Exploring DAOs as a New Kind of Institution
A Dialogue Between the Metagovernance Project and the Commons Stack
DAOs, or decentralized autonomous organizations, are technical tools written in code and run on blockchains, but they are also a new kind of social institution. Early explorations of DAOs focused more on their technical development and less on their social implications, leading to a series of failures that exposed the limitations of a rigid, “code is law” approach to the design of DAOs. In this article, we explore a more holistic approach to DAOs, one that integrates technical engineering with social design. We call this approach the institutional view.
Defining the Terminology
We have the language of cryptoeconomics for talking about the technical side of DAOs, but so far we haven’t found the right words to talk about their social side. In this article, we’d like to introduce four such words: institution, organization, constitution, and computational constitution. (Okay, technically that’s five words, but “constitution” is a repeat!) In doing so, we’ll try to fit the cryptoeconomics you know and love within the larger field of institutional economics, and show you why it’s a good idea to take into account not just technical components such as tokens or code, but also social components like constitutions and social agreements. By holistically engineering DAOs, we can avoid the trap of reductionism and open these new tools up to a world of future applications.
Institutions are defined as stable patterns for regulating human behavior. Lots of things are institutions, including many things that one does not normally think of as institutions. Courts, markets, and contracts are institutions, but so is the design of the aisles in your local supermarket or even a baby’s reflex to cry.¹ Code, too, is a kind of institution, insofar as it regulates human behavior. DAOs are computational institutions — they facilitate governance through a series of smart contracts — and like any institution they should be designed with those being governed in mind. The modern approach to institutional design was pioneered by Elinor Ostrom and others in the context of common-pool resources, but here we’ll focus explicitly on DAOs and other computational institutions. Further, we will distinguish between classes of institutions (e.g. markets, legal associations, DAOs) that follow an institutional pattern, and actual instances of institutions (the Superior Court of New Jersey, the reputation feature on Reddit). When we talk about DAO institutions in this article, we’ll be examining them as an “institutional pattern”, a class of institutional design that can have widely varying instantiations.
From an institutional perspective, DAOs are not just bundles of smart contracts; they are also social organizations, entities composed of individuals gathered for a common purpose.² This implies that the bundle of smart contracts that defines the DAO is not a complete representation of that DAO; the same code, used by different groups of people, can lead to vastly different organizations. The group of people participating in the DAO can choose to abide (or not abide³) by the rules as they are set out, and they can also collectively decide to change how the organization behaves by amending the DAO’s rules. In MolochDAO, for example, decisions regarding funding grants or accepting new members would be made and executed according to the predefined rules, whereas a decision to switch from 1-token-1-vote to 1-person-1-vote would be an adaptation to those rules. The process of constitutional amendment is amongst the most sensitive aspects of an organization’s rules.
Collectively, these on-chain (and off-chain) rule sets, smart contracts, and social decision processes are described by a constitution: a body of values and rules which govern the collective decision-making process of an organization. A constitution may be organized into one centralized document, but in many organizations it is distributed across several documents and instructions, from formal charters to codes of conduct to pinned posts by authoritative members. In an opt-in organization like a DAO, a constitution serves as a contract for participation — by participating, one implicitly or explicitly agrees to abide by the organization’s constitution. By regulating decision making in an organization, constitutions help us set the principles and rules for how we make rules (e.g. rules of order for a legislature), modify existing institutions (e.g. amendments to voting eligibility), and even design new institutions (e.g. creating an independent monitoring body). Good constitutions help institutions adapt to new circumstances, new memberships, and even new code.
Lastly, a computational constitution, implemented via constitutional code, is that portion of a constitution which is made up of software. For example, the way that a DAO runs its online voting processes (including user authentication, quorums, token weighting, proposal timing, etc) is part of its computational constitution. The underlying blockchain of a DAO has its own computational constitution, namely its consensus protocol. A computational constitution (for example, a smart contract) automates the administration of the decision-making procedure; this enforces the process but not the outcome. Among other benefits, the digital nature of DAO governance allows us to capture and translate repeatable governance patterns — constitutions as well as more basic rules and proposals — to serve as best practices for applications in various contexts. It is within the context of these patterns that we wish to develop an ontology of institutional economics that fits the DAO ecosystem, along with a map of how various projects could collaborate on and across these different layers of patterns.
Surveying the Landscape
Many groups are working on patterns and instances in DAO governance today — but they are working at different “governance layers” — across institutions, constitutions, computational constitutions, and DAO instances — that are not unlike the layers of a tech stack. At the constitutional pattern level, we have groups like the Metagovernance Project, who are building ontologies and tools to help people design constitutions for DAOs as well as for online communities far beyond the blockchain. Some of those constitutional patterns can be automated, which is where groups like the Commons Stack are using token engineering to design and test computational constitutional patterns, such as Conviction Voting. There are groups building replicable DAO patterns, such as Aragon, DAOStack, and Colony, which are working on interesting mechanisms like holographic consensus and courts for dispute resolution. We also have instances of particular DAOs such as Moloch and Metagame, which themselves have been forked repeatedly to serve the varying needs of communities eager to use these new tools.
We hope that this particular clustering of the DAO ecosystem helps to clarify what different organizations are contributing — to the organizations themselves as much as to people just learning about DAOs. We also hope that this clustering will catalyze research in this area, e.g. by identifying common and overlapping research topics. As the DAO ecosystem matures, groups will benefit from horizontal collaborations within layers, e.g. collaborations between organizations developing computational constitutions to develop common technical standards and shared infrastructure. But some of the richest future work — the killer apps, if you will — will be projects that span vertically across each of these governance layers, namely projects to build well-governed, rigorously-engineered, and fully-featured DAOs.
Context is Key
Institutions, organizations, constitutions, and computational constitutions give us a way of talking about the design of DAOs in a holistic way.
After all, a constitution (even a non-computational one) is more than a document; it’s a description of the values and principles that govern an institution’s decision-making processes. It defines roles and associated rights, the protocols for exercising those rights, and the conditions under which they can be amended (e.g. voting). But a constitution also provides a basis for a shared identity; it extols shared values and may provide vision or direction regarding future decision-making under those values without providing specific details as to what types of decisions might need to be made. In this way, a constitution can never be fully reduced to an algorithmic representation.
But not all constitutions are created equal — clearly, some constitutions help communities prosper and others lead to conflict and dissolution. In one famous set of institutions studied by Ostrom, namely the International Forestry Resources and Institutions (IFRI), different configurations of roles, rights, rules, and protocols led to widely different outcomes in different contexts. There was no single, “optimal” constitution or institution that worked across all forests and all communities. However, certain arrangements were repeated across many examples, such as monitoring systems and community consultation. We refer to such arrangements as constitutional patterns: each pattern refers to a family of related constitutions that may be applied to a local context.
In this sense, DAOs, as they exist today, tend to resort to a relatively small family of constitutional patterns related primarily to making and voting on proposals to direct the expenditure of funds. Raising funds, proposing expenditures for specific work, voting on those expenditures and verifying the completion of work, as well as lodging and resolving disputes, are necessary activities for a wide range of organizations. The institutions that underlie these activities are being implemented by organizations like 1Hive and are being assembled into reusable patterns by projects like the Commons Stack. It is not surprising that these elements of social logic have been the primary focus of computational constitutional development thus far.
Where to From Here?
These early examples have created a lot of enthusiasm around the idea of DAOs, but there is a lot of exploring ahead to discover workable institutional patterns (and anti-patterns). It’s also important to remember that the computational constitution is not the whole constitution, and that automation does not magically result in a healthy institution. We believe that taking an institutional approach, and adopting some concepts from social science, will help us to establish and steward healthy DAOs. The institutional approach can also help us organize the range of different projects in the DAO space (see diagram above), and identify common research questions. We’re researching design principles and institutional patterns in hopes of establishing a mature science and engineering practice for DAOs.⁴ Whether you’re a designer, a researcher, or just a fan of DAOs, we hope you’ll join us.
Do you agree with this framework and our choice of concepts? Are we missing something, whether an important concept, a layer in our diagram, or a particular organization working in this space? Feel free to leave comments & suggestions below!
If you would want to support the Commons Stack initiative and help us create a new way to fund public goods and open source projects, you can do this by applying to the Trusted Seed.
Acknowledgments: We would like to thank Seth Frey, Shermin Voshmgir, Griff Green, Anja Blaj, Allen Farrington, Abbey Titcomb, Grace Rachmany, Lawrence Lau, Kris Decoodt & more for their helpful comments and feedback in earlier drafts of this article.
 Many naturally-occurring patterns (e.g. pair bonds and families) can be described as institutions, but typically we do so only in cases where we can ascribe or add a human design or intent to that pattern. A mountain range is not an institution; the road that winds through it, however, is. Thunder is not an institution, but the idea of Zeus is.
 Almost all organizations function as institutions of some sort, and many institutions are structured as organizations, so many people use one word interchangeably with the other. Here, we’re using the words a little more formally. In place of “organization”, Elinor Ostrom often uses the more general term “action situation”.
 Given that a DAO has a social component, human decision-makers with open inputs can violate aspects of the social contract associated with that DAO without explicitly violating the rules which are enshrined in code. Court systems such as those being developed by Kleros and Aragon exist in part as a response to this phenomena.
 Here engineering refers to the social institution rather than the act of applying technology. Engineering as a social institution is characterized by a set of shared values, and in particular a constitution laid out in various codes of ethics upheld by the engineering profession across a range of subdisciplines.