Social Decentralization is a Design Outcome, Not a Blockchain Feature (Part 1)
Why decentralization is about coordination, not crypto
Welcome to the fifth edition of Black Box. Last time, I explored the connection between early adopters and Cameo. This time, we’ll be analyzing DAOs using a game theoretic framework.
I studied economics in college at first because it was the closest thing that the liberal arts curriculum offered to business. Fortunately, I discovered game theory as a sophomore and became genuinely obsessed, eventually taking every industrial organization class that the department offered. However, I always thought game theory was just that — a theory. My perspective on this has changed since becoming a startup strategist and advisor. I have realized that although technology changes, the way people respond to incentives does not. This is why economic principles hold in spite of the centuries.
Yet this universal truth is often forgotten or outright ignored by evangelists drunk on the hype of some new technology. Nowhere is this more flagrant than in the web3 space—and before you call me a hater, I believe that crypto has valuable use cases. The reality is many of the hottest web3 startups are simply “successful web2 company, but on the blockchain”:
Is putting something on the blockchain the only way to decentralize it, and is decentralization really the solution to all of our problems? I am not qualified to comment on the technical side1, but I think the answer to both questions is no on the social side. Specifically, I believe DAOs exhibit the same community behaviors as their non-crypto counterparts, which means tokens tend to make decentralized coordination harder instead of easier.
Designing Decentralization
One reason I love game theory is many of its results are counterintuitive2. As I conducted research for this essay, I realized social decentralization is not only common in web2 (i.e., off the blockchain), but also inversely correlated with how structured the group in question is. This was counterintuitive to me since more structure mean less freedom, which is the opposite of the libertarianism characteristic of web3-style decentralization. How does this circle square?
First, a bit of theory (no pun intended). Traditional game theory focuses on the players and finding the strategies that optimizes their payoffs in a game. But there is a field within game theory that does the reverse. In mechanism design, the objective is to structure the payoffs, or design a mechanism, such that the strategies that players use to maximize their payoffs result in the intended outcome. (A set of such payoffs is called a solution mechanism.)
Mechanism design may seem nefarious, but many economists are studying it to develop more effective policies. The most well-known example of this is Nobel Prize laureate Elinor Ostrom’s work on the tragedy of the commons. She described eight design principles for regulating the use of common pool resources such as water. I think they can be adapted to web2 social platforms and their communities as guidelines for decentralized coordination:
DP1. Social groups should have clearly defined boundaries. Most group boundaries are based on interest. This is generally enough — there is a reason people follow some Twitch streamers but not others.
DP2. Rules should fit local circumstances. Different Facebook groups and subreddits have different rules to fit their specific needs.
DP3. Rules should be established through collective decision-making. Intuitively, people are more likely to follow rules that they helped make; this is a form of internal legitimacy. In practice, this design principle manifests as a set of “starter” rules decided by the group creator that members indirectly change over time (e.g., through community feedback). Note that rules can be explicit or implicit (as in group chats), and that a community may have both.
Game theoretically, this design principle is interesting because traditional mechanism design assumes a single designer. While simpler analytically, a single designer is harder to implement than multiple designers because it requires complete knowledge of all players’ preferences3. Letting members participate as designers allows this information to remain localized. It also enables players to propose rules that align with their true preferences — a condition that any solution mechanism must satisfy because of a theorem known as the Revelation Principle4.
DP4. Rules should be enforced by monitors. Most social platforms have moderators or administrators.
DP5. Sanctions for violators should be graduated. This design principle assumes that most violators do not break the rules intentionally. For example, many Slack communities prohibit posting a message in multiple channels, but violators are more likely to get a DM reminding them of this than kicked out of the group. (Of course, repeat violators should be punished or expelled.)
DP6. Conflict resolution should be cheap and easily accessible. This is why most social platforms have moderators or administrators.
DP7. Rules should be recognized by higher authorities. This is a form of external legitimacy. It is why social platforms respect community decisions on violators even if the offending behavior is not prohibited by platform policies.
DP8. Social groups work best when nested in larger networks. Social platforms help online communities by setting baseline rules and facilitating conflict resolution for more serious issues.
Decentralized coordination is essential because a social group is defined by its purpose, and the group must carry out that purpose to have meaning and to attract and retain members. Groups with purposes as “frivolous” as sharing pictures of cute dogs have extensive rules that require some coordination to enforce. Even members of groups without explicit rules implicitly coordinate to encourage behavior in line with their purposes, discourage behavior that is not, and maintain civility.
To that point, there is a spectrum of social decentralization. Many groups do not have and do not need formal conflict resolution systems. But generally, more structure is required for more decentralization. Otherwise, the group is at risk of either failing to carry its purpose and dissolving or defaulting to its administrators’ decisions when there are disagreements.
Different web era, same social dynamics
Ostrom’s design principles can be grouped into stages of structure that stack to enable more social decentralization in web2 groups:
S0. A clear group purpose. This corresponds to DP1. Some social platforms cannot host decentralized communities because their structures make it hard for groups with purpose to form. The chaotic shouting match that is Twitter is such an example (although it is trying to change with Communities).
S1. Visible status within the group based on carrying out the purpose. This does not correspond to any design principle, but it is critical nonetheless. Recall that mechanism design works by manipulating payoffs. Money is not an appropriate payoff in most online communities, so they use social capital to incentivize coordination instead. This manifests as some kind of visible status that can be tiered, quantified (e.g., Reddit Karma), or both.
S2. Group bylaws that establish desirable and acceptable behavior, and a way to acquire status by enforcing these bylaws. The first part of this stage corresponds to DP2 and DP3, while the latter part corresponds to DP4, DP5, and DP6. Some groups align status and enforcement by granting privileges or administrative rights to members with sufficiently high status. For example, the Huffington Post used to award top comment monitors a moderator badge which would (among other things) make their comments featured.
S3. A formal conflict resolution system. This corresponds to DP7 and DP8. Few communities are large and complex enough to have a formal conflict resolution system; most rely instead on their host social platforms to mediate when serious issues arise. Wikipedia is one of the few exceptions, as I will get to shortly.
What makes these stages interesting is they also reflect how DAOs and web3 communities grow. pet3rpan, an investor at crypto fund 1kx, describes the process as a three-step funnel:
F1. Attract intrinsically motivated members. As pet3rpan notes, these are the type of people who will form strong communities. This is analogous to S0.
F2. Get members engaged. This is analogous to S1, just with tokens added to the payoffs. (Anyone familiar with the web3 space knows status is very much still part of the payoffs).
F3. Align members with community success. This is analogous to S2, but again with tokens.
If web2 social groups look like decentralized communities, enforce rules like decentralized communities, and grow like decentralized communities, then they probably are decentralized communities. But would tokens make them more decentralized? Put another way, are web3 communities better because of tokens?
Before I lay out my view, I’d love to hear what you think. Tweet your thoughts @jwang_18. See you in two weeks!
The fact that you’re reading this means you’ve probably read it already, but I’ll refer to Moxie Marlinspike on this question anyway.
My favorite is the classic Bertrand model, which proves that a duopoly can be perfectly competitive if the companies compete only on quantity (i.e., sell commodity goods).
That economics should not make the standard assumption of complete knowledge is the basis of Friedrich Hayek's The Use of Knowledge in Society, a landmark essay in the field. The essay also inspired Wikipedia founder Jimmy Wales to make the platform socially decentralized, as I’ll discuss in part 2.
In essence, the Revelation Principle states that if there does not exist a mechanism that results in the desired outcome using true preferences, then there is no solution mechanism. The intuition is players would gain nothing from lying because the solution mechanism would effectively lie for them (e.g., through additions or transfers between players) as part of their maximum payoffs.