UPDATE: See additional notes at the end of the post regarding the 2015 fundraiser and the early 2016 executive shake-up and controversy.

I'm a very avid user of Wikipedia: I view about 500-1000 Wikipedia articles a month, and have created over 200 Wikipedia articles that get a total of over 100,000 monthly pageviews. I've also used the underlying MediaWiki software powering Wikipedia to create separate wikis on some subjects (including a group theory wiki that gets about 750,000 annual pageviews). So in general, I'm a fan both of the Wikipedia project and of the software and technology that powers it.

Nonetheless, I believe that some of the common arguments people make in favor of the donating to the Wikimedia Foundation (WMF), the non-profit that hosts Wikipedia and a number of its sister sites, are flawed. In this post I attempt to outline my reasoning. I originally considered writing this post in December during the WMF's controversial fundraiser, but wasn't able to find an appropriate framing for the post, and ran out of time. A few days ago, I thought of a new and better way of writing the post, and here it is.

Table of contents

• 1. Arguments, background, and disclaimers
  • 1.1. Arguments deployed for donating to the Wikimedia Foundation
  • 1.2. Where I'm coming from: background and potential conflict of interest
  • 1.3. What I'm (not) trying to do
  • 1.4. A bit more on what this post is not about
• 2. Evaluation of the arguments
  • 2.1. Is Wikipedia in danger of death?
  • 2.2. Is the Wikimedia Foundation unusually well-run and efficient?
  • 2.3. Does the marginal and/or average donation contribute unambiguously to improving Wikipedia quality?
  • 2.4. Does Wikipedia helping you personally make it morally obligatory or desirable to donate to the WMF?
  • 2.5. Is there a Kantian imperative to donate?
• 3. Conclusion
• 4. Addenda
  • 4.1. ADDED December 2015: Remarks in light of renewed interest due to the 2015 fundraiser
  • 4.2. ADDED March 2016: Remarks in light of the executive shake-up at the WMF

1. Arguments, background, and disclaimers

1.1. Arguments deployed for donating to the Wikimedia Foundation

On December 2, 2014, Jimmy Wales, Wikipedia co-founder, sent me this email message on behalf of the WMF:

Dear Vipul,

Thank you for helping keep Wikipedia online and ad-free. I’m sure you’re busy, so I’ll get right to it. We need your help again this year. Please help us forget about fundraising and get back to improving Wikipedia.

If all our past donors simply gave again today, we wouldn't have to worry about fundraising for the rest of the year.

We are the small non-profit that runs one of the top websites in the world. We only have about 200 staff but serve 500 million users, and have costs like any other top site: servers, power, programs, and people.

Wikipedia is something special. It is like a library or a public park. It is like a temple for the mind, a place we can all go to think and learn.

To protect our independence, we'll never run ads. We take no government funds. We survive on donations from our readers. Now is the time we ask.

If Wikipedia is useful to you, please take one minute to keep it online and ad-free another year.

https://donate.wikimedia.org

Thanks,
Jimmy Wales
Wikipedia Founder

Many of my friends and acquaintances, including those who aren't much into charity or philanthropy, have made small but regular donations to the Wikimedia Foundation. I myself donated $125 to the Wikimedia Foundation around 2007/2008.

Donating to Wikipedia makes prima facie sense: Wikipedia is an extremely valuable resource to humanity. How valuable? Take a look at the numbers: English Wikipedia alone gets 10 billion views a month, and all the other language Wikipedias get another 10 billion. That's a total of about a quarter of a trillion views a year. To put it another way, the mean world resident human views a little over 30 Wikipedia pages a year (and if you restrict to the population with Internet access, the number grows to over 100). If you very crudely estimate the average value of a pageview at somewhere between 0.1 and 4 cents, then the annual value delivered by Wikipedia is somewhere between $250 million and $10 billion. In comparison, the Wikimedia Foundation's annual plan for 2014-2015 calls for $58.5 million in spending. So, looking at averages alone, we seem to be getting somewhere between a 4-fold and a 160-fold return on investment.

But does donating to the Wikimedia Foundation actually make sense, once you think more closely about it? I think it could, for some particular sets of values that the donor may hold, and some particular set of beliefs about the world. However, I think that many of the arguments that people make for donating to the Wikimedia Foundation are flawed, based on a combination of factually mistaken beliefs and incorrect reasoning. In this post, I consider some of these arguments.

1. We need to donate to the Wikimedia Foundation in order to make sure that Wikipedia continues running: I believe this is wrong, even if you ignore the issue of marginal versus average donation. Reasons: the foundation is sitting on large reserves, operations costs (including server costs and associated engineering costs) are less than 20% of the budget, and even if donations from individual donors (the main source the WMF relies on in its fundraising) reduced, there are a number of large foundations that would be willing to fill in the gap. Finally, in the highly unlikely event that the WMF did go under, backups of the content as well as publicly available copies of the underlying software could be used by others to set up a clone quickly (this is also compatible with the licensing agreement for the site's content and software, so legal issues are minimal). But in any case, if the WMF does go under, that's more likely to be due to lawsuits or malice on somebody's part rather than directly because of a lack of funds.
2. The Wikimedia Foundation is unusually well-run and efficient, so there is value in rewarding it for this by donating to it: The Wikimedia Foundation is reasonably well-run and efficient, but not extraordinarily so. Some people have made a big deal of alleged inefficiences in the WMF, but I think such apparent inefficiency is par for the course for a software organization of that size and scope. At the same time, there isn't good evidence of very high efficiency. Appropriate organizations in the broad reference class include the Mozilla Foundation and Mozilla Corporation, the WordPress Foundation and Automattic (the company that runs WordPress), and GitHub. Comparing the Wikimedia Foundation to Google or Facebook and then noting that it is much leaner and therefore more efficient is misguided.
3. The marginal and/or average donation contributes to improving content quality on Wikipedia: Donations don't go directly to content creators. Rather, they go into running the Foundation, whose main costs are staff salaries (staff includes software engineers as well as others) and grants to local chapters. One could argue that some of the activities sponsored spur greater editor participation, but the evidence for this isn't clear-cut. The software improvements that would allegedly improve editor participation have generally met with heavy backlash from the community and often been retracted. You might examine the data and come to the conclusion that donating does spur more contribution, but this isn't something that can be assumed offhand.
4. Wikipedia has helped me so much, it's important to pay back: The WMF has done a reasonable job of stewardship, but I believe that insofar as there inheres a strong moral obligation to pay back, this obligation is to content creators. Donating to the WMF does not reward past content creators.
5. If everybody donated $1, then ...: The "if everybody donated $1" argument is a dangerous slippery slope to get on. The argument is generally made from a Kantian moral imperative standpoint. But it proves too much, because if everybody gave $1 for any service that appeared to vaguely benefit them, that would lead to a lot of careless spending and a lot of overfunding for sectors of the economy whose value hadn't been proven out. This is also related to the concentrated benefits, diffuse cost problem that plagues politics.

1.2. Where I'm coming from (background and potential conflict of interest)

As I noted above, I donated about $125 to the WMF in 2007/2008. I haven't donated to the WMF since then. My reasons for donating back then were partly that the WMF had much smaller budgets and so the marginal impact of additional money (or what GiveWell and others have called room for more funding) seemed more clearly positive. That said, with my current set of beliefs and goals, I probably wouldn't donate to the WMF even in a situation similar to what it was back then. Nonetheless, I don't have any deep regrets about the donations. They were a relatively small amount and I've made far bigger and more costly financial decisions of dubious and in some cases even negative value.

I've written up my history of editing Wikipedia here. TL;DR: I edited it a bit when I was an undergraduate (around 2005 and 2006) then reduced my editing involvement for a few years, then resumed around 2012. I've created over 200 Wikipedia articles that get a total of over 100,000 monthly pageviews.

Around 2008 or 2009, I had become quite interested in issues related to Wikipedia's management problems and the challenges facing the WMF. I had grown somewhat skeptical of the WMF and of Wikipedia, which was somewhat related to my reducing my editing involvement there. Some of that skepticism lives on in modified form, but my way of thinking about the issue has changed, and I don't endorse the connotations of some of my earlier writing about Wikipedia and the WMF, even if I don't have any object-level disagreements.

1.3. What I'm (not) trying to do

This post holds plenty of potential to be construed as a smackdown of the WMF in the vein of those regularly published in venues such as Wikipediocracy, the Daily Dot, and the Register (all of which are linked here; seems like most of the external publications that cover WMF minutiae generally do so from a somewhat negative angle, and these are the three top ones). So before embarking on a more detailed justification of my claims, let me clarify a little bit more what I hope to accomplish:

• I don't intend to connote that there is anything particularly morally wrong or imprudent about making small donations to the WMF. I think that the arguments that people make in favor of it being virtuous, desirable, or necessary are misguided, but it's not a particularly evil use of money. People spend money in a variety of ways, not many of which yield the sort of great returns we aspire to in our "effective altruist"-style resource allocation. If you think of donating to the WMF as a tip (which goes to the waiters and waitresses rather than the cooks!) it doesn't seem unreasonable. (I've made the analogy with a tip in an earlier Quora answer).
• I don't intend to categorically rule out the possibility that donating to the WMF could be an effective use of resources. I don't think many of the arguments people commonly make for it are epistemically sound or factually well-informed, but there could be more sophisticated arguments and reasonable sets of goals and beliefs of the world where donating to the WMF is a great choice.
• Although it's a possibility I don't rule out, I don't claim categorically that the WMF is particularly inefficient as an organization, relative to other organizations in its reference class. Nor do I claim that it is failing to pluck any obvious low-hanging fruit. Rather, my claim is that the WMF so far hasn't produced a huge bang for the buck or shown signs of strong ability to deliver value on donations (beyond keeping the site running, which is no mean feat). I believe this is largely because the problems are inherently very hard, but I don't take a position on the extent to which intractability is a barrier.
• From a consequentialist perspective, I don't expect this post to significantly affect donations to the WMF. In the highly unlikely event that it reduces annual donations by something like $150,000 (more than a full-time software engineer's salary!) that still wouldn't dent the WMF's annual budget by more than 0.3%. But actually, even if my post influences donation decisions of $150,000, the WMF is likely to still meet its fundraising target by extending the fundraising period a little longer to make up for the forgone donations. So what could be the effect of this post? First, the set of people who might be persuaded by it is likely the set of people with the most sympathies to effective altruist ideas, so in a sense, substituting their donations away from the WMF would be more high-impact than substituting the donations of somebody who isn't interested in effective use of money in general. Second, I hope that my article helps spark a discussion, both about the WMF's activities and about some of the broader topics (philanthropy, nonprofit software organizations, obligations to give).

1.4. A bit more on what this post is not about

The Wikimedia Foundation was in the news in December 2014 for allegedly misleading and alarmist language in its fundraising drive. You can see a mailing list discussion about these concerns here and read some criticism from The Register, The Daily Dot, Wikipediocracy, NewsLines, and Make Use Of.

This post is not about that controversy, though I used some of the discussion engendered by the controversy to inform my views. In other words, I'm not focused on the question of whether the WMF's messaging is deceptive or whether it is morally culpable for the (in my view misguided) reasons people feel it is virtuous, desirable, or necessary to donate to the WMF. Interesting as that issue is, I'm more interested in articulating my reasons for believing that the arguments (as I believe many people formulate them) are flawed.

Also, I am not claiming that the views here represent a consensus view of the effective altruist community, or even a dominant view within that community (for a broader spectrum of views from within the EA community, see here, and also wait for comments on the post you are currently reading). My views and perspective are informed by beliefs that are consonant with effective altruist values, but others who identify with these values may come to fairly different conclusions. Nor do I think you need to embrace general EA principles to derive value from, or even be convinced by, my arguments in this post.

Further (unlike the case of William MacAskill's critique of the Ice Bucket Challenge), the purpose of this post is not to say something like, "instead of donating $100 to the WMF, donate it instead to a GiveWell-recommended charity." Of course, if somebody reads this post and decides to not make a donation to the WMF that they otherwise would, that leaves that extra money unallocated, and insofar as they have a specific total charity budget out of which that WMF donation was coming, they have the option of allocating the money to other charities. Otherwise, that money saved goes into their general pool of fungible money, which they can allocate to charity, saving, or consumption. But the negative decision to refrain from donating to the WMF does not imply any immediate positive decision to allocate the money thereby saved in a particular way. I do have thoughts on the latter, but this post is not the place to voice them.

2. Evaluation of the arguments

2.1. Is Wikipedia in danger of death?

Recently, I came across this Reddit piece (not actual news). It's not clear where the author's actual sympathies lay, but some of the connotatively expressed viewpoints describe what many people seem to think and assume about Wikipedia.

Writing Prompt - Not Actual News[WP] Wikipedia is shut down and all copies deleted for lack of funds and loss of net neutrality. This is the founder's "I warned you, jerks" notification.

submitted 2 days ago by sarfreer

 

• 877 comments
• share

show 500

sorted by:

best

[–]ChokingVictimWordsOnTheInternet.org 3925 points 2 days ago*x5 

According to the 2014-15 annual plan, Page 8, the annual operations cost for the Wikimedia Foundation was $8.65 million for 2013-14 (projected) and $7.68 million for 2014-15 (planned), compared to overall spending of $49.2 million ($41 million internal expenses + $8.2 million grants) for 2013-14 (projected) and $58.6 million ($50 million internal expenses + $8.2 million grants) for 2014-15 (planned). The ~$8 million operations costs include the server costs and relevant engineering costs. Note that the Internet hosting costs themselves are even lower: about $2.5 million projected for 2014 (the new plan doesn't seem to include a corresponding number).

In a Quora answer, Andreas Kolbe quoted Jimmy Wales from back in 2005:

“So, we’re doing around 1.4 billion page views monthly. So, it’s really gotten to be a huge thing. And everything is managed by the volunteers and the total monthly cost for our bandwidth is about 5,000 dollars, and that’s essentially our main cost. We could actually do without the employee … We actually hired Brion [Vibber] because he was working part-time for two years and full-time at Wikipedia so we actually hired him so he could get a life and go to the movies sometimes.”

Wikipedia pageview counts have gone up by a factor of about 15X (about 20 billion monthly page requests) and pure Internet hosting costs have gone up by a factor of about 40X (from $60,000 annually to $2.5 million annually). This is roughly expected, because Wikipedia now aims for substantially greater reliability and has become a more critical resource.

I think the total engineering operations costs are a more reliable proxy for the "cost of maintaining the website" than pure Internet hosting costs. They could overcount costs because some engineering operations aren't directly necessary to keep the site running. They could also undercount costs because, even if the goal were just to keep the site running, there would likely be other administrative overhead. I think that, extrapolating somewhat, an annual cost of about $12 million would be enough to keep the site running at its current performance, if that were the sole goal. And in the event that the WMF is able to raise no additional money, it could draw down its reserves and keep running the site for about four years (the Foundation has cash reserves of $28 million and other assets of $23 million).

However, if individual donors cut back on donations and the WMF weren't able to meet its fundraising goals, it likely would simply fill in the funding gap from larger foundation and corporate donors. The reason the WMF relies on individual donors to the extent it does (about 85%) is to maintain independence from the agendas of these large donors. But this does mean that there are other lines of defense before the WMF would need to pare itself down considerably.

Finally, if the Wikimedia Foundation had to close its doors, it wouldn't take long for another resource to create a replica of Wikipedia. The entire content of the encyclopedia is available as a public dump, updated every few weeks. There would be a painful transition but it wouldn't be a permanent loss. The content is also cached by Google and the Internet Archive. Needless to say, this is unlikely to happen, and the biggest risk factors associated with it happening are natural disasters and lawsuits rather than running out of money. But lawsuits themselves won't result in an immediate shutdown. If there were some sort of lawsuit or legal trouble, people would have ample time to prepare and respond to the site going down.

ETA: Andreas Kolbe has done a similar calculation as mine here. I hadn't seen this calculation (though I had read his Quora answer linked above) before I did my own.

2.2. Is the Wikimedia Foundation unusually well-run and efficient?

Co-founder Jimmy Wales hints at this efficiency in his letter:

We are the small non-profit that runs one of the top websites in the world. We only have about 200 staff but serve 500 million users, and have costs like any other top site: servers, power, programs, and people.

Some people might interpret this as a claim that, compared with Google or Facebook or YouTube, Wikipedia is an extraordinarily efficient-to-run website. Where Google has over 50,000 employees and Facebook over 8,000, the WMF employs a mere 200. And while Google and Facebook measure their expenses in the billions, the WMF's $50 million is a bargain.

However this is a misguided conclusion for a few reasons.

1. The complexity of the task undertaken by Google and Facebook is substantially greater than that of the task undertaken by Wikipedia in terms of page serving. The majority of Wikipedia pages are loaded statically from a periodically updated cache, and require little computation or user-specific customization. Facebook, on the other hand, has to generate every pageview dynamically based on user-specific information. It's a much harder problem. Similarly, Google's search algorithm relies on indexing huge amounts of data, and its personalized search results rank them in real time using a fairly complicated strategy. YouTube probably comes closest to Wikipedia in terms of serving a predetermined page to users, but that's video streaming, which is fairly expensive in terms of bandwidth (that said, I believe the operating costs of YouTube are probably comparable to the WMF's annual budget, though much more than WMF's operations costs).
2. On priors, we shouldn't expect that the Wikimedia Foundation should have figured out a technology or strategy to be so much more efficient than others. The median WMF employee may be somewhat better than the median Google or Facebook employee, but the top 200 employees at companies like Google or Facebook are likely to be a lot better than the 200 Wikipedia employees, if only because of the ability of substantially higher salaries to attract substantially better people.

Here are some organizations that I believe fall in the broad reference class that the WMF does with respect to scope and influence on the Internet:

• Automattic is the company that runs WordPress.com, the platform where people can create their own WordPress blogs. It has also supported the creation and development of the open source and free WordPress software, though the software itself is managed by the WordPress Foundation. According to CrunchBase, Automattic has raised a total of $310 million since its launch in 2005, of which $120 million was through secondary markets in 2013, and $160 million was a venture capital round in 2014. Automattic is a for-profit company that makes its money by offering WordPress hosting for big websites (such as TechCrunch) and its freemium WordPress.com offering. WordPress powers over 60% of websites, and WordPress.com has an Alexa rank of 33 when I last checked. It has 303 employees.
• The Mozilla Foundation is the steward for the open-source web browser Mozilla Firefox (and other related products) which powers between 10% and 33% of web pageviews. It had an annual revenue of $300 million when last reported, most of it coming from a deal to have Google as the default web search engine (the deal partner is now Yahoo!, but revenue for the deal has not yet been reported).
• GitHub is widely used by open source projects to coordinate diverse contributions, as well as by companies, particularly small and middle-sized ones, to manage their codebases. The website had an Alexa rank of 95 when last checked. The company was largely self-financed to begin with but raised $100 million in 2012. It has 241 employees.

We see that the WMF's revenue and expense numbers are within the same ballpark as those of these other organizations that are doing comparably complex stuff with reach that is roughly comparable, albeit in different ways.

I don't have a clear assessment of the quality of WMF's software engineering. Considering that they've successfully kept the site functional and running, one can be reasonably confident that they're at least reasonably competent. But the value of many of the software innovation efforts is ambiguous at best. The basic MediaWiki software was written as far back as 2001 and its rate of change has been declining (it has been fairly stable since about 2013). This isn't necessarily a bad thing, since MediaWiki is a fairly good piece of software for something like Wikipedia. But insofar as it's great software, the credit goes to a team of people who built it long before the WMF started employing full-time software engineers.

Two of the main products of recent years of software engineering effort, the Visual Editor and the Media Viewer, met with significant revolts from many vocal sections of the user community. Whether this was because the software was poor or because the community (or relevant sections thereof) were resistant to change is unclear, but either way, given that the goals of these products were to improve engagement and participation by the community, the failure of the products speaks to the Foundation's software engineering work often not furthering its stated goals. That said, there are some clear successes: the mobile version of Wikipedia and Wikipedia Zero are well-executed and generally liked.

To take another example of more personal relevance: last year, I was interested in getting information on the total number of pageviews of Wikipedia pages I've created or edited. To its credit, the WMF publishes anonymized hourly logs that provide information on how many people viewed each page. In late 2007, Domas Mituzas created a tool, stats.grok.se, that processes these logs and allows people to get easy information on the number of pageviews of a particular page in a particular month. It's currently maintained by Wikipedia user Henrik. Note: This tool, which has been the main means for accessing analytics information for Wikipedia, wasn't created or maintained by the WMF, but by volunteers (the WMF has recently come up with its own tool, but I think it's still in beta, and Wikipedia's official statistics page still recommends stats.grok.se).

In the beginning of 2014, I would manually query stats.grok.se for the monthly pageviews of all pages I created and enter the results into a spreadsheet to get an estimate of my impact. Finally, I decided to automate the process of retrieving the data, and I created the Wikipedia Views website, where you can now get the breakdown by page and month for pages I created. When I made Wikipedia Views, I didn't have any experience managing a software project (though I had edited/tweaked pre-existing projects in the past), and I still managed to pull it off, over a few weeks of a few hours a week. Given the WMF's interest in attracting and maintaining contributors, and its having a full-fledged analytics team, I would have expected this sort of tool to have come from the WMF itself a while back. (Incidentally, MediaWiki itself, which is a far more complex project, was created by Magnus Manske over two weeks because he was interested in learning PHP and applying it to the wiki content creation problem).

But what surprised me recently, when I was drafting my post on the great decline in Wikipedia pageviews (something I noticed based on stats.grok.se data as viewed through my tool, and again something that I wasn't able to find any discussion of), was that the statistics on stats.grok.se (that are still listed as the best source for pageview statistics) do not include mobile pageviews. In fact, the WMF started publishing logs including mobile and Wikipedia Zero pageviews as recently as September 2014, but stats.grok.se is still using the old logs. This, despite stats.grok.se being cited and linked to as the canonical source of information on Wikipedia pageviews. And bot pageviews still aren't excluded, even though both the filtering out of bots and the inclusion of mobile pageviews have been on the agenda since as far back as February 2013.

There are many ways of interpreting these facts. Perhaps these analytics just aren't that important to the WMF or to their users, so they are not giving the analytics a lot of attention (my own interest in analytics is arguably quite atypical). Perhaps this takes a lot more resources than I think it does. Maybe a mere $50 million isn't enough, perhaps a $500 million annual budget is needed to get a quality analytics product. Whatever the case, I want to take this example to highlight that even with this large budget, many things that some might consider basic and relatively straightforward may end up not happening. So I think this is some evidence against the WMF being unusually efficient or good at getting things done (but it may not be evidence of incompetence per se). It also means that if your goal is to get some very specific project done, such as a new analytics tool, it's better to build it yourself or fund it directly, going through the WMF only if necessary.

2.3. Does the marginal and/or average donation contribute unambiguously to improving Wikipedia quality?

Once we know that the servers that host Wikipedia can keep running for a small fraction of the annual budget of the WMF, the best argument in defense of the WMF lies in the value add of the Foundation to the underlying software and to the content created.

I already noted above that the Visual Editor, the WMF's most high-profile software engineering effort to improve user participation, was rejected by the community after several months of effort. Apart from these software efforts, the Foundation has also tried a number of other techniques to increase user participation. These seem like valiant efforts, but they don't seem to have had a clear track record of success. For instance, efforts to improve article quality appear to have stalled because there is very little consensus on what article quality even means. There is much sporadic discussion about how to improve gender ratios in Wikipedia editing, but the end results are unclear.

I don't think the difficulty with improving content quality, or expanding the range of editors and the quality of contributions, are evidence of any unique incompetence on the WMF's part. These are very hard problems, and it's possible that they can't be solved even if a lot more resources (including high-quality thought) were poured into them. Any steps taken to attract new people run the risk of alienating the current contributor core (as happened with the Visual Editor).

So the pushback against the idea that donations to the WMF improve content quality isn't intended as a criticism of the WMF's competence, but rather of the idea that improving quality is a relatively easy task that the WMF is executing well.

Outgoing WMF chief Sue Gardner has said something similar:

"Too large a proportion of the movement's money is being spent by the chapters [whereas] the value in the Wikimedia projects is primarily created by individual editors: individuals create the value for readers, which results in those readers donating money to the movement. ... I am not sure that the additional value created by movement entities such as chapters justifies the financial cost," Gardner wrote.

Worse is the risk for trading favours, taking advantage of positions and troughing in general, she said. Gardner warns that the FDC [Funds Dissemination Committee] process is "dominated by fund-seekers, does not as currently constructed offer sufficient protection against log-rolling, self-dealing, and other corrupt practices. I had hoped that this risk would be offset by the presence on the FDC of independent non-affiliated members".

Gardner questioned whether bankrolling local chapters was the way to go - and urged the community to look for alternatives.

"With such a high proportion of movement resources now funding [chapters' staff and offices], we need to ask if the benefits are turning out to be worth the cost. It's possible that a well-managed shift to some staff support can help a volunteer community stay energized and enthused, but the risks and costs of setting up bricks-and-mortar institutions also dramatically increase, alongside sometimes difficult dynamics between staff and community."

She added: "I believe we're spending a lot of money, more than is warranted by the results we've been seeing. I am concerned by the growth rates requested by the entities submitting funding requests to the FDC."

2.4. Does Wikipedia helping you personally make it morally obligatory or desirable to donate to the WMF?

First off, as noted earlier, the WMF doesn't reward either past or future content creators. Past experiments with paid editing to improve quality didn't go well. It's not clear whether such efforts were inherently doomed or the execution was flawed. But whatever the case, it seems that payment of content creators is not part of the current agenda of the WMF.

I think there could be a good case for donating to the WMF based on a sense of moral obligation if the money were going to those whose services most directly helped you (the reciprocity/returning favors perspective), and/or if it were clear that the WMF's use of funds, either at the margin or on average, were the most effective ways of furthering the goals of building the better encyclopedia that you benefited from (the effective altruist perspective). Neither of these seem true to me.

As Bryan Caplan noted in his blog post reviewing Jason Brennan's book The Ethics of Voting:

2. There are many extrapolitical ways to exercise civic virtue and contribute to the common good:
[M]any activities stereotypically considered private, such as being a conscientious employee, making art, running a for-profit business, or pursuing scientific discoveries, can also be exercises of civic virtue.  For many people, in fact, these are better ways to exercise civic virtue.

In the same way, even if there is a generic obligation to "pay back" the world for the great resource called Wikipedia, that obligation can be fulfilled through many means other than donating to the WMF.

This isn't to say that there is anything categorically wrong about donating to the WMF as a way of discharging the (perceived) obligation. I don't think it's the most effective or sensible apprach, but it's not obviously wrong.

Aside: Andreas Kolbe has proposed that, rather than donating to the WMF, you seek out people to do paid editing and pay them if the quality of the work satisfies you, using the Wikipedia Review Board to find and communicate with the relevant editors. I don't have clear thoughts on this, but I do think that if your goal is to improve content quality or quantity on Wikipedia, editing it yourself -- or paying others to directly edit it through appropriate mechanisms (and not in violation of Wikipedia's guidelines) is probably a more cost-effective approach.

The English Wikipedia has a little under 5 million articles, and all the language Wikipedias together have between 10 and 20 million articles, with a growth rate of about 10% per annum for the English Wikipedia and perhaps a little higher for other Wikipedias. So if we think of half the money going to the WMF as effectively being for the English Wikipedia, that's an annual spend of $25 million for about 500,000 new articles, so the cost per new article comes to about $50 (of course, this cost calculation is problematic because much of the new article creation is unrelated to much of the WMF's spending). At this kind of cost, creating new content yourself might well beat out making a donation.

However, the "engage in or fund content creation on Wikipedia directly" argument is not essential to my main point here.

2.5. Is there a Kantian imperative to donate?

The Kantian moral imperative to donate is about seeking a good general rule that, if everybody (or sufficiently many people) followed, the world would be a better place. From this perspective, the rule "don't donate to the WMF" isn't a good rule, because if everybody followed that, the WMF would not have any money, and Wikipedia would not be able to survive. The Kantian imperative argument is not an argument about the practical consequences of your marginal decision to not donate, but about what general principles are best for the overall working of society.

A slight variant of this is the argument: "if every Internet user who has accessed Wikipedia gave $1 a year, then clearly everybody is donating far less than the value they get from the encyclopedia, but Wikipedia would have something like $500 million a year in revenue. So a $58 million annual budget is a bargain deal in comparison."

It is true that $1 seems like a trivial sum of money for most donors relative to the value they appear to get from Wikipedia. But this type of argument proves too much. We can come up with lots of things that one could use the "if everybody paid $1" argument. But if everybody followed this argument literally, then people would rapidly spend out all their money in these $1 tips, and the providers of the relevant services would be flush with cash that they don't have good ways to allocate. I believe that this has already happened to some extent with the WMF, but even if you don't think so, it seems reasonably clear that it would happen at 10X the scale.

Rather than adhering to the Kantian imperative, I think it's better to look at the issue from the perspective of room for more funding: is the WMF in danger of failing to meet its core functions that have been known to deliver value, or of exploring directions that I think are particularly promising? The answer for me is a clear no at the current margin. But if the answer changed, either because funding went down a lot or because they started taking up much more promising projects, then I'd reconsider my decision in light of the new evidence (and blog my modified views, to set straight the public record).

I think that there is some appeal of the Kantian imperative in that it avoids complicated game-theoretic coordination issues of the sort that GiveWell outlined in its December 2014 blog posts on coordinating the contributions of several large donors. And if the WMF fell in the category of contenders for a top place to donate to (i.e., it were something I think there are strong reasons to give to, but we are just constrained on cash and there are somewhat better opportunities) then I think the sort of coordinated solution GiveWell tried to orchestrate would be worthwhile. I don't think that's the case with the WMF.

3. Conclusion

I will reiterate that this post is a reasoned criticism of donating to the WMF but not intended as a smackdown (I believe many of the problems being tackled by the WMF are intrinsically hard). Rather, I am using this post to explore what I perceive as factual and epistemic flaws in particular reasons to donate to the WMF. Given the publication venue (the Effective Altruism Forum) I believe some potential donors who read this might find the information contained here sufficient to tip them away from considering donating to the WMF, but it is likely that the majority of readers will not change their decision (some will continue to donate, some will continue to not donate).

I hope this post is informative both regarding some of the facts surrounding the WMF's situation and regarding how somebody might use effective altruist principles in evaluating a particular giving opportunity. I'm interested in hearing other people's thoughts on both these counts.

PS: I had posted about my plans to write this blog post on the Effective Altruists Forum in December 2014. You can see the post and the comments on it here.

PS2: Related to point 4 above, I have been experimenting with sponsoring other people's Wikipedia contributions.

4. Addenda

4.1. ADDED December 2015: Remarks in light of renewed interest due to the 2015 fundraiser

In its December 2015 fundraising drive, the Wikimedia Foundation hopes to raise $25 million. Skepticism of the Wikimedia Foundation's fundraising drive seems to be getting more mainstream now, with the Washington Post using a title that might have been seen on The Register a few years ago: Wikipedia has a ton of money. So why is it begging you to donate yours? Despite the moderate increase in skepticism, I expect that the WMF will be able to successfully close its fundraising drive.

However, the inherent challenge in raising more money given declining pageviews has been noted both in fundraising meta discussions and in the Washington Post and other mainstream media coverage of the fundraising round. It is possible that this will place an effective ceiling on the amount of money that can be raised through fundraising drives in the coming years.

As for the criticisms I made in my post of arguments for donating to the WMF, I believe that they mostly remain intact. The WMF has made some progress in various areas, but it's still quite unclear to me if this progress is impressive relative to the financial and manpower resources at their disposal. To take an example I've sort-of-been keeping up with, they've made a number of improvements in their analytics (they now have an API endpoint to get desktop/mobile and user/bot/spider views of a page by date) which is quite useful to me personally. But it's not clear that this is an impressive accomplishment relative to the amount of time it took and the amount of financial resources available to the Wikimedia Foundation. Also, the extent to which the various initiatives around analytics pay off in terms of better content and more pageviews remains to be seen.

It's possible that I will revise my estimate of the WMF more favorably over the coming years, but as of now I still stand by what I wrote in my post.

4.2. ADDED March 2016: Remarks in light of the executive shake-up at the WMF

In February 2016, the Wikimedia Foundation revealed that it had applied for a grant from the Knight Foundation to build improved search functionality for the Wikimedia projects. This was consistent with WMF Executive Director Lila Tretikov's goal of making the organization focus on technology so as to facilitate greater engagement by editors and readers. The ambitiousness of the project and the opportunity cost of going down that route attracted negative attention and comment from both WMF employees and influential community members. In addition, the lack of prior disclosure about the grant was not taken kindly by people who expected a higher degree of transparency form the WMF. WMF Executive Director Lila Tretikov resigned with her effective departure date the end of March 2016. Vice has two good articles on the subject (here and here) and Molly White provides a comprehensive timeline of events.

As an outsider, I don't have a strong enough view regarding how realistic the search functionality plans are. To some extent, I can see merit in both sides. I agree with Tretikov, and others, that the decline in traffic to Wikipedia is a significant issue for the future of Wikipedia and, by extension, the Wikimedia Foundation. In fact, I wrote a blog post about this decline a year ago. Given that at least part of this decline appears to be because people searching on Google are presented with knowledge cards on Google itself and therefore don't even click through to Wikipedia, providing a better in-site search experience could, potentially, address the traffic decline. It's not clear to me that this is necessarily a worse idea than others. I can also understand somewhat the need for secrecy in the initial stages of such a plan, given the potential both to alert competition about it and to alienate people within the community even before any details had been fleshed out.

But that said, given that the Wikimedia Foundation's software engineering team had not, in recent years, done anything of comparable complexity to designing a full open search engine, and the fact that similar projects by other companies took huge amounts of resources, skepticism is in order.

But regardless of which side you take in this debate, the attempt to build a search engine, and the subsequent executive fallout, suggest that it would in general be hard to do ambitious things with Wikipedia. You could argue that this is because a visionary executive director is hampered by a bureaucratic organization and community, or you could argue it is because the Executive Director lacked vision or the ability to cooperate with the community to move things forward. I think this on the whole supports my rebuttal on (2). Wikipediocracy, a website whose content I mostly don't endorse but that still provides an interesting genre of Wikipedia criticism, sums up the situation by saying that unicorns are a myth.