by Felix Machart, Aug. 13

Last year we at Greenfield were searching for a comprehensive document to get an overview on various governance approaches applied by blockchain projects, ideally in one source — but we could not find anything, so we decided to create it ourselves.

This post lays out conclusions from the paper, which is linked in full at the bottom. We have been looking at blockchain and DAO governance from various viewpoints, first elaborating what governance is, what blockchains are in context of organizational theory, reviewing extant literature (blockchain focused, as well as selected works on classic economics and organization science), laying out distinct features regarding governance of major blockchain projects to then synthesize common elements and conclusions from both theory and practice.

What is governance?

Organizational governance refers to the means that organizations deploy to influence organization members and other stakeholders to contribute to organizational goals and purposes and the means by which the goals and purposes are determined.

Corporate governance is viewed as encompassing the rights and responsibilities of stakeholders towards a firm, which compete for resources in an institutional actor-centered view of the firm. In a broader sense, to account for the plurality of institutional settings for organizing productive endeavors in society, governance are all influencing factors of institutional processes, which includes nominating the controlling or regulating parties engaged in the process of organizing production and distribution of goods and services.

Organizational governance is heavily influenced by national governance, providing the constitutional and legal framework in which firms and markets operate to resolve certain market imperfections.

Overall, governance is concerned with considerations for total welfare maximization as well as balancing the interests of a wide set of stakeholders with various needs, values and political viewpoints. Stakeholders are identified through the actual or potential harms and benefits that they experience or anticipate experiencing as a result of the organization’s actions or inactions (as opposed to “strategic” stakeholders as defined by Freeman — “without which a firm would cease to exist”). Most probably, there cannot be one optimal form of governance, as stakeholders have diverse interests and values and there exists a large trade-off space in policy decisions. Thus, different communities need to maintain distinct visions, objective functions as well as governance systems that attract certain members, enable them to provide input as well as exit if there is no longer a fit between the collective and the individual.

In order to build governance systems with and for blockchains it is crucial to ask what they are. Are blockchains institutional technology to serve the private interests of their initiators and community? Are blockchains a public good? Digital commodities? If the “community” owns the platform/blockchain, are there stakeholders who are affected but not properly represented?

We view blockchains as institutional and governance technology that can govern collective action but also needs to be governed (ultimately by humans) in order to represent the governed. Thus, one can separate governance by blockchain infrastructure (akin to organizational and mechanism design — the structuring of communication, coordination, and control) and governance of the infrastructure.

Blockchains continue the trend that open-source communities started in creating distributed innovation systems, which push the firm out of the center of prominence in coordinating value creation.

Governance by the infrastructure — distributed innovation systems on blockchains

Three distinct components define a new era of organizational design driving distributed innovation systems, that span open-source communities, platforms and open organizations that have been inspired by them: 1) interface design (mediating interaction within and across systems, as well as their different sub-systems and actors); 2) participatory architectures (enabling peers to articulate ideas and contribute meaningfully, or to provide other resources such as computing power or data); 3) evaluative infrastructures (accounting mechanisms judging quality and value).

Distributed innovation systems that are based on governance by blockchain infrastructure can potentially solve the problem of more platform monopolies emerging, by making the logically monopolistic network contestable based on the threat of forks (of the single protocol/ruleset, as well as data that is owned by users) which makes the market for middlemen more competitive, while benefiting from decentralized infrastructure and control. Thus, blockchains facilitate the networking of distributed peers, without the market power costs that usually come with centralized operators.

From reducing transaction costs due to distant search to alleviating monopolized trust

Generally, inefficiencies in trade arise when parties do not reach the best possible collective outcome through transaction costs such as search costs, complexities in contract drafting and enforcement or incentive problems like the hold-up problem.

The internet has largely tackled the frictions that arise through search costs, while blockchains are interesting for resolving the frictions that arise due to a lack of trust. The hold-up problem, one premier reason why firms exist (as contracts are usually incomplete), could be solved in a novel way, by using blockchains as neutral, shared databases, on which crucial components of distributed innovation systems depend, namely architectures for collaboration as well as evaluation.

Rules are clearly and transparently defined and to the largest extent immutable (viewing the case of The DAO hack in Ethereum and the resulting hard-fork to manually reimburse victims as an exception and symptom of an early ecosystem). Thus, business relationships ought to be more predictable and leverage to renegotiate of a stronger partner/platform operator can be mitigated (which allows rational actors to enter into such a relationship in the first place, as no re-negotiation is expected).

Contracts can potentially be designed in a more complete way, leveraging commons libraries of contingencies that cover more edge cases and automated enforcement. However, many organizations will still depend on incomplete contracts and human judgement. Aragon Agreements is an example of natural language contracts, that are interpreted by Aragon court, a game-theory based, distributed arbitration court to address incomplete contracts and human judgement in a decentralized way.

Permissionless public infrastructure creates resilience

A protocol or smart contract can set incentives, by which it defines an objective function, and programmatic rules, according to which stakeholders need to operate, quite similar to the “nexus of contracts” that represents a firm. One stark difference from firms is, however, that DAOs either represent permissionless layer 1 infrastructure or are built upon such, which allows for increased scalability (low friction to join, automated rules enforcement), resilience (while some peers can drop out, others can easily come in to gain the reward) and innovation (functionalities can be easily integrated and stacked upon each other).

Besides pure profit motives, which appeal to extrinsic motivation, intrinsic (the pleasure of executing a task) as well as identified (to achieve a shared goal) motivation are strong drivers of human behavior (which are especially important in the open-source space). The essence of a DAO goes beyond the pure code that defines its rules, and members are often attracted by ideology and stories, which they then in turn shape as well.

Balancing immutability/stability with change

One can debate the extent to which blockchains are and should be immutable, which has a large effect on governance by the infrastructure. If one of the most crucial aspects of a layer 1 DAOs/blockchain is to securely store and transfer digital property rights (which promotes access to capital and economic development), as well as represent reliable and legitimate institutions for layer 2+ DAOs, it should favor stability over change (as stable systems/institutions have historically been more economically successful and less change opens fewer avenues for questioning legitimacy of an institution). However, systems will always need to evolve, as circumstances and the people they serve change, which necessitates governance of the infrastructure.

Governance of the infrastructure

Communities have been experimenting with various systems to govern the infrastructure, from informal to formal processes, from loosely coupled off-chain to tightly coupled on-chain systems. There is no doubt that voting plays an important role in gauging community sentiment and to condense individual preferences into a picture that reflects the aggregate. The main questions are, 1) which stakeholders are entitled to vote (to the largest extent token-holders in permissionless systems, as “1-person-1-vote” has so far been either reliant on centralized KYC or not sybil resistant) and 2) how tightly coupled is the result of the vote with a protocol change. As for the 2nd dimension, there is a strong argument regarding the importance of the default behavior of client software that powers full-nodes (which have shown a dominant position in enforcing consensus rules as in the case of the UASF) and whether those stakeholders should be required to opt-in to a new set of rules or required to opt-out actively in order to avoid a change (given the former will be biased towards the change, as there is a strong Schelling point/coordination flag towards the tightly coupled result). Non-mining full-nodes or users running full-nodes that do not hold a significant number of tokens do not have a sybil resistant mechanism to signal their stance towards an upgrade such as token-holders. Thus, the requirement for full-nodes to actively opt-in to a protocol change represents an important counterbalance of power that leads towards a more holistic stakeholder representation (though with a bias towards the status-quo, that can lead to stasis in the worst case). Tightly coupled on-chain voting, however, forces nodes to implement change if decided so by token-holders (if they do not opt-out and fork off, which might be more difficult to coordinate).

DAOs like Pocket network are exploring to organize around a constitution which lays out the social contract of the community as an “off-chain protocol”, while its consensus mechanism defines the explicit contracts (complete contracts) of using the on-chain protocol. To balance stakeholder interests legislature (token voting) is limited by the judiciary (Aragon court interpreting the constitution), while a foundation is set up to implement valid decisions across chains (executive), which is inspired by trias politica.

Stakeholder lock-in as a measure of interest alignment & legitimacy of governance rights

Agency theory commands that the providers of specialized inputs should be in charge of governance, which has traditionally been providers of physical capital/investors, as they are locked-in with their investments (sunk costs — inputs are not useful otherwise or it is costly to make them so). In the knowledge age and digital sphere it is increasingly specialized knowledge, data and the use of standards, but also capital that is being invested, staked or locked-up for extended amounts of time to contribute to network security and value creation that represents those specialized inputs or sunk costs.

While user and app developer lock-in through user-controlled data, open code and the ability to fork is considerably decreased compared to centralized platforms, it still exists due to network effects and coordination problems that make it harder to fork together with large stakeholder groups. Thus, the requirement to opt-in to protocol changes by full-nodes that are run by and represent power users (strategic stakeholders — sophisticated end-users, exchanges, app developers) such as in non-tightly coupled voting as in off-chain governance, might be necessary as long as there is no sybil resistant way to include them in tightly-coupled on-chain governance. Core developers that provide their specialized input arguably have strong influence in both tightly coupled on-chain as well as loosely coupled off-chain voting, due to information advantages, as well as them actually implementing upgrades. It is true that tightly coupled on-chain governance can also lean towards conservatism and stability by requiring higher quorums and/or supermajorities, however it is still only token-holders that vote, as opposed to a broader strategic stakeholder base.

As the level of lock-in depends on the strength of network effects, a network can arguably be governed in a more centralized way early in its evolution (which allows a faster pace of change and iteration), however it is important to engage a broader stakeholder base as the network scales in order to set up a sustainably successful ecosystem (which can be established through the vision & roadmap that is communicated early on, through which stakeholders self-select to adopt the protocol — this can then create a Schelling point for stakeholders to converge on).

Stakeholder representation vs. scalability of decision making

There are large differences in participation rates in voting, due to differences in community size and composition and differences in voting systems as well as quorum requirements. Direct democratic approaches suffer from low participation as decision-making involves cognitive costs, so some amount of representation by councils, boards and/or through liquid democracy might be necessary in many cases in order to enable effective decision making (e.g. voting power delegation in Compound). Liquid democracy structures are blending the best of representative and direct democracy, allowing anyone to directly vote on an issue, if she feels to be well informed or there is a strong opinion, as well as allowing for delegating one’s voice to a representative (like a liquid management board or protocol politicians). Thus, voters who are rationally ignorant might have enough incentive to at least delegate, while the cognitive load of deciding on a matter is outsourced to the delegate (which increases turnout). Voters who are motivated to decide for themselves can still do so, which increases decentralized input. There have been votes that were able to be influenced by large individual voters, which becomes easier as there is low participation. Adaptive quorum biasing, introduced by Polkadot is an interesting approach to autonomously protect against low turn-out votes having a result that does not reflect the whole. Making an elected board decide and let them be overridden by a sufficient threshold of voters is also a pragmatic way of leaving uncontroversial and day-to-day decisions to agents that focus on the matter, while keeping the option of decentralized direct decision-making. One could argue Nexus Mutual does not represent a true DAO as it has elements of centralization and lacks true autonomy and permissionless-ness (limiting influence of individual voters to 5% by KYC). However, also other projects show centralized and permissioned aspects, highlighting the fact that decentralization is multi-faceted as well as a spectrum and not absolute. Finally, token/vote distribution (incl. the level of participation and decentralization of full-nodes) is the most crucial factor in determining the extent to which power is decentralized and communities ought to target and select long-term, mission aligned and knowledgeable investors (time-locking for voting can be a feasible proxy for long-term alignment).

Multi-layer Governance

One of the most important questions is what is the highest institution a blockchain or blockchain-based organization is embedded in? One cannot view any of the layers in isolation, but mutually intertwined and people’s behavior is informed by the sum of this complex system. If a DAO aspires to be autonomous in the sense that it understands its purpose as representing a completely self-sovereign entity, a governance approach that balances all stakeholder interests (not only strategic) with democratic (1p1v) legitimacy might be required in order to limit negative external effects such as carbon emissions of PoW (while it is questionable how far ranging the voting population is defined and decentralized identity has not been solved yet).

This stands in contrast to the view of Vitalik Buterin that if Ethereum becomes something political, which rules you can debate, its utility gets reduced considerably (as a neutral, immutable state-machine). In fact, cryptocurrencies have been started due to the frustrations with the current imperfections in governance, such as collusion and bribery. However, in practice, already The DAO hack and subsequent split into Eth/EthClassic has shown the political aspect. Vitalik makes the distinction that an entity that changes/evolves due to the collusion of 51% of actors is a DO while DAOs are resistant to collusion (which he argues to be especially important, given plutocracy favoring effects such as wealth concentration and bribery). This hints towards him wanting Ethereum to be more of a DAO than a DO that can explicitly be influenced by a majority of e.g. token holders through on-chain governance (one has to keep in mind though that the term DAO is used loosely in practice and we see projects with on-chain governance also as such — see paper). He highlights that a base-layer chain should be innovative in the short term and stable in the long-term, while satisfying hard crypto-economic guarantees in order to be reliable. Other community members, however, frequently highlight the fact that autonomous software is to be avoided, due to ex-ante unknown external effects. Rough consensus and running code with core developers balancing stakeholder interests is resistant to be captured by hostile takeover of token-based voting rights, but excessive power of developers on the other hand can also be a risk, while still being kept in check by users/nodes needing to adopt a given update.

In contrast to the viewpoint that blockchain networks could be platforms for governing society, members of the Decred and Tezos communities argue that cryptocurrency networks are rather platforms for coordinating digital commodities and thus better compared with corporate governance than national governance. A DAO that subordinates itself to a higher-order governance system that is democratically legitimated (akin to a nation state), can operate based on token-holder based on-chain governance, while being held in check. However, bribery and manipulation of news channels is a real phenomenon that must be solved for crowd participation to be effective in any case. The approach that a cryptonetwork is regulated (by states) at the edges (node/exchange) level basically translates to a setting in which actors are governed by various layers, with state level regulation as a meta-layer finally trumping blockchain incentives for an individual. Still, a decentralized network as a whole can remain resilient to changes in legislation in any particular jurisdiction. It is however crucial, for a cryptonetwork to maintain effective interfaces with jurisdictions (e.g. legal wrappers, compliance tools, stakeholders negotiating/taking into account rules and protocols) in which it is supposed to be adopted, as interoperability drives network effects and thus value.

Ultimately, governance is about social consensus

In the end, social consensus is what defines a cryptonetwork. The option to fork is the most crucial instrument of last resort to force decision-makers to take stakeholders into account, while effective governance gathers maximum stakeholder voice in order to avoid exit. Coordination mechanisms that allow for coordinated switching to a new fork are important to make the threat of a fork realistic. A default setting of not to update, as in loosely coupled off-chain governance creates a coordination flag towards stability, important for base-layer institutions.

As network effects exist, there are strong benefits to remain unified instead of splitting communities. As a result, if values and objectives amongst communities are similar enough to agree on common protocol parameters, there is considerable value to be created by preventing factions to split. Thus, value can be maximized if decision-making processes enable finding common ground. Designing good interfaces and participatory infrastructures that enable the aggregation of viewpoints of diverse stakeholder bases are of utmost importance. Also, effectively filtering through proposals and gauging community sentiment before resources are spent to develop an upgrade that might not be adopted are crucial.

What is more, management changes from hierarchical relationships to community management through diplomacy “to govern the ungovernable — the anarchical society — through discursive and cultural practices“, as open network-organizations made up of self-sovereign actors (to whom the locus of control shifts, due to them being allowed to exit and self-select to a new protocol, for which exist lower barriers to entry). In open networks of peers with distributed leadership and agency a manager-diplomat needs to be creating the conditions for collective action to occur. Community management in blockchain ecosystems is therefore of utmost importance, not only to attract the best possible community (with a fit regarding values and skills), but also to moderate and translate amongst different stakeholders in order to crystallize common or opposing viewpoints and unite the community to a coherent whole.

It is culture that represents the deep values driving participants (i.e. the social protocol that runs on people’s minds). Thus, culture shapes all governance layers above and especially also which protocols participants self-select towards. A community’s culture (like a company culture) gets built over a longer time horizon, is mostly implicit/tacit (exists in the participants’ minds, potentially only subconsciously) and is thus hard to codify and as a result, hard to imitate (as opposed to open-source code).

Bootstrapping DAOs could be summarized to be establishing a vision, purpose and culture (creating memes), setting objective functions (incentive systems) and communicating those to enthusiastic initial members (potentially incentivized by early token issuance) until resource gravity/network effects kick in to self-sustain growth.

The discussion around governance is to some extent highly ideological, as no optimal form of governance exists — considerations for efficiency, as well as trade-offs among stakeholders have to be taken into account, which makes the process inherently political. Thus, there is room for a diversified set of DAOs, covering broad ranges of objective functions and value systems, that will be also reflected in their governance systems. In a space where code and data are open (and/or user-owned), governance driven by values, norms as well as community composition and thus access to knowledge, offers sustainable competitive advantage.

For more details, read the full paper, which is engraved here on Arweave.

Many thanks especially to Mario Laul, Patrick Rawson, Jack Laing, Sebastian Blum, David Fauchier and Lane Rettig for their helpful feedback as well as to everyone inspiring this research through conversations and prior work.