tldr: Recursive decentralization and autonomy is the key to creating trustworthy DAOs that "do what they say on the tin" and solve meaningful problems at scale.
It's a commonplace observation that our society faces a crisis of institutions. We no longer feel able to trust our governments, our media, many businesses, religious organizations, or technology platforms. We see them as unreliable servants of our interests, and careless masters of the public goods they are responsible for.
As society itself has become more complex, it is harder than ever to understand what some of these institutions even do, and to hold them accountable for their intentions or their actions. This results in what economists call a "low-trust society", in which our inability to trust institutions begins to erode our ability to engage in mutually beneficial cooperation in general. Suspicion of corruption, bias, ulterior motives or plain incompetence, becomes the default.
Some of this results from the effects of globalization and the internet. Many more of our interactions are with organizations that we only know through their digital presence, and while that makes it easy to access them, it also makes it harder to hold them accountable when things go wrong. Sometimes, much like the Wizard of Oz, there appears to be nobody behind the curtain in the first place. When our personal data is stolen from supposedly secure online services, when our pandemic preparedness plans turn out to be inoperable, when "transient" inflation turns out not to be quite so transient, when transparent falsehoods become campaign slogans, who can we hold to account?
In Making Policy for a Low-Trust World, Matt Yglesias argues that we can't restore trust simply by adding more layers of procedure. Additional procedures can serve to increase the very complexity that makes accountability impossible to establish. Instead, he argues for institutional simplicity, or the "does what it says on the tin" principle:
It’s easy for everyone, whether they agree with you or disagree with you, to understand what it is you say you are doing.
It’s easy for everyone to see whether or not you are, in fact, doing what you said you would do.
It’s easy for you and your team to meet the goal of doing the thing that you said you would do.
This is good advice.
This is also advice that the most effective web3 and crypto systems have tended to follow. The oldest and simplest example is Bitcoin itself. It is easy to understand what the Bitcoin system does - provide a global ledger for transactions in a supply-limited digital currency. It's easy to see if this is actually happening, whether by validation of the chain itself, or by audit of the source code. And it's easy to meet the goal because the incentives designed into the system and the culture of open source push the system in the right direction. Bitcoin does what it says on the tin, and even the people who don't like what's in the tin can agree on it.
Bitcoin achieves this in part by being autonomous - it is very difficult to shut down or subvert the system from outside - and decentralized - it is very difficult to corrupt or take over the system from within. These two properties, of autonomy and decentralization, form part of the model for DAOs: decentralized autonomous organizations. That third part - organization - refers to the fact that they must coordinate some useful activity.
If DAOs are going to provide alternatives for a wide range of social institutions, they must be able to pass Yglesias' three tests: to be able to explain what they're doing, make that verifiable, and succeed in doing it. As an outsider to the DAO, I want to know that what it is promising to me and how to be sure that it is keeping those promises. As an insider, I want to be able to know the DAO's purpose and to be able to use my skills and knowledge to pursue it.
There are some cases where this is straightforward: the organization's plans are simple, and the individuals within the organization can spontaneously act toward the achievement of these goals even without coordination. Organizations built around markets tend to work like this, with financial incentives coordinating activity. To paraphrase Adam Smith,"it is not from the benevolence of the miner that we expect our blocks to be filled, but from their regard to their own interest". With narrowly-defined roles and direct economic incentives, it's possible for many people to coordinate without requiring any explicit coordination.
Outside of these special cases, the problem is trickier. There are times when we want to do things that have not been done before, so we can't possibly rely on predefined roles or automatic economic incentives. Decisions will need to be taken, agreements will need to be made, and if the activity we wish to carry out is on a large scale, then many decisions and agreements will need to be taken and made. This is necessarily human-scale activity, involving people talking, planning, and deciding together. We want the DAO to act with integrity, meaning that it can seriously commit to its plans while coordinating the different perspectives and capabilities of its members.
Our belief is that decentralization and autonomy are not principles that apply only to the DAO as a whole, but to every level of activity within the DAO. For a DAO to act with integrity, the members of the DAO must also be autonomous, and decision-making must be recursively decentralized.
At the smallest level is an individual member of the DAO. In the absence of a central planner, the member has total autonomy: they can intend to work on whatever problem they believe to be important, in whichever way they think will be most effective. If resources are required, they can request them from the DAO's treasury, which will use a pre-agreed procedure such as voting to decide on whether to grant the resources. At this level, autonomy consists of being able to decide on one's intentions, and decentralization comes from the independence of the member's intent.
If the member cannot solve the entire problem, or fully exploit an opportunity, without assistance from others, then they must share their intent with someone else. If they can persuade someone else to help them, we can say that they have a "shared intent". Since it is shared directly by the two collaborators, it is autonomous with respect to outside influence. And since neither collaborator is commanding the other, they do not infringe each other's autonomy, and no central authority is created. At this level, autonomy consists of being able to decide who to collaborate with, and decentralization comes from the independence of this shared intent.
The challenge is now one of scaling beyond two collaborators. If the collaborators cannot achieve everything they need between themselves, then they must break their shared intent down into smaller units, some of which can be shared with other collaborators. To do this, they must persuade others to take on responsibility for those parts. The relationship between the original intent and these new parts is one of dependency: the original intent now depends on the success of further intents for its own success. Since this process is recursive, we can break very large objectives down into smaller ones, thereby creating a network of intents. At this level, autonomy consists of being able to decide how one's intents depend on intents being undertaken by others, and decentralization comes from the fact that this network is produced by peer-to-peer agreement and collaboration.
At the level of the whole DAO, it is necessary to decide which intents will be granted resources. Again, these must be decisions taken by autonomous agents without central control, participating in a governance system such as voting. A common difficulty in such situations is the risk of centralization by granting a large amount of resources to a single individual for the purpose of pursuing a particular project or objective. In order to avoid this, a DAO might prefer to make smaller grants. However, with a recursive network of shared intents already in place, it is possible to fund the network rather than an individual person. Each collaborator would receive the resources they require. Thus it is possible for the DAO to pursue its aims by voting to distribute resources across networks of its own members, for the purpose of achieving objectives that are both individually and collectively meaningful, without relying on central planning.
In our sketch above, we show how it is possible to construct a DAO by tying fine-grained resource allocation to networks of individual and shared intents freely chosen by the members, with the scalability to tackle projects of any size and duration. Or, respectively, a DAO that is truly decentralized, autonomous, and organized.
Such a DAO shares the "does what it says on the tin" properties of Bitcoin and other effective open institutions. Because the members of the DAO make their own plans, and decide together on what activities to fund, it is easy to understand what the DAO say it is doing, whether it has done it, and for each member to understand their personal responsibility.
In principle, it is just as possible to audit such an organization's network of intents as it would be to audit the source code of Bitcoin, or of a smart contract. It might require some technical skill to do so, but the major challenge lies in making the organization's structure legible in the first place. By examining what the organization says it's doing, and how this breaks down into individual actions, we can have much greater trust that the organization will keep its promises and serve our interests.