Ethereum as Collective Intelligence
A framework for decision-making and governance
[NOTE: Working draft. Latest Update July 1, 2018. Feedback welcomed.]
Governance is a significant challenge for all public blockchain networks, and for the Ethereum community, it is one that continues to stimulate debate and discussion. To be sure, some progress has been made. We have a general idea of what governance means in this context and given that we’re talking about decisions made by and for a decentralized network of people, there is particular emphasis on coordinating participation around and managing the legitimacy of — the decision-making process.
Two general approaches to blockchain governance have been defined and it’s clear which model Ethereum favors (also here, here). Nevertheless, experiments with on-chain governance continue (i.e., Decred, Dfinity, Polkadot). Others start off-chain with the intention to shift to on-chain (i.e., 0x).
Recently, an important first step was taken in an attempt to address governance challenges head-on and also make explicit, the values held by the Ethereum community (see here for a nice summary). And yet, this meeting also highlighted where more work is needed.
Current Challenges
Clarification of decision-making process: EIP editorial judgement
The Ethereum Improvement Proposal (EIP) process has come a long way since its inception and is continually being streamlined and improved (both Hudson Jameson and Nick Johnson provide an overview of EIP’s progress here; historical context here). The main challenges at this point appear to be judging community consensus and further clarifying criteria for accepting changes. According to Jameson, current criteria for accepting changes are A) adoption of proposal in multiple clients and B) wide deployment. While criteria for accepting changes technically exist, it appears that editors often have to make a judgement calls and this especially becomes tricky for contentious issues, particularly when editorial judgment is dependent on community consensus (Johnson explains here).
Equally important is communicating all this to the community at large to address these, currently, unanswered questions.
Gauging community consensus: Better signaling systems
The challenges EIP editors face in exercising judgement may be ameliorated with better systems to gauge community consensus (or lack thereof), particularly around contentious issues. Jameson discussed this issue in his recent talk at Zcon0.
More participation in decision-making
Another challenge is the lack of participation in the decision-making process and there appear to be several reasons for this. First, there could be a lack of expertise, or perhaps a hesitancy to take on the responsibility of being an EIP editor (see here for recent discussion on potential fiduciary responsibilities of public blockchain software developers). Lack of legal clarity certainly does not help.
Values & Constitution
Many have pointed out a need for the Ethereum community to outline its values to guide decision-making (e.g., mission statement, constitution). Yet, I find interesting this observation by Adriana Z. Robertson, a law professor, who was present at EIP0 Summit:
The last thing I took away was a genuine belief that, despite repeated assertions to the contrary, Ethereum already has a constitution (source).
Collective Intelligence: A framework
The work of governance centers around decision-making. Having a clear process around decisions; participating in decision-making; having clear signals from the community for certain decisions; communicating the process and outcomes of decisions.
To be deliberate in our approach, it might be helpful to consult a growing body of literature on collective intelligence that has provided an analysis of decision-making in online groups.
Here are some definitions of collective intelligence to illustrate how it might be relevant for the Ethereum community:
- a form of universally distributed intelligence, constantly enhanced, coordinated in real time, and resulting in the effective mobilization of skills (Levy, 1994)
- a group of human beings [carrying] out a task as if the group, itself, were a coherent, intelligent organism working with one mind, rather than a collection of independent agents (Smith, 1994)
- shared or group intelligence that emerges from the collaboration, collective efforts, and competition of many individuals and appears in consensus decision making. The term appears in sociobiology, political science and in context of mass peer review and crowdsourcing applications (Wikipedia, 2018)
Collective intelligence is when large, organized groups of people, often online, work together to achieve an objective, such as:
- To keep Ethereum The Best It Can Technically Be. (source)
- To bring together various Ethererum teams who play major roles in determining the direction of the protocol. (source)
- [To] build a more globally accessible, more free and more trustworthy Internet. (source)
- [To nurture] the vision of the serverless internet, the decentralised web. (source)
Perhaps the most famous example that does not rely on cryptographic methods is Wikipedia. Other examples can be found here.
This framework starts by posing four basic questions, the answers to which are building blocks for any system of collective action. They can be used to design traditional organizations or, for our purposes, online communities that form around decentralized public infrastructure (for further information, see here).
At the very least, they be used to understand how our current decision-making processes, and by extension, governance system, works. The questions are:
- What is being done?
- Who is doing it?
- Why are they doing it?
- How is it being done?
What is being done?
The two primary building blocks here are Create and Decide. To simplify the EIP workflow, the major tasks either involve creating a proposed change or deciding whether that change gets passed.
Let’s take a relatively simple example: an Ethereum Improvement Proposal (EIP) for a non-protocol change. Then we’ll follow-up with a more complex example later.
What is the task?
EIP authors will create and propose a change. Their job is to vet the idea to make sure it hasn’t been covered in prior research as well as gauge applicability to the entire community by scanning through public forums (i.e., subreddits, issues repository, Gitter channels, Ethereum Magicians).
Then EIP editors will read, check if a submitted EIP is sound and ready, provide feedback and provide instructions for next steps. Moreover, EIP editors will correct for structure, grammar, spelling, or markup mistakes (see here). Without passing judgement, focusing on administrative/editorial aspects, the editor decides if an EIP gets merged into draft.
Who is doing it?
The primary building blocks here are Hierarchy and Crowd.
Hierarchy is when a particular person or group perform the task, while crowd is when anyone in the community can performs the task.
Similar to other prominent open source projects, EIP authorship is performed by anyone from the crowd, while editorial duties are performed by a hierarchy — there are currently six editors, self-appointed or chosen based on expertise, who are performing these duties. The All Devs Call, a forum to discuss EIPs that have been merged into draft, is also invite-only group with some decision-making influence.
In traditional organizations, hierarchy is the default option that is appropriate when only few people have the necessary skills/expertise or when there are no mechanisms in place to facilitate crowd work. The trade-off here is efficiency (hierarchy) for decentralization (crowd).
Jameson has pointed out that he has ‘too much power’ (referencing gate keeping responsibilities in the All Devs Call) and there is a general desire to streamline the process (make it more inclusive) as well as spread out the responsibilities of EIP editors (as Ethereum continues to grow, having just six editors is probably not sustainable).
Why are they doing it?
There are several broad, high-level motivations that compel people to participate in collective intelligence systems including: intrinsic motivation (pure enjoyment of writing code that the community will use), contributing to a large cause (a more free and trustworthy internet), socializing with peers (conferences), recognition from peers, and monetary reasons.
Traditional organization have relied on monetary compensation. Open-source movements have appealed to other motivations (intrinsic, contribution to larger cause, social recognition). Ethereum is more aligned with the latter, but the point is not to debate the merits of one over the other, but to see if different building blocks can be combined in ways that serve the overall purpose.
For example, monetary mechanisms may be needed for more difficult roles (i.e., EIP editors) or, if the goal is to encourage more participation in technical discussions, a system that fosters social recognition / social reputation may help. Mechanisms we’re familiar with include twitter ‘likes’, ‘retweets’, ‘followers’, other ‘emojis’, ‘upvotes’ etc.; Wikipedia has its own system.
How is it being done?
This final question, ‘How is it being done’, is linked to the first question, ‘What is being done’ (i.e., Create or Decide). For things being created, the response to ‘How’ is Collection or Collaboration. Collection are tasks that individuals can do independently of others, collaboration are tasks that are interdependent in nature.
It appears that EIPs are submitted independently of each other. For the most part, EIPs have one main author. These would fall under Collection tasks.
However, perhaps it’s possible that more complex EIPs with interdependencies could also be submitted. There are also some EIPs that have more than one author so these cases would fall under Collaboration tasks.
Finally, for ‘Decide’ tasks, there are essentially two ‘How’ responses: Group Decision and Individual Decision. The Group Decisions are when inputs from the crowd are taken, then assembled to generate a decision that holds for the group. Examples of Group Decision include: Voting, Consensus, Averaging and Prediction Markets.
For Individual Decisions, certain members make decisions (e.g., EIP editors), and while they would like decisions to be informed by the crowd, technically do not require actual crowd input. Examples of Individual Decisions include: Markets and Social Networks.
With the current EIP process, editors make (individual) decisions (to merge pull requests into drafts) after they’ve made an attempt to gauge community consensus. Individuals decisions from EIP editors are informed by gauging community sentiment through various Social Networks channels (i.e., Twitter, Reddit, Facebook Groups).
NOTE: The following sources have been consulted in an attempt to capture the EIP workflows as accurately as possible: 1. Thread on ‘Reviewing the current EIP Process’, 2. EIP Purpose and Guidelines, 3. A Review of the Current Ethereum Governance Processes (EIPs & All Devs Call) recorded at EIP0
Also, Dan Finlay, made a helpful flow chart:
Protocol Changes
To recap, the previous section discusses a non-protocol change EIP using a collective intelligence framework to understand decision-making in an EIP workflow, from the perspective of both an author and an editor.
It may be helpful to have an additional example.
For more complexity (and realism), we’ll use the example of an EIP attempting to make protocol changes. Because EIP for protocol changes can conceivably lead to contentious hard forks, they get at governance challenges the community is currently wrestling with.
Going through the four questions, I’ve summarized the process in this table:
What is being done: Create
A couple of observations.
First, the process of pushing an EIP for protocol change involves a lot of moving parts, so hopefully breaking the task into chunks allows us to (more) objectively examine our decision-making process.
Focusing on the (Create) rows, we can see that creating, submitting and editing an EIP draft can be done by anyone from the community (Crowd), but it requires that person championing his own submission, by vetting the idea. At the moment, most EIP submissions are self-contained, independent of other submissions (Collection). One question to ponder is whether it is possible that EIPs can be a collaborative task where submissions involve interdependencies and multi-party collaboration. What abilities would be required to coordinate such EIPs? Would this encourage more (or less) EIPs? Would this encourage or hinder innovation?
Next up are a multitude of players involved in deciding whether the EIP gets merged into draft, recommended for implementation, implemented on a client and, ultimately, adopted by end-users.
What is being done: Decide
When looking at the (Decide) rows, there is an opportunity to clarify what exactly we mean by consensus. As Jameson mentioned above, in his role as EIP editor, he has to scan through various social media channels to gauge community consensus. Yet, it appears consensus must also be achieved within the All Core Devs Call as well as the Client Dev Team.
Each level of consensus, in the wider community and within a more immediate small group (All Core Devs Call or Client Dev Team), has its own set of challenges. In the former, we might need to build tools that aggregate, analyze, interpret and display online community sentiment. I’d argue we also need online platforms for constructive debate and discussion. The latter becomes more challenging as the group gets larger.
When distinguishing consensus in different areas of the system, it seems entirely possible to improve the process of gauging community consensus separately from the social consensus required of core developers throughout the EIP workflow. As Micah Zoltu points out about the current system:
This process allows people who have spent a lot of time thinking about an issue the least amount of inertia pushing against them and those who have spent less time more inertia to fight against.
There’s been discussion of various voting techniques and even prediction markets as a means aggregating community consensus and achieving more decentralization. It would be helpful to clarify, at which stage in the decision workflow should there be a shift from consensus to voting? Does this imply that decision-making power transfers from EIP editors / core developers to the crowd? That may be a plus for decentralization, but what are the trade-offs?
If we are to experiment with governance, perhaps it’s prudent to experiment in stages. One of the arguments against voting is a lack of participation and, perhaps worse, uninformed participation. Introducing changes at specific stages in the EIP workflow could encourage participation and over time, as more people get educated about Ethereum, more meaningful participation could be cultivated.
For example, having the crowd get involved in earlier stages of the EIP workflow — like whether to merge into draft — would allow for risk-minimal experimentation with group decision-making before moving onto higher stakes level like recommending changes to a client.
Finally, using a collective intelligence lens allows us to continue to be honest about where power resides in the decision-making process.
Parting Thoughts
“Blockchain governance” has been conceptually stretched to the point of near meaninglessness. — @nic__carter
I’ve tried to dig into a specific aspect of governance — decision-making — and highlight how the Ethereum can more effectively act as a collective, borrowing ideas from the multi-disciplinary field of Collective Intelligence. Hopefully this framework demonstrates how we can begin to systematically begin to address the challenges of governance.
To be sure, governance challenges will require multi-disciplinary efforts. I’ve written here, on the potential contributions from social and organizational psychology, and next, I’m hoping to draw in insights from human-computer interaction as well as artificial intelligence. Here are some resources I think could help us better gauge community consensus:
- Reflect (better discussion through restatement)
- Consider It (a web forum that elevates your community’s opinion)
- Opinion Space (help communities generate and exchange ideas about important issues)
- Widescope (a social platform for serious conversations on the web)
Stay tuned!
