Consensus in Tech

Photo of Sue Gardner (c) Victoria Will for the Wikimedia Foundation.
Photo of Sue Gardner (c) Victoria Will for the Wikimedia Foundation.

An Interview with Wikipedia’s Sue Gardner

From 2007 until 2014 Sue Gardner served as the executive director of the Wikimedia Foundation, the organization that operates Wikipedia. Today she is an advisor to the Wikimedia Foundation and on the boards of various internet-related nonprofits. She currently works with the privacy nonprofit The Tor Project and with First Look Media, the news project founded by Pierre Omidyar that publishes The Intercept.


Friends Journal: How did Wikimedia come to develop a culture of consensus decision making?

Wikipedia is part of the free software movement, which is fundamentally about collaboration and openness. Not all free software projects are consensus-oriented, but they all have a general bias in that direction, and Wikipedia’s no exception. Its founder, Jimmy Wales, designed it to be participatory and hierarchy-free, and of course that led directly to people self-organizing. I would say its consensus-oriented practices developed naturally out of that.

I should also probably say that anecdotally, I’ve always felt that Quakers are highly represented on Wikipedia, relative to their presence in the general population. I have met a half-dozen Quaker Wikipedians, and Quaker influence on the project is sometimes explicitly acknowledged in our policies and discussions. It wouldn’t surprise me if Wikipedia’s practices were heavily influenced, early on, by the participation of Quaker Wikipedians.


Friends Journal: Can you give an example of how consensus is used at Wikimedia? Is it used with staff or board or among the volunteer editors?

Sure. So, for example, when Chelsea Manning changed her name from Bradley and publicly announced that she is a woman, a Wikipedia editor immediately changed her article to reflect her new name and her gender. Another editor disagreed with that decision and reverted it. That kicked off a long discussion on the article talk page where editors debated how the article should characterize Chelsea Manning’s name and gender. Debates like that are super-common on Wikipedia, and they are ordinarily pretty structured. Someone formally opens and frames the discussion; everyone is invited to give input over several days or weeks; people modify their positions as they’re influenced by things other people have said, and ultimately a new, previously uninvolved editor comes along and summarizes what’s been said and gives a sense of  the resolution or outcome. If that summary is felt to be a reasonable reflection of what happened, then that decision is what ends up happening.

The board of the Wikimedia Foundation does not operate according to a pure consensus process. It has a normal voting model—majority-rule—and it’s not unusual for unanimity to be lacking. Typically people will vote their conscience, rather than pretending there is consensus when there isn’t. But even in that context, the Wikimedia board of trustees will aim to achieve as much agreement as possible, and will only very reluctantly move things forward in the face of sustained internal disagreement. It was  common when I was executive director for the board to do things like take a straw vote to surface people’s initial views and then discuss how to reach a compromise that everyone could support. We didn’t always achieve that, but there was always a good faith attempt.


Friends Journal: How do you practice consensus with an online volunteer editor pool spread across the globe, all in different time zones?

It actually works really beautifully. I think the asynchronous, distributed nature of the discussion works in Wikipedia’s favour: it cools things down and results in a better, more thoughtful conversation. That said, there are aspects to the Wikipedia experience that work against good discussion. Sometimes people are writing in a second or third language. Text provides only limited signals: you miss facial expressions, tone of voice, and body language. You ordinarily don’t know anything about the people you’re talking with. And all that contributes to our human tendency to assume the worst. For example, if someone is extremely succinct, you might experience that as brusque or rude, whereas maybe they are just typing on a phone, or were raised in a culture that’s super-direct, or are sick or experiencing physical pain. So there are upsides and downsides for Wikipedia.


Friends Journal: Quakers have used a form of consensus for hundreds of years, and Quaker-inspired activists codified a secular process in the 1970s. How did you come to learn of the Quaker influence in consensus practice?

When I came to Wikipedia, I didn’t have any personal experience with consensus-oriented process, but I knew of course that it predated us. And so I started looking around for models. There was remarkably little that I could find: the Polder system and a few others. Then I stumbled across Michael J. Sheeran’s Beyond Majority Rule, and found it super-useful. So I started reading other books on Quaker practices, and adopting some of them for my work at Wikipedia.


Friends Journal: In 2010, you went into a Quaker process reading binge and then attended a clerking workshop at the Ben Lomond Quaker Center led by Elizabeth Boardman and Eric Moon (both well-known to Friends Journal readers!). What similarities and differences did you find with Wiki culture? Did you bring back any new concepts to Wikimedia?

Yes! What was funny was that a lot of Quaker practices seemed to be already in effect at Wikipedia. They weren’t necessarily adapted directly from the Quakers, but I think Wikipedians to some degree arrived at the same place by having similar goals and doing similar thinking. So, for example, in our clerking there was a long-standing Wikipedian practice of summarizing. There would be a long discussion “on-wiki” (using the wiki software), and at some point somebody would come along and attempt to summarize the major arguments and reflect back to the group a sense of where we were at. That is very similar to Quaker clerking. For example, it was important that the summarizing person try to be unbiased: that they try to neutrally reflect the arguments rather than advocating for a particular position. They needed to be credible and trusted by the group.

Phoebe Ayers, a board member, and I went down to the Ben Lomond Quaker Center and participated in a weekend workshop on clerking. That was awesome! And afterwards, we tried to implement some of the basic principles and practices inside the Wikimedia Foundation board discussions. Phoebe and I worked really hard to create a safe, trusting environment, because our observation of the Quaker processes told us that doing that would enable our group to make better decisions, as well as decisions that “stuck.” We also talked about practices that we had previously used informally, such as encouraging quiet people to speak and talkative people to listen. We made that an explicit part of our board meetings, and it was incredibly useful and effective.


Friends Journal: You said that you have thought that the process has a bias towards the status quo at Wikimedia? Have you ever seen that challenged?

Consensus decision making is slow. The benefits outweigh the costs in my experience: the decisions are better; people are more committed to them; and people are happier and more engaged overall. But it is slow, and yes, there’s a bias towards the status quo.

A Quaker once told me a story about a person in his community who was blocking an important decision that everyone else agreed to. He told me that the person blocking the decision was elderly, and so they decided to just park the decision until after that person had died. I was horrified by that but simultaneously also admired it. (And I think that community felt the same way.) Consensus decision making is founded, I think, in a fundamental agreement that when people feel very, very strongly about something, you don’t force it. You don’t push ahead without them. There is a cost to that, but you can’t  have a consensus process without that cost.

I have seen the status quo successfully challenged in big ways at Wikipedia. There are a half-dozen times, that I’m aware of, that somebody led a dramatic change inside Wikipedia. The people who did that successfully were always ones who were already widely respected in the Wikipedia community. They had put in a lot of hours; they knew the community very well; and they had built up a lot of trust. In those circumstances, the Wikipedia community sometimes has been willing to park objections and extend sufficient trust to allow something to happen that it might have been suspicious of in different circumstances. Jimmy Wales was able to do that, and Erik Moeller also. There are maybe a handful of others, but it is very rare. When you have that ability, you need to exercise it sparingly or you risk losing it.


Friends Journal: You’ve written that you don’t think that consensus decision making is very adaptable for mass collaboration—to use the tech phrase—that it’s not scalable. Do you think there are viable models for consensus on a larger scale?

I think it’s really hard. It is just much easier to achieve consensus when people have been able to build personal, individual trust among themselves. That happens slowly over time, and it’s usually the result of a history of personal interactions. I think the way to scale consensus decision making is by putting in place standards and practices that everybody agrees to, so there is a bedrock of agreement on process before the work starts. The Occupy movement did a good job with some practices, like the progressive stack practices. But with larger groups, consensus decision making is harder and more fragile.


Friends Journal: The early Quaker movement tossed aside a lot of what was considered normal church organization—ministers, creeds, sacraments—and then went on to build a kind of collective knowledge of best practices that became normative Quaker behavior, a process that feels almost wiki-like. As you learned about Friends, did you see any surprising cross-over?

Yes, lots, and it probably evolved in a similar way. Wikipedia, of course, threw away a lot of received or conventional wisdom about how to build an encyclopedia: that you needed to be an expert; that you needed a vetting process pre-publication; that being wrong was a mortal sin; that somebody knew the truth, and others did not. But then over time Wikipedia also built up its own encyclopedia-making practices, conventions, norms, and best practices that I would say were better and more robust than the ones that preceded them. But that isn’t unambiguously good, right? It is good because it reflects expertise that was hard-earned over time, but the community becomes less accessible to outsiders. In general, it’s important for teams or groups to have counter-pressures to aid development and avoid bureaucracy, which is a kind of priesthood with insiders and outsiders. It’s hard to institute that kind of counter-pressure, but without it you risk becoming rigid.


Friends Journal: Do you know of other media or technology companies using consensus?

In the Bay Area and perhaps in software development generally, there is a lot of interest in new ways of organizing work. Knowledge work is different: it isn’t widget-making, and so the industrial manufacturing processes don’t work. Free software communities often use consensus or semi-consensus process, although often they combine it with what’s called kind of jokingly a “benevolent dictator.” This is somebody very trusted, often the founder, who is given social or cultural permission to break deadlocks and resolve disagreements unilaterally. The benevolent dictator is a kind of safety valve for the culture, because he or she enables stuff to move forward when agreement is hard to reach. Agile methodology has some aspects of consensus decision making to it. And I know that a few companies like Medium (which is both a tech company and a media company) are using holocracy, and my understanding is that some implementations of holocracy have similarities to consensus-based practices. Ev Williams, the founder of Medium, has written and said some interesting stuff about how he uses holocracy and why he likes it.


Friends Journal: What do you think Quakers can learn from Wiki and/or media or tech culture?

I don’t feel knowledgeable enough about Quaker culture to presume to say. And there are aspects of Quaker culture that are clearly superior. For example, the Quakers famously value the contributions and expertise of older women. That is awesome because the rest of the culture—very much including tech culture—typically over-values the contributions of young men at the expense of the contributions of women and older people. Also tech culture, including Wikipedia culture, is not always great at collaborating well. Bluntness is highly valued; being right is highly valued; it is sometimes “fighty.” Tech culture could learn about trust-building and healthy communication from the Quakers.

One thing that tech culture has that the Quakers, I think, do not have is an agreement that it’s okay to break things, to mess up, and to make mistakes. Quaker culture strikes me as fairly risk-averse and cautious. In Silicon Valley, there’s an ideology that it’s totally okay to break things. Sometimes that’s just a cover for being reckless or flouting consumer protections or whatever. But having social permission to mess around and iterate-towards-better is a great strength of tech culture.


Friends Journal: And the big question: do you ever visit or have the urge to visit a Quaker meeting for worship?

I have often considered going to a San Francisco meeting! I’m an atheist, but I love the Quakers I’ve met and really enjoyed the few days I spent at the Ben Lomond Center. So one day I may turn up at a meeting.


Friends Journal: This is a question from one of our readers via Twitter: The tech industry seems to thrive on speed and impulse. Can discernment really be given more than lip service in that context?

A while ago I read Rosabeth Moss Kanter’s book Men and Women of the Corporation, which I think was first published in the seventies. She talks about external environments in organizations that are super-changeable, disruptive, and fast-moving creating organizations that are internally homogenous. Everybody is the same: same age, same race, same gender, same beliefs, same leisure activities. This strikes me as totally understandable, because you can’t operate in total chaos. If the external environment is tremendously chaotic and it’s hard to know, for example, what your users may want tomorrow, then it helps to be able to count on understanding the person at the next desk.

So I think that’s the situation in Silicon Valley. It’s especially important for start-ups to be agile in order to flourish, and that is partly the reason for their internal similarity: so many white and Asian men; so many from the Ivy League; so many shared orthodoxies, myths, and ideologies. Everybody wears a hoodie and reads Hacker News. In the long run, that kind of homogeneity is terrible for business, precisely because it excludes so many useful viewpoints, experiences, and skills. It’s also ethically problematic because it excludes people who should have the opportunity to  contribute. But in the short term, it’s an adaptation that’s regrettable but also understandable and successful.

So this is a very long way to answer your question. The answer is in the context in which Silicon Valley operates where there is very little interest in consensus-type processes. To some degree there is an interest in inclusion but not consensus. In addition to the external environment, there is also an inexperience among Silicon Valley managers, many of whom aren’t yet very sophisticated or thoughtful about this stuff.


Friends Journal: Here’s another reader question: Have you seen any ways in which this process has been agile and a helpful space for minority voices?

Oh, what a great question! I was going to say that the consensus process is great for minority voices, but then I thought about it for a second and realized that’s totally untrue. Consensus process, unless it’s very carefully done, can be terrible for marginalized or non-majority people, because “everybody feels” like X is correct, or the “sense of the room” is that X is correct, or whatever. Certainly we have had that experience at Wikipedia, where a group of majority culture people have, through consensus process, happily arrived at an incorrect answer, because particular voices went unheard. One example is Wikipedia’s deciding to name an article about the river known as the Ganges. Inside India and throughout Asia, that river is best-known as Ganga. But the English Wikipedia article calls it Ganges, presumably because that’s what British and North American people call it, and there are more British and North American Wikipedians than Indian ones. That is an example of minority voices going unheard inside a consensus process. I would argue that the root problem there is the under-representation of certain voices, which leads to systemic bias and low-quality decisions. But the systemic bias could have been mitigated if the voices of the under-represented people had been deliberately amplified in some way. That’s something the Quakers get right more often than Wikipedia does, and it’s something the Occupy movement did very well.

The consensus process isn’t agile, and I don’t think it is the right tool for every decision. It is a hammer, and sometimes you need a screwdriver. So, for example, if you’re making a decision that is heavily dependent on a particular kind of expertise and knowledge, that shouldn’t be a consensus decision.  Examples of questions better handled through a consensus process would be whether Wikipedia should comply with Chinese government censorship in order to make the site more easily available to large numbers of people in China and whether Wikipedia should accept advertising so it can afford to hire large numbers of engineers to make the site really awesome. Those are important values-based decisions that shouldn’t be made by a small group. In a community like Wikipedia, in my opinion, they require consensus. Questions about the visual design of the mobile app or where to open the new data center are expertise-driven decisions, and in my opinion shouldn’t be made by consensus that includes people who don’t have that expertise. Those decisions should be made by experts. So, a true, full-on consensus decision making process is not the right tool for every decision made. And it’s important to be explicit about what process you’re using  and why, because otherwise you risk confounding people’s expectations and damaging trust.

2 thoughts on “Consensus in Tech

  1. Thanks so much for this article. I am a techie, trained in Microsoft speak. In most tech companies the staff are so competitive that they are ugly to work with. So glad to hear that at least one tech working environment has created a quasi-safe place that doesn’t destroy its goals to enhance the world. I use Wikipedia a lot and have donated to assist in their mission.
    Nyla Dartt

  2. I know another open source project that practices consensus decision-making: Kubuntu.

    Kubuntu is an operating system based on Linux and derived from the popular Ubuntu system. Kubuntu was started by Jonathan Riddell, a Quaker in Scotland.

    I haven’t been involved in 8 years, but what I remember is that Kubuntu had a board which made decisions by consensus in public meetings held on internet relay chat (IRC).

Leave a Reply

Your email address will not be published. Required fields are marked *

Maximum of 400 words or 2000 characters.

Comments on may be used in the Forum of the print magazine and may be edited for length and clarity.