Experiments in algorithmic governance continue
Trying not to fail at Decentralised Autonomous Organisations (DAOs)
29 July, 2021
“Decentralised Autonomous Organisations” (DAOs) are a new form of politically decentralised organisations that promise “autonomy”. In 2017, Quinn DuPont authored a book chapter on “Experiments in algorithmic governance: A history and ethnography of “The DAO,” a failed decentralized autonomous organization”. Yet, these experiments continue. This blog series explores the questions, ‘are decentralised autonomous organisations resilient, and if so, when and how do they break?’.
This blog is in two parts. Part 1 explores the historical background to the emergence of the idea of modern-day DAOs in blockchain communities today. I emphasise the dynamics of political decentralisation in DAOs which originates with the cypherpunk movement, as well as the tension between the role of humans and algorithmic governance in the concept of “autonomous” organisations, which is rooted in the longstanding “on-chain” and “off-chain” governance debate. I emphasise DAOs as socio-technical construct, and explore some of the common approaches to establishing a DAO, including DAOFirst and “exit to DAO”. I then further refine the approach to resilience in DAOs, by setting out common vulnerabilities for further ETHnographic enquiry.
Part 2 will observe DAOs in action to explore if DAOs are resilient (meaning participants continue to participate towards said objectives), and if so, what enables resilience between the human and automated components of DAOs, or what hinders it.
This piece contributes to building a DAO lexicon, and exploring key themes, tensions, and dynamics in DAOs, in relation to the human outcomes for those participating in them.
The blog precedes a conversation on “DAO implementation”, which I moderate for “Smart Contract Research Summit”, August 5-7, 2021. It also contributes to early analysis on my PhD on “Resilience in Decentralised Technologies”. Constructive engagement is most welcome.
Suggested citation: Nabben, K. “Experiments in algorithmic governance continue: Trying not to fail at Decentralised Autonomous Organisations (DAOs)”. Substack. 29 July, 2021. Available online: https://kelsienabben.substack.com/p/experiments-in-algorithmic-governance.
“Decentralised Autonomous Organisations” (DAOs) are a revolutionary way of self-organising, dreamed in the tradition of the dominant libertarian imaginaries of the cypherpunks, where governance is administered by software code in a network of people. Blockchain-based tools are at the stage where they are mature enough to support rapid experimentation regarding this concept. DAOs are often promised to be participatory, highly democratic systems that offer operational efficiency. DAOs are socio-technical hybrids, that operate in both the “on-chain” and “off-chain” world. The nature of these systems is still emergent, and the human outcomes of these systems remains understudied. This piece explores the question, “are DAOs resilient?”.
In my studies of resilience in decentralised technology communities, I argue that resilience is the ability of a socio-technical system to adapt and transform in response to threat or crisis. In DAOs, resilience is the ability of the DAO to persist in response to social or technical attacks or shocks. I observe vulnerabilities as a relational notion to resilience in order to study socio-technical systems. I approach governance as the field of action for coordination and control of a complex system. DAOs, require establishing the field of action (initial settings), as well as the technical, social, and cryptoeconomic actions possible.
What is a DAO?
The acronym “DAO” stands for “Decentralised Autonomous Organisation”. The initial, basic purpose of a DAO is that of a virtual entity where members have the right to spend funds and modify code. Although there is no formal definition of a DAO as these institutional forms are still evolving, a DAO is a model for coordinating amongst peers with no central intermediary, towards a stated objective.
In 2013, co-founder of Bitshares, Steem, and EOS blockchain Dan Larimer described Bitcoin as a type of DAO, using the metaphor of cryptocurrencies as shares in a “Decentralised Autonomous Corporation” (DAC) with the goal of earning profit for shareholders by providing services on the free market. Five days later, then author at Bitcoin Magazine Vitalik Buterin (now co-founder of the Ethereum blockchain), pointed out that corporations are “nothing more than people and contracts all the way down”.
The concept of DAOs has since been popularised by blockchain communities, especially in the Ethereum ecosystem. The software language of the Ethereum protocol allows automated, smart contracts for the enactment of composable governance processes and mechanism, and DAOs are proliferating as an open field of experimentation in automation, governance, and autonomy. DAOs have been referred to as a site of algorithmic governance for further ethnographic enquiry.
The history & purpose of “Decentralised, Autonomous, Organisations”
Decentralised Autonomous Organisations enable things to be organised in a politically decentralised manner. “Decentralised” refers to architectural decentralisation of physical computing hardware, or “nodes” in the peer-to-peer network, as well as freedom from coercive authorities or intermediaries that may hold power to influence a system.
“Autonomous” refers to independence, or self-governance of individuals and the organisation itself. This is not be not to be confused with automated, although the automatic execution of rules leads to autonomy. According to Buterin, the idea of autonomous systems pre-date blockchain communities (see Robin Hanson’s futarchy, “a mechanism for organizational governance via prediction markets”, “automaton” self-operating machines, and the novel series Daemon.
DAOs are a continuation (the logical extension) of the cypherpunk ideal of cyber and physical autonomy. The cypherpunks deeply explored ideas of automated, digital markets and physical world outcomes in the 1990s, in the lead up to the invention of Bitcoin. Some even advocated for the abolition of property rights and creation of “Temporary Autonomous Zones”, of ad hoc, self-governing territories of non-hierarchical social systems that “elude formal structures of control”. DAOs operate in digital space but extend to influence the physical world.
In 2014, Buterin describes that “instead of a hierarchical structure managed by a set of humans interacting in person and controlling property via the legal system, a decentralised organization involves a set of humans interacting with each other according to a protocol specified in code, and enforced on the blockchain.” While these ideas existing in business, economics, cybernetics, and politics prior to the open-source blockchain communities of today, decentralised, public blockchains make them possible.
One way to analyse DAOs is in terms of what is being organised, what is being decentralised, and who, or what, are being made autonomous? This includes whether autonomy is about individual autonomy, or collective autonomy, and what trade-offs are required to optimise for each.
The role of humans in autonomous organisations
The idea of DAOs is applicable to both corporations and communities. DAOs have internal capital (or property). The primary stakeholders of a DAO are investors, employees, and customers.
The tension in decentralised governance that Buterin emphasizes is, how much do we really need people in algorithmically programmable organisation? While some human action is necessary for higher order specialized tasks (not the other way around), people are increasingly less essential in the day-to-day operations of an organisation in the post-industrial era.
“In an autonomous agent, there is no necessary specific human involvement at all; that is to say, while some degree of human effort might be necessary to build the hardware that the agent runs on, there is no need for any humans to exist that are aware of the agent’s existence.” – Vitalik Buterin, 2014.
The science fiction dream is that “autonomous agents” would be actors, or stakeholders, in these organisations. A fully autonomous agent is the idea of fully Artificial General Intelligence (AGI). This encapsulates the very essence of resilience, in that the agent could adapt to circumstances, to transform and survive to meet its aims, into perpetuity. Observing attempts at “Decentralised Autonomous Organisations” reveals early dynamics of the social outcomes, benefits, and concerns, to hypothesize about the role of autonomous systems, and the possibility of autonomous agents, as coordination infrastructure.
DAOs “think” for themselves, with automation at the centre, and humans at the edges.
DAOs, DACs, DAs and More: An Incomplete Terminology Guide - Vitalik Buterin, 2014.
“This vision stands in stark contrast to the state and corporate sponsored surveillance super-structures which are the primary applications of advancements in artificial intelligence funded and deployed by centralized institutions” state Voshmgir, Zargham, & Emmett, in “Conceptual Models for DAO2DAO Relations”, 2021.
Public decentralised blockchains, smart contracts, and the concept of DAOs provide the possibility to explore how much of an organisation’s bylaws can be translated into software code and executed by smart contracts, to function autonomously from human direction.
What is being automated in blockchain institutions (such as DAOs) is some measure of trust in the system, to enable scalability of digital social institutions and the advancement of society. In the well-known essay titled “Money, blockchains, and social scalability”, Nick Szabo states that blockchains reduce human “vulnerabilities to our fellow participants, intermediaries, and outsiders”, which increases efficiency of resources and thus, social scalability. “Trust minimization is reducing the vulnerability of participants to each other’s and to outsiders’ and intermediaries’ potential for harmful behavior” - Nick Szabo. DAOs intend to reduce human vulnerabilities, through blockchain enabled cryptographic trust minimization and automated efficiencies, to produce scalable, independent, self-directed social institutions and societies.
It remains to be seen whether individual and collective autonomy are indeed being enabled by these new, cyber-physical institutional forms. The next section will explore types of DAOs, the common approaches taken to create a DAO, as well as a definition of resilience in DAOs, for further case study analysis.
DAO governance: Technical on-chain and social off-chain
Scholars have referred to two main types of DAOs as participatory and algorithmic. Participatory DAOs are managed by distributed consensus through smart contracts to signal the preferences of members. Algorithmic DAOs aim to be entirely algorithmically governed, with the underlying smart contracts dictating the entire functionality of a DAO.
An example of a participatory DAO might be GitcoinDAO, which has a council of governors, which can allocate governance tokens to stewards, deliberates proposals on forums, votes on “Snapshot” (an “off-chain” voting tool), and then executed voted proposals, such as allocating funds. An example of an algorithmic DAO is DxDAO, which launched with no pre-defined members, and is completely community run without any intervention of a project or team, which controls a decentralised trading protocol and other DeFi tools. The aim of DxDAO, as well as other DAOs which may define themselves as “algorithmic”, is to be “as widely distributed as possible from day one”.
Yet, all DAOs are both participatory and algorithmic. This is what is unique about DAOs. DAOs require participation, which is a political process where the actors involved in decision-making processes are positioned towards each other through power relationships that occurs “off-chain”, and algorithmic governance, which is the point at which decisions are reflected and executed “on-chain” through the use of smart contracts. There is necessity and value in both the technical (including cryptographic and algorithmic) components of DAO structures and governance, and the social (including normative decision making, or decision making in line with cultural values of what is acceptable).
What this distinction in thinking between these two component points to is a much deeper cultural dynamic about the value and risks of social components versus the value of technical components in blockchain communities. This binary tension has repeated throughout blockchain history and is rooted in the history of blockchain governance.
A brief history of blockchain governance as “on-chain” and “off-chain”
Historically, blockchain governance refers to the rules among blockchain of how the software code of a blockchain protocol is changes. Ideas around blockchain governance are heavily embedded in culture, such as Bitcoin’s ideals of immutability, and minimal trust between people, which result in an ideology of “code is law”. This debate about the role of people to intervene in algorithms is also referred to as “on-chain” versus “off-chain” governance.
“On-chain” governance is when governance rules are made explicit in software code, and blockchain nodes automatically execute code upgrades in the protocol in response to on-chain coin holders voting processes. In contrast, “off-chain” governance is when rules are much less formal, and the non-code-based processes of how ideas are shared, discussed, and evaluated outside of formal, recorded, transparent decisions are eventually reflected in nodes decisions upgrade their software to pass on changes to the protocol. Therefore, nodes in the peer-to-peer network are active participants in governance processes.
The premier example of the tension between algorithmic prominence, and normative (establishing norms, according to shared values) decision-making is “The DAO” hack. A bug in the code which allowed millions of dollars to be drained from the smart contract resulted in arguments in the community between whether it is right or wrong to change the record of transactions. This led to a “hard fork” of the Ethereum protocol, which established Ethereum and “Ethereum Classic”, as well as early Ethereum CTO Gavin Wood founded his own protocol known as “Polka Dot”.
In the aftermath of the hack and ongoing debates on governance, Buterin stated that “people who think that the purpose of blockchains is to completely expunge soft mushy human intuitions and feelings in favor of completely algorithmic governance (emphasis on "completely") are absolutely crazy”.
Today, the Ethereum community typically embraces its “off-chain”, human-oriented processes. For example, the Ethereum Foundation maintains an essential coordination role in the development of the protocol, such as hiring research and development staff and chairing regular “Ethereum Improvement Proposal” (EIP) meetings. The advantages of this approach were evident when the community needed to coordinate a rough social consensus miners, exchanges, and node operators to upgrade their software to “fork” the protocol after “The DAO” hack. One disadvantage of a bias towards “off-chain” governance is mere efficiency and scalability, with the Founder of Ethereum Vitalik Buterin stating that the Ethereum2.0 protocol upgrade has been a significantly slower process than originally anticipated.
In contrast, other blockchain communities such as Polka Dot and Tezos embrace on-chain governance, whereby protocol upgrades are determined through referendum type votes, and software code is immutable, meaning it cannot be changed once deployed. An advantage of this is representation and transparency. A drawback of this approach is to reduce governance to voting, when it is in fact multiple layers of social and technical components and processes interacting in the coordination and control of a system, which can’t be anticipated in advance, and are therefore impossible to encode into the rules of protocol in advance.
Blockchain communities orientate around the value of decentralisation of political power. On-chain governance processes establish explicit rules of governance and offer transparency. In contrast, off-chain governance processes are opaque, messy, and inherently political, but in part unavoidable, and perhaps helpful, in leading, educating, and establishing clear and trusted direction in projects. The affordance given to social and technical aspects of governance in DAOs is not just a system design choice but an ideological choice. Participants in blockchain communities and governance develop cultural norms, which affect information and incentives.
It remains to be analysed what outcomes and emphasis on “off-chain” or “on-chain” governance processes produces, and which is more resilient under certain circumstances.
“DAOism” – modern day DAOs
It’s safe to say that DAOs have become a “quasi-cyber-religion”.
Modern DAOs are commonly “decentralised applications” deployed on top of the layer 1 protocol blockchain. The capabilities that public blockchains enable for experimentation with DAOs is evidenced by the plethora of approaches to “DAOs” popping up in blockchain communities today.
To obtain legal status as a registered corporate entity (in the US), a DAO must:
· be deployed on a public blockchain,
· provide a unique public address (has) so anyone can view their operations,
· software code must be open-source, software code must be audited, laypeople able to read smart contract variables and token restrictions,
· governance must be decentralised in the technical architecture of the DAO,
· at least one DAO member,
· a contact point,
· a binding dispute resolution mechanism for participants, and,
· a dispute resolution mechanism for interacting with third parties, outside of the DAO.
There are two simultaneous types of governance at play in blockchain infrastructure as a socio-technical systems: governance by the infrastructure and governance of the infrastructure. This duality is evident in both the role and function of existing DAOs.
DAOs exist for venture capital allocation (FlamingoDAO), funding public goods (GitcoinDAO), managing “Decentralised Finance” (DeFi) protocol, such as Automated Market Makers (UniswapDAO), funding life extension research (VitaDAO), and building themselves (1Hive). For lists, see here, here, here, and here.
These can be further categorised, although definitions remain broad. DAO participants govern the rules of the protocol, as well as providing labour within the DAO, to be governed by the protocol.
Autonomy via composability
DAOs are developing towards the ideal of autonomy through the composable nature of blockchain infrastructure. A DAO, such as 1Hive, can run on a protocol, such as Ethereum. Unique identity can be verified through decentralised applications like BrightID. It can be publicly funded through a bonding curve, fair launch, or retroactive airdrop. Labour contributions can be recognised and measured in “cred”. Disputes can be arbitrated in decentralised courts through (such as Aragon or Kleros). And insurance can be supplied in the case of hacks or bugs in the code that lead to losses. These provide modular, composable, multi-layered mechanisms for executing specific governance processes that transition between physical space and cyberspace.
The next section outlines the common approaches to ‘creating’ or ‘becoming’ a DAO through the “DAOFirst” or “exit to DAO” approaches.
Just DAO it (birth)
“How to start a DAO”, Tweet by Kevin Owocki, Co-founder and CEO of Gitcoin, now GitcoinDAO.
“DAO First” verses “exit to DAO”
“A general DAO is like a living creature, with its own subjective mind that senses, perceives, thinks, makes sense and takes decisions.” - Matan Field, DAOStack.
DAOs are undertaking two main routes towards becoming a DAO. One is the “DAO First” approach. The other is to transition through a progressive route to decentralisation, which I term, “exit to DAO”.
The “DAO First” approach
Kain Warwick, Founder of Synthetix DeFi protocol has pioneered the articulation of the “DAO First” approach. DAO First refers to determining the initial rules of token distribution and capital formation from the outset of designing a DAO, rather than the alternative of setting up and transitioning from a traditional company to a DAO.
“Exit to DAO” (progressive decentralisation)
The alternative to a DAO First approach is progressive decentralisation. This approach is seen in numerous projects that began as a product or idea, and given the nascent maturity of DAO tooling, now wish to “decentralise” decision making power to the community. Progressive decentralisation is when token distribution and capital formation rules are determined after a project has a community. The complexity of this is that decisions are required to be made with the community to be seen as legitimately decentralising the project, but can be embroiled in ambiguity, protocol politicking, and bureaucracy (or “fucking yourself”, according to Kain).
I refer to “progressive decentralisation” as “exit to DAO” (which has affectionally been coined “DExit” by Rich Brown from SCRF). In “exit to community” by Nathan Schneider and Morshed Mannan, there are three tools to allow ventures to hand over control to their community. These are: i. the trust model, where ownership is centralized in a trust that is governed by a trustee; ii. the federation model, businesses composed of multiple independent business entities, and, iii. and the tokenization model, using blockchain technologies. In DAOs, we are seeing all three tools at play: from multisig wallet holders as “trustees”, token whales ensuring the stability of the system, and tokenisation to facilitate participation by the broader community. The risk that I point out is that the rapid exit of the project founders to “decentralise” to a DAO places numerous substrates of these tools for institutional infrastructure design at risk.
While there is a lot of emphasis on starting DAOs, their common attack vectors, and the ways in which they can extract value, garner less attention.
The remainder of this piece seeks to explore the DAO dynamics that create and address vulnerabilities in DAOs, as a means to understanding what social and technical aspects of DAO structures or governance (as decision making rules and processes) are stabilizing or undermining to resilience in these socio-technical algorithmic assemblages.
Defining resilience in DAOs
DAOs live and die by participation. A resilient DAO is one that maintain a critical mass of participants in order to continue to pursue its stated purpose and objectives. “Acts in a decentralized decision-making system will be denoted resilient if it ensures all decisions made in the DAO to agree with its approximate opinion” states Matan Field, Co-Founder of DAOStack. Without a stated purpose, and engaged participants, measures for resilience and ‘success’ are unclear. Thus, a resilient is one that acts in line with its own mission and governance purpose.
The tension in DAO communities is how to build scalable, decentralised, autonomous organisations, that effectively balance social and technical dynamics to achieve their stated purpose. Thus, defining the governance surface, and enacting governance are key levers of resilience.
DAOs are vulnerable to both social and technical attack.
Social attacks include collusion and sybil attacks. Collusion attacks are when a significant percentage of a certain type of members collude specifically to direct the DAOs activities. Sybil attacks are where people create multiple identities to undermine the rules of reputation and exploit the system. Bribes and centralization are also seen as threats to DAOs, which may be a consequence of mechanism design (information and incentives). A new term has emerged to describe governance vulnerabilities specifically, known as “governance extractible value”, to describe capital structure exploitation and short-termism.
Technical attacks can be software code bugs that create security errors, as demonstrated in “The DAO” hack.
This initial analysis of general DAO vulnerabilities surfaces the importance of culture for participation, and thus resilience, in DAOs.
Essential components for DAO life
“when a company becomes a DAO”, Tweet by Kevin Owocki, Co-founder and CEO of Gitcoin, now GitcoinDAO.
A culture of participation
DAOs are an assemblage of “people, tools, and processes”, according to Michael Zargham, Founder of Blockscience. This includes culture, rules and incentives encoded as technical parameters, and information, towards a clear objective or purpose. These elements then co-construct one-another in evolutionary, adaptive processes, sometimes strengthening the life of the DAO (participation towards said objective), and sometimes hindering it. DAOs must incentivise participation in order to live.
Profit and culture are two, major incentives for participation in DAOs. Participation is an essential component of DAOs. As DAOs are both social and technical, some of the aspects that motivate participation in DAOs is culture and profit. DAOs typically have both to survive.
Everybody hears about cryptocurrency when the price is up. DAOs use digital token incentives to establish desired behaviour for participants and punish undesirable behaviour (such as ‘slashing’ in Proof-of-Stake). DAOs are partly comprised of degenerates, “a subculture associated with a disreputable corner of decentralized finance known for pump and dump schemes”, known as “DeFi Degens”.
The other ingredient is culture. Culture is understood as the social behaviour and norms of a human society, expressed in the knowledge, beliefs, and customs of individuals and groups. For example, FlamingoDAO “aims to support, purchase, archive, collect, and potentially tokenize important pieces of this ecosystem”, specifically, NFTs. It is investing in recognising and preserving valuable cultural artefacts. Yet, it is comprised of registered investors and recognising the profitability of its investments in these cultural assets. In contrast, other DAOs such as 1Hive use digital tokens to incentivise labour within the DAO, for the development, growth, and sustainability of itself.
In “After the Digital Tornado”, Kevin Werbach argues that governance has to be “by-design” in blockchain systems, it cannot be an afterthought. Perhaps, culture too must be by-design, as it is what carries communities through adaptations and engages participants throughout shocks and transformations to the system.
Part 1 of this blog has outlined key historical and cultural dynamics of DAOs, especially the tension between human and algorithmic influence in DAO governance. I have outlined the history of the modern blockchain community concept of DAOs, the tension between human and algorithmic components in DAO governance as steeped in “on-chain” and “off-chain” governance debates, and outlines some common approaches to establishing a DAO, which are useful delineations for later case studies.
This has established a basic DAO lexicon by which to observe the elements and categories of DAOs against certain vulnerabilities to observe and understand resilience in DAOs as examples of attempts at political decentralization via algorthmic governance.
Part 2 will observe common social and technical vulnerabilities in DAOs in action through ethnographic case studies. This includes looking for cases of DAO death, and resurrection. Analysing how different types of DAOs (on-chain and off-chain governance) are responding to these will provide a better understanding of if resilience is possible through politically decentralised, autonomous organisations, and the dynamics, risks, and trade-offs for participants and DAO in this phenomenon.
With thanks to the team at Blockscience for early discussion and feedback on some ideas included in this piece, especially Zargham, Burrrata, David Sisson, Jeff Emmett, and Jessica Zartler, as well as the RMIT Blockchain Innovation Hub, especially my supervisors Ellie Rennie and Chris Berg, and Marta Poblet for reviewing. Thank you also to Rich Brown and Eugene Leventhal from Smart Contract Research Forum for encouraging me to share this investigation.