Search and destroy: the Knowledge Engine and the undoing of Lila Tretikov
This article was originally posted on the author's blog and is republished with his permission. The views expressed in this article are his alone and do not reflect any official opinions of this publication.
The Wikimedia Foundation is in open revolt. While the day-to-day volunteer efforts of editing Wikipedia pages continue as ever, the non-profit Foundation, or WMF, is in the midst of a crisis it’s never seen before. In recent weeks, WMF staff departures have accelerated. Within just the past 48 hours, employees have begun speaking openly on the web about their lack of confidence in the leadership of executive director Lila Tretikov.
All in all, it’s been a terrible, horrible, no good, very bad start to 2016. Controversy in the first weeks of the year focused on the unexplained dismissal from the WMF Board of Trustees of James Heilman, a popular representative of Wikipedia’s volunteer base, before shifting to the unpopular appointment to the WMF Board of Arnnon Geshuri, whose involvement in an anti-competitive scheme as a Google executive led him to resign the position amidst outcry from the staff and community (see previous Signpost coverage).
But other issues remained unresolved: WMF employee dissatisfaction with Tretikov was becoming better known beyond the walls of its San Francisco headquarters, while questions mounted about the origin, status and intent of a little-known initiative officially called Discovery, but previously (and more notoriously) known as the “Knowledge Engine”. What was it all about? How do all these things tie together? What on Earth is going on here?
Deep breath.
The strange thing about the Knowledge Engine is that, until very recently, basically nobody knew anything about it—including the vast majority of WMF staff. Not until Heilman identified it as a central issue surrounding his departure from the Board had anyone outside the WMF staff ever heard of it—though in May 2015, Risker observed that a team called “Search and Discovery” was “extraordinarily well-staffed with a disproportionate number of engineers at the same time as other areas seem to be wanting for them”. This despite the fact that, as we know now, the WMF had sought funding from the Knight Foundation of many millions of dollars, receiving just $250,000 in a grant not disclosed until months later. As recently as this month, a well-considered but still in-the-dark Signpost article asked: “So, what’s a knowledge engine anyway?”
After several months of not knowing anything was amiss, followed by weeks of painful acrimony, we think we have the answer: as of February 2016, the mysterious project is in fact a WMF staff-run project to improve Wikipedia’s on-site search with some modest outside funding, which sounds like a good idea; sure, Wikipedia’s on-site search engine isn’t maybe the best, but we also know at some point it was an ambitious project to create a brand new search engine as an alternative to Google. Sometime in 2015, the WMF submitted a proposal to the Knight Foundation asking for a substantial amount of money to fund this project. It is described in still-emerging documents from this grant request as a “search engine”, and several early mock-ups seemed to suggest this was in fact the idea, as seen in internal documents revealed for the first time by the Signpost last week.
Why would Wikipedia consider building a search engine, anyway? The most likely answer is fear of being too dependent on Google, which sends Wikipedia at least a third of its total traffic. In recent years, Google has started providing answers to queries directly on the search engine results page (SERPs), often powered by Wikipedia, thereby short-circuiting visits to Wikipedia itself. Tretikov herself, in a rambly January 29 comment on her Meta-Wiki account page, identified “readership decline” as Wikipedia’s most recent challenge.[1]
It’s an understandable position: if you are the leader of an organization whose success has been largely described in terms of its overall traffic, any decline in traffic may be equated with a decline in Wikimedia’s ability to fulfill its mission. I submit this is short-sighted: that Wikipedia has an educational mission whose impact cannot be measured solely in terms of traffic. That Google borrows information from Wikipedia—though they are not alone in this—in such a way that it answers people’s questions before they have to actually click through to en.wikipedia.org is still a win for Wikipedia, even if it reduces the (already low) probability that a reader will become a Wikipedia contributor.[2]
The logic is twisted, but you can follow it: most readers find Wikipedia through a search engine, so if the search engine that helped make Wikipedia the success it is today changes its mind and starts pointing elsewhere, better to get ahead of things and create a new alternative that people will use. I guess? If we accept this reasoning, we still have to confront questions like: Is this actually something the WMF can accomplish? Is this within the WMF’s scope? Is this something that will help Wikipedia accomplish its mission? These are much harder questions for WMF to answer—in part because the answers are “no”, “no”, and “no”—and would absolutely have to be shared with the Wikimedia Board of Trustees ahead of time and, for political reasons, socialized within the Wikipedia community itself. The incident surrounding Heilman’s departure suggests the former was an issue, and the ongoing furor is because the latter obviously did not occur.
Meanwhile, the extreme unwillingness of Lila Tretikov and even Jimmy Wales to talk about it is, in fact, tearing the Wikimedia Foundation apart. Tretikov has lost all remaining credibility with Wikimedia staff and close community observers, not that she had much to begin with. As this week comes to an end, more staffers are quitting, remaining ones are complaining in public, and it seems impossible to imagine Lila Tretikov remaining in charge much longer.
If you need a detailed timeline of events to understand how we got here, I am pleased to say you’ll find just what you’re looking for below, although I’m afraid this whole thing is too large and multifaceted to do proper justice within the space of this already very long article. A full accounting may go back to the mid-2000s, when Jimmy Wales harbored ambitions of building his own search engine—Wikiasari in 2006 and Wikia Search in 2008. It certainly would include a full accounting of the many high-profile WMF staffers to leave since late 2014, and the role Tretikov played in each. It would include a careful examination of what the WMF can and should do in Wikipedia’s name, and an evaluation of how the evolving app-focused Internet raises questions about Wikipedia’s own future.
I think that’s more than I can accomplish here.
Instead I want to focus on what’s happening this week. But first we have to fill in some of the blanks. To do so, you’ll want to wind back the clock a few weeks:
- Let’s start on January 25, when Jimmy Wales called Heilman’s claims that transparency issues were at the core of his dismissal “utter fucking bullshit”. Jimmy Wales is known for occasionally lashing out at pestering editors on his Talk page, and this certainly seems to be one of those times.
- On January 29 Tretikov made her first public, community-facing statement about the Knight Foundation grant, which was welcomed for showing some self-reflection but also raised more questions than it answered.
- On February 1 WMF developer Frances Hocutt stated, on Tretikov’s discussion page no less, that employees were being “censured for speaking in ways that I have found sharply critical but still fundamentally honest and civil”.
- Don’t skip the aforementioned “So, what’s a knowledge engine anyway?” investigation by Andreas Kolbe for The Signpost, published February 8, still the most comprehensive evaluation of this multifaceted controversy.
- We then jump ahead to February 11, when Wales was still doing his “Baghdad Bob” routine, publicly insisting to Wikipedia editors that any suggestion WMF had ever considered building a search engine was “a total lie”.[3]
- Just hours later, WMF Communications uploaded the Knight Foundation grant agreement itself to the WMF’s own wiki, confirming for the first time, in public, that WMF was describing the project as “the Internet’s first transparent search engine”. The Signpost has the most detailed breakdown not only of the grant agreement, but also three supplemental documents which were leaked to the Signpost but have not been made public at this time.
- Also read this powerfully-argued blog post by Wikipedia veteran Liam Wyatt (Wittylama) about the poor strategic decision-making that led to the current controversy.[4]
- You might then have a look at The Register, always snarky, but with a decent summary of where things stood last week, just before it became newsworthy. I definitely recommend this February 15 story by Vice’s Motherboard about the fiasco (and this follow-up) but skip this Newsweek story except to see how the media was, for a brief moment, cluelessly reporting that Wikipedia was taking on Google.
- However incomplete, I think this upshot from The Verge is a good enough summary, at least for public purposes:
Whether Wikimedia’s plans just naturally evolved [away from the search engine project] or whether it was responding to the community’s response is difficult to say, but the organization is now, at least, claiming it does not want to square up to Google, but just improve its own product.
- As all this was unfolding, the exodus of key WMF staff was accelerating. On February 8, Tretikov announced on Wikimedia-l that Luis Villa, head of the Community Engagement department and previously a member of the WMF’s legal team, would be leaving.
- At least Tretikov seemed to be in control of that one. Because the next day Anna Koval, a manager of the education program, announced her own departure on the mailing list.
- And then on Friday, February 12, a very big resignation letter dropped on the Wikimedia-l mailing list: that of Siko Bouterse, another veteran leader who had long provided a crucial link between the Wikipedia volunteer community and the professional WMF staff. Careful with her words, Bouterse wrote:
Transparency, integrity, community and free knowledge remain deeply important to me, and I believe I will be better placed to represent those values in a volunteer capacity at this time.
- Messing up my timeline a bit, but still worth noting: Hocutt, the developer who had made public internal fears about silencing dissent, announced her own (albeit temporary) departure in yet another Wikimedia-l post on February 17, noting her leave was “due in part to stress caused by the recent uncertainty and organizational departures.”
Finally, on February 16, Lila Tretikov published an open letter (co-authored by Vice President of Product Wes Moran) on the Wikimedia blog titled “Clarity on the future of Wikimedia search”. Alas, it wasn’t terribly clarifying: it seemed aimed at the clueless mainstream journalists like the one at Newsweek, and not at the Wikipedia community who knew which information gaps actually needed to be filled in. It began:
Over the past few weeks, the Wikimedia community has engaged in a discussion of the Wikimedia Foundation’s plans for search and discovery on the Wikimedia projects.
Well, that is certainly one way to put it! Put another way, you have been backed into a corner defending the untenable proposition that Wikipedia has never considered building a search engine, and now that the mainstream press is reporting, based on your own documents, that you are building a search engine, one certainly has to say something about it.
After much boilerplate about the growth of Wikipedia and its many achievements, Tretikov and Moran finally get around to the point:
What are we not doing? We’re not building a global crawler search engine. We’re not building another, separate Wikimedia project. … Despite headlines, we are not trying to compete with other platforms, including Google.
This seems to be true, insofar as there is no search project currently. However, Wales had previously locked himself into the position that there was never a search project originating from WMF, and by now we know that is obviously false. Without any acknowledgement in this letter, it is useless. But it’s worse than that:
Community feedback was planned as part of the Knowledge Engine grant, and is essential to identifying the opportunities for improvement in our existing search capacity.
We are 10 months past the initial plans for this far-reaching, mission statement-busting project, six months past the award of a grant to pursue this quixotic effort, and not two months removed from the violent ejection of a Board trustee over the matter… and all you can say is “feedback was planned”?
Finally, the closest thing to acknowledging the Knowledge Engine was, at some point, actually a search engine:
It is true that our path to this point has not always been smooth, especially through the ideation phase.
And nothing more.
The first comment on the post was brutal, bordering on uncivil, from a retired editor. It concluded:
You are either:
a. Flat out lying, and hoping we don’t actually read the grant,
b. Have misled the Knight Foundation as to your intentions for their grant money, or
c. Seriously incompetent and should never be put in charge of writing a grant application
None of these options look good for the WMF.
A few hours later, a member of WMF’s Discovery team gamely stepped forward and tried to offer a plausible explanation for how the grant request did not necessarily imply a Google-competitive search engine project—damage control, essentially—but still had to concede the wording of the grant did not make Tretikov or WMF look good: “It is ambiguous. I can’t speak to the intent of the authors and while there are current WMF staff listed, they are not the sole authors of the document.”
Finally, a day later a true hero emerged in Max Semenik, another Discovery team engineer, mostly unknown to the community, and who was willing to take off his PR hat to say what everyone pretty much knew:
Yes, there were plans of making an internet search engine. I don’t understand why we’re still trying to avoid giving a direct answer about it. …
The whole project didn’t live long and was ditched soon after the Search team was created, after FY15/16 budget was finalized, and it did not have the money allocated for such work … However, ideas and wording from that search engine plan made their way to numerous discovery team documents and were never fully expelled. …
In the hindsight, I think our continued use of Knowledge Engine name is misleading and should have ended when internet search engine plans were ditched. No, we’re really not working on internet search engine.
Now that sounds like a real answer! What’s more, it also provides the outlines of a believable story as to why the Knight Foundation grant included language about the search engine, even if it wasn’t then the plan. This is transparency of a sort! But it’s transparency of the last-ditch kind. That it had to come from a low-level engineer indicates there is a major problem, and speaks to the fact that the WMF simply cannot go on this way.
At a time when Wikipedia has already-existing problems, the WMF was asking for money to basically create a whole new set of problems. That is the mark of an organization, if not a movement, adrift. Clearly, they pitched a search engine to Knight, and they asked for millions—I have heard the number placed at $100 million over 5 years—later reduced to $12 million, of which Knight provided $250K to build a plan—essentially a pat on the head: ‘since we like you, here’s a few bucks to come up with a better idea’.
Mysteries remain: where did the idea come from, who championed it, when did it die—or when did it recede and what happened afterward? One answer is supplied in another comment on this public thread (!) from yet another WMF team member (!) pointing a finger at former VP of Engineering Damon Sicore as having “secretly shopped around grandiose ideas about a free knowledge search engine, which eventually evolved into the reorg creating the Discovery team.” Sicore left in July 2015. A big remaining question, for which there is no answer at this time: when the actual grant was submitted to the Knight Foundation.
An argument I have heard in recent days is that it’s common in grant-making to try for everything you can and see what actually sticks. This may be true, but if so, it doesn’t seem to have been worth it. That WMF leadership felt they had to hide the fact later on also underlines the mistake they knew they were making.
Another big question: how does this affect Wikipedia’s public reputation, particularly among donors, most especially among foundations? You have to think the answer is a lot. The WMF looks like the Keystone Kops. Why would you give it money? And right now, the Knight Foundation specifically must be asking what it’s got itself into.
Within the last 24 hours, the trickle of public criticism about Tretikov has become a widening stream. Some of it is taking place in the above comment thread, plenty is still happening at Wikimedia-l, but a lot of it has moved to a semi-private Facebook group called Wikipedia Weekly, where staffers previously not known for voicing internal dissent have been speaking quite frankly about how bad things are at 149 New Montgomery Street.
Yesterday afternoon on the mailing list, a developer named Ori Livneh replied to a plea for calm by community Board trustee Dariusz Jemielniak by explaining why they could not remain silent:
My peers in the Technology department work incredibly hard to provide value for readers and editors, and we have very good results to show for it. Less than two years ago it took an average of six seconds to save an edit to an article; it is about one second now. (MediaWiki deployments are currently halted over a 200-300ms regression!). Page load times improved by 30-40% in the past year, which earned us plaudits in the press and in professional circles.
…
This is happening in spite of — not thanks to — dysfunction at the top. If you don’t believe me, all you have to do is wait: an exodus of people from Engineering won’t be long now. Our initial astonishment at the Board’s unwillingness to acknowledge and address this dysfunction is wearing off. The slips and failures are not generalized and diffuse. They are local and specific, and their location has been indicated to you repeatedly.
Shortly thereafter Asaf Bartov—one of WMF’s more outspoken staffers, even prior to the last 48 hours—voiced his agreement and turned his comments back to Jemielniak:
Thank you, Ori. +1 to everything you said. We have been laboring under significant dysfunction for more than a year now, and are now in crisis. We are losing precious colleagues, time, money, *even more* community trust than we had previously squandered, and health (literally; the board HR committee has been sent some details). Please act. If for some reason the board cannot act, please state that reason. Signal to us, community and staff, by concrete words if not by deeds, that you understand the magnitude of the problem.
And then, about 10 minutes later, Lila Tretikov posted to this very conversation thread, and this is all she had to say:
For a few 2015 accomplishments by the product/technical teams you can see them listed here:
https://meta.wikimedia.org/wiki/2015_Wikimedia_Foundation_Product_and_Technology_Highlights
That is the complete text of her emailed post. That is really all she had to say, in a public thread specifically criticizing her leadership and all but explicitly calling for her removal. One gets the feeling, at this point, even Lila Tretikov just wants it to be over.
In the early morning hours of February 19, a WMF software engineer named Kunal Mehta wrote an impassioned, rather forlorn post on his personal blog, titled: “Why am I still here?”:
Honestly, I don’t understand why the current leadership hasn’t left yet. Why would you want to work at a place where 93% of your employees don’t believe you’re doing a good job, and others have called you a liar (with proof to back it up) to your face, in front of the entire staff? I don’t know everything that’s going on right now, but we’re sick right now and desperately need to move on. …
I love, and will always love Wikimedia, but I can’t say the same about the current state of the Wikimedia Foundation. I’ve been around for nearly nine years now (nearly half my life), and it feels like that world is slowly crumbling away and I’m powerless to stop it.
And that’s why there is really just no way Lila Tretikov can continue to lead the WMF. A week ago, the thinking was: the Board of Trustees chose her over James Heilman, so they’re really sticking with her. At the time it also seemed like the Knowledge Engine was a going concern, and their support for her owed to their insistence on moving ahead with the project above community and staff objections. Knowing what we do now, it’s inexplicable. The thinking now is: she obviously has to go, and the only reason the Board might have for not acting on it would be legal considerations.
For the sake of Wikipedia’s future, the Wikimedia Foundation needs new leadership. Lila Tretikov must resign, or she must be replaced. This is the most challenging article I’ve ever had to write. The next one, I hope, will be about the start of the turnaround.
- Notes
- ^ “Our aim was to begin exploring new initiatives that could help address the challenges that Wikipedia is facing, especially as other sources and methods arise for people to acquire knowledge. If you haven’t yet, please have a look at the recent data and metrics which illustrate the downward trajectory our movement faces with readership decline (since 2013), editor decline (since 2007, which we stabilized for English Wikipedia in 2015), and our long standing struggle with conversion from reading to editing. These risks rank very high on my list of priorities, because they threaten the very core of our mission.”
- ^ See this comment from WMF’s Dario Taraborelli, who argues: “[T]raffic per se is not the goal, the question should be about how to drive back human attention to the source”.
- ^ Full quote: “To make this very clear: no one in top positions has proposed or is proposing that WMF should get into the general “searching” or to try to “be google”. It’s an interesting hypothetical which has not been part of any serious strategy proposal, nor even discussed at the board level, nor proposed to the board by staff, nor a part of any grant, etc. It’s a total lie.”
- ^ “It seems to me extremely damaging that Lila has approached an external organisation for funding a new search engine (however you want to define it), without first having a strategic plan in place. Either the Board knew about this and didn’t see a problem, or they were incorrectly informed about the grant’s purpose. Either is very bad.”
Discuss this story
75%the top 25% stays the same, everyone above that gets cut down tothe 75th percentilewhat the lower 75% make. If some major personalities want to leave, great - if not, we can reevaluate and try a bigger cut. The problem with laying out big bucks for decision makers is they make decisions, and we don't really want people making decisions, we just want them closing out tasks in Phabricator that have been sitting around unanswered for years on end. The problem with laying out big bucks for career software designers is they want to make careers, have some big piece of software with a fancy name they can say they built for a top website, not say they closed out a bunch of miscellaneous user requests.The community consultation was particularly effective in highlighting the needs and wants of our readers, as well as those of the editing community, and this has informed the ongoing strategy design process - a process that has deep community input. That strategy, in turn, informs funding decisions.
The Community Resources Team surveyed the community and discussed with them their technical priorities, and tailored their current Idea Lab Campaign accordingly.
The WMF have accepted the FDC's proposal that the WMF submit to the same reporting standard they expect of their chapters.
Lila could and should have been more candid about the Knowledge Engine project as the idea was evolving, and I hope she's learned from that, but under her the WMF has developed institutional structures that are intrinsically respectful of and responsive to the volunteers and readers. I hope she survives this crisis but, if she doesn't, I hope those structures do. --Anthonyhcole (talk · contribs · email) 02:27, 20 February 2016 (UTC)[reply]
Here’s some very important specifics pertaining to the grant process:
The widely circulated grant agreement PDF was not written by the WMF. It was written by junior staff members at the Knight Foundation. Here’s how it works: you have a conversation with staff at the Knight Foundation about ideas you have for program areas you’d like to fund, throw these ideas back and forth through a phone call or two, and then send the Knight Foundation a 1 page summary of what you’re looking to do. Knight junior staff then turns this into a document that they send to their VP, and then once the VP signs it, it’s done.
Importantly, the language on the grant agreement is not written by the grantee, but instead by the Knight Foundation, and usually by junior staff. I could show you some of our old grant agreements, and you’d be blown away by how “off” the language is on the agreement versus on the proposals we sent in. The grant agreement language is designed to be informal, and is written largely based on conversations.
Given the funding amount of $250k, this was *not* a long, drawn out grant process. This grant must have gone from “quick first chat” to “grant agreement” in a week or less. Grants of less than 250k are not approved by the Knight Foundation Board and are instead approved by VPs. They happen very quickly. This is likely why many people at the WMF felt blindsided.
The last Knight Foundation grant I got took two weeks from conversation to grant agreement. The final text in the grant agreement was written by staff at the Knight Foundation and had several important mistakes present in it.
I’m not sure why senior WMF didn’t explain this more clearly. My best guess is they didn’t want to malign the Knight Foundation. Nothing about this grant process seems incompetent to me.
I could not disagree more with your call for Lila Tretikov’s resignation. It’s completely ridiculous. This is just a growing pain associated with WMF applying for foundation funding, something they’ve only done a literal *handful* of times in the past. Grant funding is an extremely important part of WMF being able to innovate outside of existing budget areas. Grants are also typically opaque. I don’t think the WMF has applied for many grants, at least based on its size. I’m willing to bet that no other nonprofit the size of WMF has taken as little grant money as WMF has. So this seems like both a growing pain and a necessary growing pain. 2600:1010:B001:974D:A1FF:B07E:3146:626B (talk) 12:20, 20 February 2016 (UTC) (not logged in so as to not burn my own grant relationship bridges)[reply]
A story in desperate search for a conspiracy. Sorry, but this is completely out of proportions. Jeblad (talk) 22:18, 20 February 2016 (UTC)[reply]
Nov. 20 Grant for payment within 60 days and the announcement
The timeline some are discussing seems a little off. The Knight Grant was not completed until it was signed on November 20, 2015 (November was also when the Wikimedia FAQ page was created [2]). They then had 60 days to payout, and during that 60 days Knight made the joint announcement, which says it is:
So, regardless of the boilerplate in the agreement - Knight and WMF clearly have a common understanding that it is for early research into the Wikipedia.org search function of Wikipedia (aka., Wikimedia) projects, not any sort of Google. -- Alanscottwalker (talk) 22:24, 20 February 2016 (UTC)[reply]
"I strongly object to the notion that I have been unwilling to discuss the Knight grant, "knowledge engine" or anything else. There is no evidence offered for this because it simply isn't true.--Jimbo Wales (talk) 13:40, 20 February 2016 (UTC)"[reply]