just split resources between them but discover solutions for providing for each stakeholder’s needs and negotiate between short-term desires and long-term needs.

Across these phases, different stakeholders don’t (need to) engage equally. Customers can signal problems and occasionally offer solutions, but it’s Contributors who have the bandwidth to refine a solution and implement it. Same for Investors, they can offer insights and maybe be part of approval but will lack enough bandwidth to engage throughout and/or context to do it effectively.

Throughput and attention costs are critical considerations

The initial design of DAOs was based on airdropping tokens to everyone (making every stakeholder also an investor) and then expecting them all to participate equally in a direct democracy governance system. This design fails in most cases because most stakeholders have other stuff to do than participate in governance, they will free ride often, and expecting everyone to assess every decision is just unrealistic. The attention costs of direct democracy are too big, which leads stakeholders to disengage and hence leads to poor decisions. And to make things worse, when attention costs are too high, the number of decisions that can be made (decision throughput) becomes very low, which makes the organization slow and ineffective, leading to failure.

Seeing these problems, DAOs tried to correct the issue by implementing delegate systems, effectively creating a system of representative democracy. However, this new design brings its own set of problems. Namely, delegates become a new stakeholder group looking to maximize their interests, and so don’t necessarily act for the benefit of Customers, Investors, or (other) Contributors. High attention costs to pick delegates and hold them accountable lead to entrenched power and favor skilled politicians at the start of the DAO over those providing valuable work over the long term.

Delegates themselves can still free-ride or otherwise ignore the DAO, which has resulted in the creation of multiple delegate incentive programs. Although these programs can compensate for some of the system design failures, they are still prone to abuse and perverse incentives (e.g. adding noise).

Prediction markets and Futarchy often come then as the next logical bandaid, trying to incentivize participation and produce better decisions. However, both of these mechanisms leave the hard problem unsolved: how do you identify what needs to be worked on and how to measure it? In essence, reality is messy and complex, and trying to reduce it to a finance game leads to oversimplification. From a complexity science perspective, trying to suppress variety leads to that variety popping up somewhere else, often in the form of unintended consequences and negative externalities. Concretely, this can manifest in Futarchy systems by introducing all the biases that traders have, such as short-termism, wild swings with volatility, and only large organizations being properly assessed while smaller organizations can struggle to attract enough participants (because the attention costs are still high and trading requires many participants).

Polarization and preconceived ideas

Perhaps the biggest issue with current governance systems is that they are designed to choose between preconceived ideas rather than fostering new and better solutions. Direct and representative democracy tends to polarize people and have them entrenched in fixed positions and complex arrangements of interests in political parties. These systems discourage innovation and discourse.

Towards a solution

To address the dysfunctions mentioned above, we need a solution for governance challenges but importantly, also for organization design to segment the type of decisions to be made with each mechanism. My current view of what’s an ideal design goes as follows below. Different projects will need alternative designs, but this is a widely applicable template: 

• A separation between high-level decision-making (constitutional proposals and big financial decisions), everyday operational decisions, and conflict resolution so we can increase governance throughput and provide checks and balances. In essence, we propose a separation of powers between Policy (legislative), Operations (executive), and Conflict Resolution (judicial).
  
  • Policy (a.k.a. legislative, meta governance, constitutional) based on methods that include all stakeholders and comprise a full governance process (not just voting). The design trade-offs for Policy should favor decision quality over speed.
  • Operations (aka. Executive) optimized for fast decisions and quick iteration, as well as continuity so that the learnings compound and there’s clear accountability.
  • Conflict resolution (a.k.a. Judicial) optimized for legitimacy and transformative justice i.e. the point is not just to dictate who is right and who’s wrong but to lead to real conflict resolution and improvements to the system.
• Additional systems are also needed for oversight, strategic innovation and R&D, and supporting individuals in the system (as per the Viable System Model. (More about it here). I have included these functions within the executive to keep the framework simple.
• A final element is exit: the mechanism through which stakeholders can leave the organization.

In practice, this design can look as follows:

Policy (Legislative) - multi-stakeholder and deliberation based

• A system for collecting pain points and issues: this relies on a relatively simple database where stakeholders can log in concerns and pains and the system clusters them. We have an MVP of this via SimScore.xyz.
• Multi-stakeholder and deliberation-based assemblies where key issues for the Legislative to work on are unpacked (problem definition work) and then solutions are devised in the form of policies, guidelines, and rules. The assemblies include plenaries and information packs where experts with different points of view inform/pitch assembly members their ideas, so the assemblies are expert-informed but not expert-led. Additionally, participants in the assemblies are rewarded for their participation. We have an MVP of AI-assisted, collective sense-making in Harmonica.chat.
• When it comes to ratifying a decision from the assemblies, we use multi-factor voting:  membership to each stakeholder class is quantified with appropriate metrics like usage for Customers, token holding for Investors, and contribution over the past 3-6 months for Contributors. We also add a metric for Context that quantifies how aware someone is of the situation. We have an MVP of a Sybil-resistant and spam-resistant metric for Context via Reputation NFTs by togethercrew.com. Additional improvement to the ratification system can come from: 
  
  • Adding topic-dependent delegation in a liquid democracy system i.e. enabling chains of delegation based on the specific topics of each decision.
  • Rewarding active governance participants -both active participants and delegators who review their delegation periodically - through moderate inflation.
  • Using specific voting strategies based on the type of decision made. This area still needs a lot of research but some examples to illustrate:
    
    • Elections or picking between options: rank choice voting (with a Condorcet algorithm and backing by Borda when the Condorcet algorithm doesn’t produce a winner)
    • Adjusting a parameter (quantitative and discrete variable): vote a number (i.e. picking the median amongst the numbers selected. How this intersects with voting power is yet to be explored).
    • Answering an open-ended question e.g. what should be the DAO’s mission: aggregation of qualitative inputs and selection by a similarity algorithm e.g. using SimScore.
• For the hardcore Governerds, the legislative here is conceived as both a System 5 (Identity) and System 4 (Future/Vision) in the Viable System Model, so functions for continuously gathering data from the outside of the organization and converging into a vision are important. So far my thinking is that the same mechanism(s) described above are fit for both the purpose of generating policy and vision alignment, and as long as both categories of work are understood and neither forgotten (via facilitation in the social layer), the system will be viable. In practice, we might need to separate System 5 and 4 more but current DAO governance practice is used to a single mechanism and so shouldn’t be divided too quickly or risk overwhelming participants. I already made that mistake, introducing too fast a grants council and DAO voting as dual mechanisms in Aragon in 2021.

Operations (Executive) - small teams with entrepreneurial incentives and a few workgroups

• Initiatives and programs are executed by the Executive through the creation of small, focused units with high autonomy and interdependence. Ideally, these teams are entrepreneurial, meaning they manage their own P&L and hiring, and are owned partially by the team contributors, partially by the overall organization, and if needed also partially by investors in the team. The ideal configuration results in the Units buying and selling from each other and from external customers and suppliers, enabling market dynamics to dictate capital allocation and not asking the overall organization to figure out the split of resources. A great case study on this approach is Haier, and generally, our research has led us to converge on the idea of Swarms.
• Accountability & Data: A system for accountability is also set up with the capability for both scheduled and random audits, aggregating and structuring reporting data, and whistleblower program(s). This system aims to bring transparency and orchestrate data, and is complemented by the entrepreneurial design of the operating units in the Executive.

Conflict Resolution (Judicial)

• Handled by a neutral third party, with steps for facilitation, mediation, arbitration, and a direct link to pass learnings to the Legislative to resolve root causes. This system works in combination with the Audit function of the accountability system in the executive.

Exit

• One of the least researched areas. Although including mechanisms for lockups and some form of secondary market seem ideal to enable orderly exits. Ragequit mechanisms where a token holder can exchange their tokens for a share of the treasury are only effective when most of the value is locked in the treasury and so of very limited use in most cases. Advancements around fractionalized IP and onchain IP could transform this landscape but those are still very embryonic.

The Path Forward: Building Resilient DAO Governance 

The proposed three-branch governance model - combining a deliberative legislative system, an entrepreneurial executive structure, and a transformative judicial process - offers a practical framework for addressing the current shortcomings of DAO governance. 

By acknowledging the complexity of multi-stakeholder dynamics and moving beyond simple voting mechanisms, this system can potentially unlock higher decision quality while maintaining operational efficiency. Through carefully designed incentives, separation of powers, and mechanisms for continuous learning and adaptation, DAOs can evolve from their current state of dysfunction into truly effective self-governing communities. The key lies not in copying existing models or implementing quick fixes, but in thoughtfully designing systems that balance stakeholder interests while optimizing for both participation and throughput. While this vision will take time and effort to refine and implement fully, it provides a first-principles approach instead of bandaids for the next generation of ecosystems and organizations.

‍

This article was written by Daniel Ospina, an Instigator at RnDAO with more than 10 years of experience in Organization Design.