Slashdot Mirror

Today, read the continuation of Larry Sanger's account of the early history of Nupedia and Wikipedia (below), in which Sanger talks about the difficulties of governance in a large, free-wheeling project, some final attempts to save Nupedia, and how he came to resign from the organization. (And if you missed it, you might want to start with yesterday's installment.)

Contents:
Why Wikipedia started working
A series of controversies
The governance challenge
My resignation and final few months with the project
Some final attempts to save Nupedia
Conclusions

Why Wikipedia started working

This is a good place to explain why Wikipedia actually got started and why it worked (and still does work, at least as well as it does). The explanation involves a combination of quite a few factors, some borrowed from the open source movement, some borrowed from wiki software and culture, and some more idiosyncratic:

1. Open content license. We promised contributors that their work would always remain free for others to read. This, as is well known, motivates people to work for the good of the world--and for the many people who would like to teach the whole world, that's a pretty strong motivation.
2. Focus on the encyclopedia. We said that we were creating an encyclopedia, not a dictionary, etc., and we encouraged people to stick to creating the encyclopedia and not use the project as a debate forum.
3. Openness. Anyone could contribute. Everyone was specifically made to feel welcome. (E.g., we encouraged the habit of writing on new contributors' user pages, "Welcome to Wikipedia!" etc.) There was no sense that someone would be turned away for not being bright enough, or not being a good enough writer, or whatever.
4. Ease of editing. Wikis are pretty easy for most people to figure out. In other collaborative systems (like Nupedia), you have to learn all about the system first. Wikipedia had an almost flat learning curve.
5. Collaborate radically; don't sign articles. Radical collaboration, in which (in principle) anyone can edit any part of anyone else's work, is one of the great innovations of the open source software movement. On Wikipedia, radical collaboration made it possible for work to move forward on all fronts at the same time, to avoid the big bottleneck that is the individual author, and to burnish articles on popular topics to a fine luster.
6. Offer unedited, unapproved content for further development. This is required if one wishes to collaborate radically. We encouraged putting up their unfinished drafts--as long as they were at least roughly correct--with the idea that they can only improve if there are others collaborating. This is a classic principle of open source software. It helped get Wikipedia started and helped keep it moving. This is why so many original drafts of Wikipedia articles were basically garbage (no offense to anyone--some of my own drafts were sometimes garbage), and also why it is surprising to the uninitiated that many articles have turned out very well indeed.
7. Neutrality. A firm neutrality policy made it possible for people of widely divergent opinions to work together, without constantly fighting. It's a way to keep the peace.
8. Start with a core of good people. I think it was essential that we began the project with a core group of intelligent good writers who understood what an encyclopedia should look like, and who were basically decent human beings.
9. Enjoy the Google effect. We had little to do with this, but had Google not sent us an increasing amount of traffic each time they spidered the growing website, we would not have grown nearly as fast as we did. (See below.)

That's pretty much it. The focus on the encyclopedia provided the task and the open content license provided a natural motivation: people work hard if they believe they are teaching the world stuff. Openness and ease of editing made it easy for new people to join in and get to work. Collaboration helped move work forward quickly and efficiently, and posting unedited drafts made collaboration possible. The fact that we started with a core of good people from Nupedia meant that the project could develop a functional, cooperative community. Neutrality made it easy for people to work together with relatively little conflict. And the Google effect provided a steady supply of "fresh blood"--who in turn supplied increasing amounts of content.

Probably, all or nearly all other project rules were either optional, or straightforward applications of these principles. The project probably would still have succeeded nicely even if it had moderated or tweaked some of the above principles. For instance, radical openness, that is, being open even to those who brazenly flouted and disrespected the project's mission, was surely not necessary; after all, without them, the project would have been more welcoming to the many people who felt they could not work with such difficult people. And if we had required people to sign in, that would not have made very much difference (although it probably would have made some in the beginning; the project wouldn't have grown as fast). Of course we didn't have to use the GNU FDL for the license. Certainly, we did not need to set the community up initially as an anarchy governed by some vague consensus: instead, we could have adopted a charter from the very start. The project could have been managed quite differently; there could have been specially-designated and well-qualified editors. The project could have officially encouraged and deferred to experts. An article approval process could have been adopted without threatening the principle of posting unedited content for collaboration. Certainly, many of the later bells and whistles--the arbitration committee, a three-revert rule, having administrators with the particular configuration of rights they have, etc.--were not absolutely necessary to adopt in the precise forms they took. These differences would not have threatened the basic principles that made the project work, listed above.

So the basic principles that explain why Wikipedia could start working--and still does work--are relatively simple, few in number, and above all general. The more specific principles that Wikipedia wound up with was a matter of historical accident. There was a great deal of "wiggle room." Those intent on studying or replicating the Wikipedia model would do well to bear that in mind.

A series of controversies

So much for the very early history of Wikipedia; the next phase involved rapid growth and some serious internal controversies over policy and authority. If Wikipedia's basic policy was settled upon in the first nine months, its culture was solidified into something closer to its present form in the next nine.

The project continued to grow. We had 6000 articles by July 8; 8000 by August 7; 11,200 by September 9; and 13,000 by October 4. Consulting the website logs, we noted a Google effect: each time Google spidered the website, more pages would be indexed; the greater the number of pages indexed, the more people arrived at the project; the more people involved in the project, the more pages there were to index. In addition to this source of new contributors, Wikipedia was Slashdotted several times, and had large influxes of new users particularly after two articles I wrote for Kuro5hin were posted on Slashdot: "Britannica or Nupedia? The Future of Free Encyclopedias (July 25, 2001) and Wikipedia is wide open. Why is it growing so fast? Why isn't it full of nonsense? (September 24, 2001).

This growth brought difficult challenges, challenges that perhaps I did not sufficiently anticipate and plan for. Some of our earliest contributors were academics and other highly-qualified people, and it seems to me that they were slowly worn down and driven away by having to deal with difficult people on the project. I hope they will not mind that I mention their names, but the two that stick in my mind are J. Hoffman Kemp and Michael Tinkler, a couple of Ph.D. historians. They helped to set what I think was a good precedent for the project in that they wrote about their own areas of expertise, and they contributed under their own, real names. The latter has the salutary effect of making the contributor more serious and more apt to take responsibility for his or her contributions. They are also very nice people, but did not "suffer fools gladly," as the phrase goes. Consequently, they wound up in some pretty silly disputes that would have driven less patient people away instantly. So there was a growing problem: persistent and difficult contributors tend to drive away many better, more valuable contributors; Kemp and Tinkler were only two examples. There were many more who quietly came and quietly left. Short of removing the problem contributors altogether--which we did only in the very worst cases--there was no easy solution, under the system as we had set it up. And I am sorry to have to admit that those aspects of the system that led to this problem were as much my responsibility as anyone else's. Obviously, I would not design the system the same way if given the chance again.

As a result, I grew both more protective of the project and increasingly sensitive to abuse of the system. As I tried to exercise what little authority I claimed, as a corrective to such abuse, many newer arrivals on the scene made great sport of challenging my authority. One of the earliest challenges happened in late summer, 2001. The front page of Wikipedia--then open to anyone to edit, like any other page on the project--was occasionally vandalized with infantile graffiti. Someone then tried to make an archive of the vandalism that had been done to the front page of Wikipedia. I maintained that to make such an archive would be to encourage such vandalism, so I deleted the archive. This occasioned much debate. Then a user made the archive a "subpage" of his own user page--and user pages were generally held to be the bailiwick of the user. Consequently I deleted that subpage, which occasioned a further hue and cry that, perhaps, I was abusing my authority. The vandalism-enshrining user in question proceeded to create a "deleted pages" page, on which the deleted vandalism archives were listed, as if to accuse me of trying to act without public scrutiny; but this was, of course, perfectly acceptable to me. At the time, I thought that this controversy was just as silly as it will sound to most people reading this; I thought that I needed only to "put my foot down" a little harder and, as had happened for the first six months of the project, participants would fall into line. What I did not realize was that this was to be only the first in a long series of controversies, the ultimate upshot of which was to undermine my own moral authority over the project and to make the project as safe as possible for the most abusive and contentious contributors.

Throughout this and other early controversies, much of the debate about project policy was conducted on the wiki itself. Other debates were conducted on mailing lists, Wikipedia-L and then later, for the English language project, WikiEN-L. In addition, people had taken to putting their own essays on Wikipedia, as subpages of their user pages. These too were occasioning debate. It seemed to me, and many other contributors, that this debate was distracting the community from our main goal: to create an encyclopedia. Consequently I proposed that we move the debate to another wiki that was to be created specifically for that purpose--what became known as the "meta wiki." This proposal was very widely supported, so we set it up.

As it happened, the meta-wiki became even more uncontrolled than Wikipedia itself, and for many months was continually infested with contributions by people that can only be called "trolls." That epithet came to be discouraged, however, for reasons soon to be explained. The existence of trolls was a problem we felt we should tolerate--and deal with only verbally, not with harsh penalties--for the sake of encouraging the broadest amount of participation. In the first years, only the worst trolls were ever expelled from the project. I do not know whether this policy has been changed as a result of the operation of the much-later installed Arbitration Committee.

The reasons the meta-wiki became (at least temporarily) more uncontrolled are not far to seek. First, it had no specific purpose, other than to host project debate and essays that do not belong on the main wiki--which was not enough to make anyone care very much about it. Second, because many people did not care what happened on the meta-wiki, they did not do the very necessary weeding that takes place on Wikipedia; besides, as the meta-wiki was a repository of opinion, people felt less comfortable editing or deleting what was, after all, only opinion.

What happened was that project policy discussions moved almost exclusively to the project mailing lists. There is a reason why this was a superior solution to having much debate on an uncontrolled, "unmoderated" wiki. On a wiki, contributions exist in perpetuity, as it were, or until they are deleted or radically changed; consequently, anyone new to a discussion sees the first contribution first. So whoever starts a new page for discussion also, to a great extent, sets the tone and agenda of the discussion. Moreover, nasty, heated exchanges live on forever on a wiki, festering like an open wound, unless deliberately toned down afterwards; if the same exchange takes place on a mailing list, it slips mercifully and quietly into the archives.

At about the same time that we decided to start the meta-wiki, and soon after the vandalism archive affair, I was thinking a great deal about Wikipedia's apparent anarchy, and I wrote an essay titled, "Is Wikipedia an experiment in anarchy?" This and the discussion that ensued tended to ossify positions with regard to the authority issue: I and a few others agreed that Jimmy and I should have special authority within the system, to settle policy issues that needed settling. Jimmy was relatively quiet about this issue; this, I think, was probably because his authority was generally not in question, but mine was, because I was "in the trenches" and continuing to encourage good habits and solidify policy positions.

By November or December of 2001, Wikipedia was growing so fast and the subject of regular news reporting, even by the likes of The New York Times and MIT's Technology Review; after the two major Slashdottings earlier in the year, we knew that large influxes of members could have a tendency to change the nature of the project, and not necessarily for the better. If there were some major news coverage--an evening news story in the U.S., for example--there might be many new people who would need to be taught about Wikipedia's standards and positive cultural aspects. So I proposed what I thought was a humorously-named "Wikipedia Militia" which would manage new (and very welcome) "invasions" by new contributors. By this time, however, there was a small core group of people who were constantly on the watch for anything that smacked the least bit of authoritarianism; consequently, the name, and various aspects of how the proposal was presented, were vigorously debated. Eventually, we switched to "The Wikipedia Welcoming Committee" and finally, the "Volunteer Fire Department"--which eventually, it seems, fell into disuse.

The governance challenge

After the September Slashdotting, I composed a page originally called "Our Replies to Our Critics" (and now called "Replies to Common Objections"), in which I addressed the problem that "cranks and partisans" might abuse the system:

Moreover--and this is something that you might not be able to understand very well if you haven't actually experienced it--there is a fair bit of (mostly friendly) peer pressure, and community standards are constantly being reinforced. The cranks and partisans, etc., are not simply outgunned. They also receive considerable opprobrium if they abuse the system.

This reflects very well the conception I had in September 2001 of Wikipedia's culture; the reply above was as much hopeful and prescriptive as descriptive. But it turned out to be only partly true. As difficult users began to have more of a "run of the place," in late 2001 and 2002, opprobrium was in fact meted out only piecemeal and inconsistently. It seemed that participation in the community was becoming increasingly a struggle over principles, rather than a shared effort toward shared goals. Any attempt to enforce what should have been set policy--neutrality, no original research, and no wholesale deletion without explanation--was frequently if not usually met with resistance. It was difficult to claim the moral high ground in a dispute, because the basic project principles were constantly coming under attack. Consequently, Wikipedia's environment was not cooperative but instead competitive, and the competition often concerned what sort of community Wikipedia should be: radically anarchical and uncontrolled, or instead more singlemindedly devoted to building an encyclopedia. Sadly, few among those who would love to work on Wikipedia could thrive in such a protean environment.

It is one thing to lack any equivalent to "police" and "courts" that can quickly and effectively eliminate abuse; such enforcement systems were rarely entertained in Wikipedia's early years, because according to the wiki ideal, users can effectively police each other. It is another thing altogether to lack a community ethos that is unified in its commitment to its basic ideals, so that the community's champions could claim a moral high ground. So why was there no such unified community ethos and no uncontroversial "moral high ground"? I think it was a simple consequence of the fact that the community was to be largely self-organizing and to set its own policy by consensus. Any loud minority, even a persistent minority of one person, can remove the appearance of consensus. In fact, I recall that (in October 2002, after I resigned) I felt compelled by ongoing controversies to request that Jimmy declare that certain policies were in fact non-negotiable, which he did. Unfortunately, this declaration was too little, too late, in my opinion.

By late 2001, I had gained both friends and detractors. I think I had become, within the project, a symbol of opposition to anarchism, of the enforcement of standards, and consequently of the exercise of authority in a radically open project. But I was still trying to manage the project as I always had--by force of personality and "moral" authority. So when people arrived who clearly and openly disrespected established policy, I was, in my frustration, very short with them; and when the project continued to try to establish new policies, my role in articulating those policies and actually establishing them (attempting to express a "consensus") was challenged. This undermined what moral authority I had. I felt my job was on the line, and the project continued in turmoil day in and day out; from my point of view, fires were spreading everywhere, and as I had become a somewhat controversial figure, I did not have quite enough allies to help me put them out. Consequently I was rather too peremptory and short with some users. This, however, exacerbated the problem, because the attitude could not be backed up by punishment; harsh words from a leader are empty threats if unenforceable; I thereby handed my anti-authoritarian "wiki-anarchist" opponents an advantage, because--ironically--they were able to portray me as dictatorial, when I was anything but. I came to the view, finally and belatedly, that it would be better to "ignore the trolls." But as it turns out, this is particularly hard to do on a wiki, because, again, unlike on an e-mail list, trollish contributions do not just disappear into the archives; they sit out in the open, as available as the first day they appeared, and "festering." Attempts to delete or radically edit such contributions were often met by reposting the earlier, problem version: the ability to do that is a necessary feature of collaboration. Persistent trolls could, thus, be a serious problem, particularly if they were able to draw a sympathetic audience. And there was often an audience of sympathizers: contributors who philosophically were opposed to nearly any exercise of authority, but who were not trolls themselves.

It is surely very ironic that it was I personally who (initially) so strongly supported the lack any enforceable rules in the community. Some legal theorists would maintain that a community that lacks enforceable rules lacks any law at all. In retrospect it is clear that there was a fundamental problem with my role in the system: to have real authority, I needed both to be able to enforce the rules and, for both fairness and the perception of fairness, there needed to be clear rules from the beginning. But, by my own design, I had very early on rejected the label "editor-in-chief" and much real enforcement authority; a year into the game, it would have been difficult if not impossible to claim enforcement authority over active but problem users. Moreover, I was the author of the "ignore all rules" rule. My early rejection of any enforcement authority, my attempt to portray myself and behave as just another user who happened to have some special moral authority in the project, and my rejection of rules--these were all clearly mistakes on my part. They did, I think, help the project get off the ground; but I really needed a more subtle and forward-looking understanding of how an extremely open, decentralized project might work.

In retrospect, I wish I had taken Teddy Roosevelt's advice: "Speak softly and carry a big stick." Since my "stick" was very small, I suppose I felt compelled to "speak loudly," which I regret. (This was not such a problem, by the way, on Nupedia; partly, that was because there were not nearly as many problem users on Nupedia, but partly it was because there was clear enforcement authority.) As it turns out, it was Jimmy who spoke softly and carried the big stick; he first exercised "enforcement authority." Since he was relatively silent throughout these controversies, he was the "good cop," and I was the "bad cop": that, in fact, is precisely how he (privately) described our relationship. Eventually, I became sick of this arrangement. Because Jimmy had remained relatively toward the background in the early days of the project, and showed that he was willing to exercise enforcement authority upon occasion, he was never so ripe for attack as I was.

Perhaps the root cause of the governance problem was that we did not realize well enough that a community would form, nor did we think carefully about what this entailed. For months I denied that Wikipedia was a community, claiming that it was, instead, only an encyclopedia project, and that there should not be any serious governance problems if people would simply stick to the task of making an encyclopedia. This was strictly wishful thinking. In fact, Wikipedia was from the beginning and is both a community and an encyclopedia project. And for a community attempting to achieve something, to be serious, effective, and fair, a charter seems necessary. In short, a collaborative community would do well to think of itself as a polity with everything that that entails: a representative legislative, a competent and fair judiciary, and an effective executive, all defined in advance by a charter. There are special requirements of nearly every serious community, however, best served by relevant experts; and so I think a prominent role for the relevant experts should be written into the charter. I would recommend all of this to anyone launching a serious online community. But indeed, in January 2001, we were in both "uncharted" and "unchartered" territory. The world, I think, will be able to benefit from this and our other initial mistakes.

But in fairness to ourselves, it was a good idea to allow the community to decide by experience and consensus what article content rules to endorse. This allowed us to generate a very sensible set of article content rules. To be clear, I think it was not such a good idea to apply the same thinking to the organization of the community itself; we should have acknowledged that a community would form, that it would have certain persistent and difficult issues that would need to be solved, and that a lack of any effective founding community charter might result in chaos.

My resignation and final few months with the project

Throughout the governance controversy, I was preparing for my wedding, which happened December 1, 2001. A few days after I arrived back from my honeymoon, I was informed that I should probably start looking for another job, because Bomis was having to lay off most of its workers; they had 10-12 workers at the end of 2000, and by the beginning of 2002 they were back to their original 4-5. My salary was reduced in December and then halved in January. This seemed inevitable because Wikipedia was not bringing in any money at all for Bomis, even if Wikipedia was becoming even more of a publicly-recognized, if still modest success. Our first anniversary came just before we announced having 20,000 articles, and I was invited to talk about the project at Stanford on January 16 (text here; you might notice that I was still plugging the notion of using Nupedia to vet Wikipedia articles, as an answer to the objection that Wikipedia articles are unreliable).

I was officially laid off at the beginning of February, which I announced a few weeks later. I had continued on as a volunteer; Wikipedia and Nupedia were, after all, volunteer projects. But I was laboring in the aftermath of the governance controversies of the previous fall and winter, which promised to make the job of a volunteer project leader even more difficult. Moreover, I had to look for a real job. So throughout the month of February I considered resigning altogether.

But Jimmy had told me the previous December that Bomis would start trying to sell ads on Wikipedia in order to pay for my job. Even in that horrible market for Internet advertising, there were already enough pageviews on Wikipedia that advertising proceeds might have provided me a very meager living. We knew that this would be extremely controversial, because so many of the people who are involved in open source and open content projects absolutely hate the idea of advertising on the web pages of free projects, even to support project organizers. In fact, when this advertising plan was announced, in late February of 2002, the Spanish Wikipedia was forked (something I urged them not to do).

Bomis was not successful in selling any ads for Wikipedia anyway--you might recall that early 2002 was at about the very bottom of the market for Internet advertising. I also had some hope that we might, finally, set up the project's managing nonprofit, which we had discussed doing for a long time (and which eventually did come into being: Wikimedia). The job of setting up the nonprofit was left to me, but ongoing controversies seemed to eat up any time I had for Wikipedia, and frankly I had no idea where to begin. So, after a month without pay, I announced my general resignation; I completely stayed away from the project for a few months.

Just by the way, Wikipedia's offshoot projects--a dictionary, a textbook project, a quotation project, a public domain book repository, etc.--were all started in 2002 or later, and I cannot claim any credit for them. I did supply the name "Wiktionary" in April 2001, more or less on a whim. I quickly disavowed any responsibility for leading any such project, and it seems the Wiktionary project did not start up for another year and a half (December 12, 2002). My view now is that Webster's and the OED are quite good enough as far as English dictionaries go, and there will always be excellent free dictionaries in every language online. To try to develop a dictionary by collaboration among random Internet users, particularly in a completely uncontrolled wiki format, now strikes me as a nonstarter. I confess I am now puzzled why I didn't think so instantly; it was no doubt because I simply was throwing out ideas as they occurred to me, and also because we had too many dictionary definition-type entries in Wikipedia. (So why not give people a place to put their dictionary definitions?--Perhaps that's what I was thinking, but it hardly seems like a good justification for starting a project.) But Jimmy's first reaction was properly skeptical regarding the use of wikis and Ruth Ifcher made a stronger criticism very nicely. Dictionaries, even more than encyclopedias, must be extremely reliable to be even minimally usable; without direct oversight by linguists, a public dictionary project seems pointless. As to the other projects, they are mostly conducted using wikis and according to some of the basic founding principles of Wikipedia. But other sorts of project--for example, textbook projects, quotation repositories, and archives--necessarily require quite different specifications from those of an encyclopedia. For example, the fact that the wiki format works for encyclopedia development hardly means that it is appropriate for the hosting of public domain books. Since the same texts are available in many other places online, such as the wonderful Project Gutenberg, why would anyone choose to read The Iliad on a wiki, which could have been subtly changed by any random passer-by, without any oversight by someone who had access to an authoritative text? There is a fact about the way the text actually reads; so is editing via wiki software more apt to increase or reduce the number of errors over other systems, such as Project Gutenberg's? I do not mean to dismiss any such efforts. I simply think that considerable thought needs to be put into exactly how those other projects should be organized: the wiki format is not a magic pill that somehow makes all problems go away. Wiki is just one software paradigm, which must be adapted, supplemented, changed, or replaced in order to solve the unique set of problems a project faces.

In the spring, a controversy erupted. Caring as I did--and as I still do--about the future of free encyclopedias, I felt compelled to get involved. The controversy featured a troll who was putting up huge numbers of screeds on the "meta-wiki" and on Wikipedia as well. The controversy began with a discussion of what to do about, and how to react to, this particular troll. I maintained that one should not "feed the troll," and that the troll should be "outed" (it was an anonymous user, but it was not hard to use Google to determine the identity of the troll) and shamed.

There resulted a broader controversy about how to treat problem users generally. There were, as I recall, two main schools of thought. One, to which I adhered and still adhere, was that bona fide trolls should be "named and shamed" and, if they were unresponsive to shaming, they should be removed from the project (by a fair process) sooner rather than later. We held that a collaborative project requires commitment to ethical standards which are--as all ethical standards ultimately are--socially established by pointing out violations of those standards. Hence naming and shaming. A second school of thought held that all Wikipedia contributors, even the most difficult, should be treated respectfully and with so-called WikiLove. Hence trolls were not to be identified as such (since "troll" is a term of abuse), and were to be removed from the project only after a long (and painful) public discussion. For the latter school, it seemed to me, the only really egregious faux pas one could commit in the project was to suggest that there were objective standards that could be enforced via "shaming."

I felt at the time that the prevalence of the second school entailed rejection of both objective standards and rules-based authority. It is impossible to explain why one is removing some partisan screeds from the wiki without, in some way, identifying it as a partisan screed, and pointing out that such productions are inconsistent with the neutrality policy. This will necessarily be received as less than respectful and "loving," especially if one must engage the troll himself in a long, drawn-out dispute; in a very long dispute with any trollish type, it is only a matter of time before some epithet gets bandied about, since they are so darned useful (and accurate) when applied to trollish types. More generally, the very application of rules, or laws, entails a moral judgment, or what for its effectiveness must have the force of a moral judgment. I suppose I agree with those legal theorists who say that there is necessarily, in its core, a moral component to the law. Consequently, the new policy of "WikiLove" handed trolls and other difficult users a very effective weapon for purposes of combatting those who attempted to enforce rules. After all, any forthright declaration that a user is doing something that is clearly against established conventions--posting screeds, falsehoods, nonsense, personal opinion, etc.--is nearly always going to appear disrespectful, because such a declaration involves a moral accusation. The only way to avoid such an appearance of disrespect, perhaps, is to step very lightly and use much flattery and qualifications: "Now don't get me wrong, I think you're doing a good job overall, but it seems to me that in this particular case, your contribution is slightly inconsistent with the neutrality policy." Suppose the offender replies: "So what? I disagree with the neutrality policy." Or: "I disagree. What I wrote is perfectly neutral. Who do you think you are, anyway?" It is a very rare person who can practice "WikiLove" in such a case. In Wikipedia's developing culture, if anyone reacted out of frustration, or merely attempted to apply the law as a moral instrument, as laws typically are applied, he would become the problem, and a much more serious problem, than mere violations of the neutrality policy, say. The result is that, on pain of becoming persona non grata in the community, one had to treat brazen, self-conscious violators of basic policy with particular respect. It was a perfect coup for the resident wiki-anarchists. I again left the project for several months.

In fall of 2002, I had started teaching at a local community college, and with some extra time on my hands, I started editing Wikipedia a little and engaging in mailing list discussions. I think my first new post to Wikipedia-L, from September 1, 2002, was "Why the free encyclopedia movement needs to be more like the free software movement." In it I argued that the free software movement is led and dominated by highly-qualified programmers, and that the "free encyclopedia movement"--that is, Wikipedia, Nupedia, and other newer projects--needs to move in that direction. I suggested that Nupedia be redesigned to release "approved" versions of Wikipedia articles; Wikipedia itself was not to be touched. This proposal met with a very cool reception. After a few months of discussion, Jimmy himself was "intending to revive Nupedia in the near future" and "thinking very much along the lines of what is being discussed here." Unfortunately, this never happened.

By November or December, I think, I proposed, and Magnus Manske very helpfully coded, an expert-controlled approval process for Wikipedia that was in fact to be independent of both Nupedia and Wikipedia. It would not have affected the Wikipedia editorial process. It would have lived in a separate namespace or domain, as an independent add-on project for Wikipedia. Without explaining the details, expert reviewers, the recruitment of which I would organize, would examine Wikipedia articles and approve or disapprove of particular versions of those articles. We set up a mailing list, Sifter-L (archives no longer online, apparently), which for several weeks discussed policy issues.

There was not a great deal of support for the proposal on Wikipedia-L. There was little or no excitement that the new project might bring into Wikipedia a fresh crop of subject area specialists. But that was fine as far as I was concerned, since the project was to operate independently of Wikipedia. Still, I had the very distinct sense that any specialists arriving on the scene would not necessarily be met with open arms--particularly if before approving an article they wished to make whatever changes to articles that they felt necessary. There were even a few Wikipedians who made it clear that experts should not expect to be treated any differently than anyone else, even when writing about their areas of expertise.

I then considered whether the interaction between Wikipedians and the new reviewers might be a problem after all. Surely, I thought, most specialists would want to edit even very good articles before approving them (in the independent system). This would require that the reviewers interact with Wikipedians. Wikipedia's culture had become such that disrespect of expertise was tolerated, and, again, trolls were merely warned, but very politely (in keeping with the policy of WikiLove), that they please ought to stop their inflammatory behavior. Trolls would certainly find ripe targets in expert reviewers, I thought. I recalled that patient, well-educated Wikipedians like J. Hoffmann Kemp and Michael Tinkler had been driven off the project not only by trolls but by some of the more abrasive and disrespectful regulars. I then considered: could I in good conscience really ask academics, who are very busy, to engage in this activity that would probably annoy most of them and do nothing to contribute to their academic careers? Recruiting for Nupedia was very easy by comparison, and caused me no such pangs of conscience.

I believe it was this problem that finally prompted me, in I believe January of 2003, to inform Jimmy as follows (by private e-mail): I was breaking with the project altogether; the only way he could prevent this, I told him, was that he personally crack down on problem users, and make the project more officially welcoming to experts. I also told him that I did not expect this information to change his mind, and that I did not mean to issue an ultimatum. And in fact our exchange did not change his mind. I concluded that we had a fundamental philosophical disagreement about how the project should be run. I respected and still respect his view. That is where matters ended, and it was then that I broke with Wikipedia altogether.

Some final attempts to save Nupedia

Nevertheless, I was interested in pursuing Nupedia's development. It still seemed rescuable to me.

I recall two incidents in which I tried to have Nupedia revived, in 2002 or 2003, but I don't recall exactly. First, I approached Jimmy with the offer to try to find a buyer/managing organization for Nupedia. The suggestion was that, since Bomis did not have enough money to support it, and since Jimmy did not appear to have any specific intentions with the project other than to let it run on the system set up in 2000-1, I might be able to find a university or other organization that would take on the responsibility. I do not recall the details, but we did not pursue this possibility. Second, and later, I offered to buy Nupedia myself--that is, the domain name, the membership list, and whatever other proprietary material Bomis might have controlled. I wanted to start it up again as a simpler, more streamlined, but still fully peer-reviewed project; I thought, moreover, that if I owned it I might be able to give it to a suitable sponsoring educational or nonprofit institution. Jimmy seemed cool to the idea, and did not ask for any specific offers.

Perhaps it is, therefore, not entirely accurate to say that Nupedia died due to the inefficiency of its system. To some extent it was also allowed to die, even after it was clear that its former editor-in-chief expressed an interest in continuing the project under an entirely different system. The result was that, without a leader or organization that could support its mission, Nupedia died a slow death. The server it lived on had some trouble in 2003, and as a result the website went offline. For whatever reason, the website was never brought up again after that.

I obviously cannot speak for Jimmy, but I will say that, if he was worried that Nupedia would essentially fork Wikipedia--again, I don't claim that he had that concern--then it seems to me that such a concern would not have justified letting Nupedia wither untended. The projects, Wikipedia and Nupedia, were naturally complementary parts of a single, symbiotic whole. That at least is how I always regarded them, indeed, from the very founding of Wikipedia. From the founding of Wikipedia, I always thought Wikipedia without Nupedia would have been unreliable, and that Nupedia without Wikipedia would have been unproductive. Together they were to be an "unstoppable high-quality article-creation juggernaut."

It is still disappointing to me, that we made plans and promises to thousands of Nupedians, including hundreds of extremely well-qualified people, some of them leaders in their fields. We spent many thousands of person-hours, all told, on the project. I apologize to those people, and I can only hope that they will find some future open content encyclopedia project worthy of their participation, one that will show the world the potential that Nupedia had.

Conclusions

I have some advice for anyone who would like to start new projects on the model of Wikipedia.

You can learn from Wikipedia's success; so, first and most importantly, see above for considerations about why Wikipedia works.

But you can also learn from our mistakes. The following primarily concerns project governance, because governance issues are, in my opinion, the primary failing of Wikipedia. Bear in mind, also, that these are only rough guidelines, for those who are starting projects that have enough resemblance to Wikipedia. These are not perfectly general rules:

• If you intend to create a very large, complex project, establish early on that there will be some non-negotiable policy. Wikis and collaborative projects necessarily build communities, and once a community becomes large enough, it absolutely must have rules to keep order and to keep people at work on the mission of the project. "Force of personality" might be enough to make a small group of people hang together; for better or worse, however, clearly enunciated rules are needed to make larger groups of people hang together.
• There is some policy that, with forethought, can be easily predicted will be necessary. Articulate this policy as soon as possible. Indeed, consider making a project charter to make it clear from the beginning what the basic principles governing the project will be. This will help the community to run more smoothly and allow participants to self-select correctly.
• Establish any necessary authority early and clearly. Managers should not be afraid to enforce the project charter by removing people from the project; as soon as it becomes necessary, it should be done. Standards that are not enforced in any way do not exist in any robust sense. Do not tolerate deliberate disruption from those who oppose your aims; tell them to start their own project; there's a potentially infinite amount of cyberspace.
• As any disagreements among project managers are apt to be publicly visible in a collaborative project, and as this is apt to undermine the (very important) moral authority of at least one manager, make sure management is on the same page from the beginning--preferably before launch. This requires a great deal of thinking through issues together.
• In knowledge-creation projects, and perhaps many other kinds of projects, make special roles for experts from the very beginning; do not attempt to add those roles later, as an afterthought. Specialists are one of your most important resources, and it is irrational not to use them as much as you can. Preferably, design the charter so that they are included and encouraged. Moreover, make the volunteer project management a meritocracy, and not based on longevity but based on the ability to lead and contribute to the project; that is the only condition under which very many of the best qualified people will want to participate.

Another point needs more in-depth development.

Radical and untried new ideas require constant refinement and adaptation in order to succeed; the first proposal is very rarely the best, and project designers must learn from their mistakes and constantly redesign better projects. Nupedia's Advisory Board failed to admit to inherent flaws in its system, and its delay in admission shut the window of opportunity to its improvement. And it seems to me that the Wikipedia community fell into a mistake by thinking that just one or two features--the wiki feature and the neutrality policy and a few other things--explained Wikipedia's success, and that those features can thus be applied with no significant changes to new projects. But there is no substitute for constant creativity and problem-solving--nor for honesty about what problems need solving. The honesty to recognize problems and creativity in solving them are, after all, what made Wikipedia succeed in the first place.

This is a crucial point: if you use a tool or model from another project, think through very carefully how that tool or model should be adapted. Do not assume that you need to use every feature, or every aspect of the surrounding culture, that you are borrowing. Wikipedia borrowed rather too much from (1) the culture of wikis, (2) unmoderated online discussions, and (3) free-wheeling online culture generally. To be sure, Wikipedia is also a product of those cultures, and works as well as it does largely because of what it borrowed from those cultures. But it also shares some of its more serious current flaws with such cultures. Those planning new projects, or wanting to overhaul old ones, might well bear in mind that a certain cultural context, including the context that has grown up around a tool, just might not be right for that project. Let me elaborate.

(1) Consider first the culture of wikis. On the one hand, I said we wanted to determine the best rules, and experience would help us determine that; so we had no rules to begin with. On the other hand, one might add that another reason we began without rules was that we were partaking in the extremely uncontrolled, free-wheeling nature of "traditional" wikis. I think that's right. But there is an excellent reason why an encyclopedia project should not partake in that extremely uncontrolled nature of wiki culture, and why it should adopt actually enforceable rules: unlike traditional wikis, encyclopedia projects have a very specific aim, with very specific constraints, and efficient work toward that aim, within those constraints, practically requires the adoption of enforceable rules. The mere fact that most wikis, when Wikipedia was created, did not have enforceable rules hardly meant that one could not innovate further, and create one that did have rules.

(2) Moreover, Jimmy and I and most of the first participants on Wikipedia were veterans of unmoderated Internet discussion groups, and hence, naturally, we could appreciate the advantages of letting a virtual community develop in the absence of any real (enforcement) authority. In unmoderated forums there is often found a sense, among some participants, that any attempt to oust a particularly troublesome user amounts to unjustifiable censorship. The result is that the existence of many unmoderated forums online has created a small army of people militantly opposed to the slightest restriction on speech, who feel that they do and should have a right to say whatever they like, wherever they like, online. Any attempt to create and enforce rules for Internet projects, when that small army is ready to cry "censorship," will seem daring or even outrageous in many contexts online. But there is an excellent reason why such anarchy is inappropriate for many projects, including encyclopedia projects, even one that is self-policing like a wiki: there simply must be a way to enforce rules in order for rules to be effective. Given that encyclopedia project development happens almost entirely using words, nearly any rules will also be restrictions on speech. Anyone who advocates many enforceable rules on a collaborative project, in the cultural context of an Internet filled with so many unmoderated discussion groups, can be made to seem reactionary. But this is only a result of that cultural context; in any other context, the existence of rules would be perfectly natural and unobjectionable.

(3) Finally, and generally speaking, the Internet is a great leveller. Since social interaction can proceed among complete strangers who cannot so much as see each other, things that seem to matter in many "meatspace" discussions, such as age, social status, and level of education, are often dismissed as unimportant online. Many Internet forums, chatrooms, and blogs are populated by people who are identified by only a "handle," and any suggestion that communication should be restricted or in any way altered in accordance with "expertise" or "authority" is likely to be met with outrage, in most forums. But there are several excellent and obvious reasons why expertise does need special consideration in an encyclopedia project, and in other collaborative projects. First, there are many subjects that dilettantes cannot write about credibly; I, for example, could not write very credibly about astronomy or speleology, but I have a passing interest in both. If I am working only with other dilettantes, our articles are apt to remain amateurish at best; we can fill in the gaps in each other's knowledge, and do research, but the results will remain problematic until someone with more knowledge of the subject contributes. Second, there are very many specialized subjects about which no one but experts has any significant knowledge at all. Third, it is only the opinions of experts that will be trusted by most of the public as authoritative in determining whether an article is generally reliable or not. Moreover, the standards of public credibility are not likely to be changed by the widespread use of Wikipedia or by online debate about the reliability of Wikipedia. Like them or hate them, those are the facts. But if one points these facts out online, culturally "levelled" as it is, particularly in forums or projects like Wikipedia which go out of their way to ignore individual differences among people, one finds a frosty reception at best.

Consider, if you will, that it was because Wikipedia was started in the context of the ingrained cultures of wikis, of unmoderated discussion forums, and of the levelling, anti-elitist influence of the Internet at large, that it was very difficult for us to exercise the maximal amount of creativity that a maximally successful project would require. In establishing a new cultural context, we were deeply constrained by the old. Now, to be sure, I have said above and many times elsewhere that Wikipedia did not have to adopt the particular conjunction of policies that it did. But it is not surprising that it did adopt its particular conjunction of policies, considering the conjunction of influences on its development. So it would have required much more explanation and persuasion, and indeed, much more struggle, for us to, for example, have persuaded potential participants that some persons, even in a wiki environment, should have special rights that others do not. So powerful is the influence of cultural context that there are quite a few people whose lack of imagination is such that they believe I simply must not understand "why Wikipedia works" if I am willing to suggest that it does not have to work in precisely the way it does work. Constantly-reinforced cultural habits die very hard indeed, and place very strong constraints upon what can be imagined, and what bare possibilities seem even worth thinking about.

But it was our willingness to exercise our creativity and follow our imagination, and create what is, to some extent, a new kind of culture, that led to Wikipedia's success. For the overall project of creating open content encyclopedias--and indeed, for the fantastic collaborative Internet that has yet to be created--to reach its full potential, the process of identifying mistakes honestly and creatively seeking solutions must be ramped up and continued unabated.

Many thanks to Larry Sanger and to O'Reilly for this memoir.