Wikipedia:Wikipedia Signpost/2016-04-24/Op-ed
Knowledge Engine and the Wales–Heilman emails
The Wikimedia Foundation board's communications in the wake of the removal of the community-selected trustee James Heilman have consisted mainly of silence. Jimmy Wales has been the most notable exception, having made strongly worded statements along with promises to provide further information – promises that have remained unfulfilled.
The "gaslighting" email
In recent weeks, onwiki debates on the Knowledge Engine and Heilman's dismissal have largely died down. The most recent substantial discussion took place three weeks ago on Wales' talk page, when Wales set out to explain certain comments he had made in a February 29 email sent to Heilman and Pete Forsyth, shortly after the resignation of executive director Lila Tretikov.
As Signpost readers will recall, Forsyth took the controversial decision to forward Wales' mail to the Wikimedia-l mailing list. Forsyth felt it could provide "important insight into the dynamics surrounding Heilman's dismissal". In the ensuing debate, a number of people pronounced themselves horrified by the tone and content of Wales' email, likening it to "gaslighting" (a form of mental abuse). Others criticised Forsyth for his decision to publish it.
WMF strategy consultant brings background in crisis reputation management; Team behind popular WMF software put "on pause"
6 February 2017
Knowledge Engine and the Wales–Heilman emails
24 April 2016
[UPDATED] WMF in limbo as decision on Tretikov nears
24 February 2016
Search and destroy: the Knowledge Engine and the undoing of Lila Tretikov
17 February 2016
New internal documents raise questions about the origins of the Knowledge Engine
10 February 2016
An in-depth look at the newly revealed documents |
In his email, Wales cast a string of aspersions on Heilman's character before taking particular issue with a February 24 post by Heilman in a Wikipedia Weekly Facebook discussion of the Knowledge Engine project.
The Facebook discussion
Heilman's Facebook comment had a context. In the discussion (accessible only to logged-in Facebook users), Liam Wyatt said he was unsure that Wales could be characterised "as having been 'kept in the dark'" about the Knowledge Engine project. "James has said that the board as a whole was presented with these plans – that it was described as 'a moonshot' and that they were presented with cost estimates in the tens-of-millions," Wyatt added, pinging Heilman in his post. Heilman then replied minutes later that he had indeed asked Board members in October whether they understood "that we were building a 'search engine' as before Oct I did not realize we were. JW said that he understood this all along and it was something we needed to do."
The post appears to have angered Wales. In his email, he wrote to Heilman:
Attentive readers will note that the phrase "Google-competing search engine" appears nowhere in Heilman's post. Heilman was responding to a post that said there was a search engine project that the board was told would cost tens of millions of dollars.
Selective quoting
When Peter Damian challenged Wales to dig up the exact quote, Wales produced it, and to back up his point published excerpts from the October email conversation, with selected quotes from Heilman and himself.
Heilman asserted that Wales' summary of the exchange was "far from complete", and "not an accurate representation of the overall discussion". He asked Wales whether he would have any objection to the complete exchange being posted, so the parts Wales had quoted could be seen in context.
Wales raised no such objection, and the full exchange, as made available to the Signpost by Heilman, is published below. It shows that the accusations Wales levelled at Heilman for his Facebook post were groundless and contrived. In the actual conversation, Wales said to Heilman that –
- the ambitious vision of a search engine project as presented to the Knight Foundation, offering "a unique search experience that will go beyond what Google and Bing are already providing their users", corresponds exactly to what was approved by the board;
- he is "broadly supportive" of that strategic vision;
- the scope of the project goes well beyond a Wikimedia-internal search function;
- it includes building a natural-language question answering system akin to Google's Knowledge Graph and answer boxes;
- the project is motivated by a desire to compete with Google: "users don't come to us", Wales said, because "Google just tells the answers";
- the project is a very major financial investment for the Foundation (Wales later confirmed onwiki that it was in the ballpark of $35 million, spread out over several years).
At the same time, Wales omitted to mention in his summary the concerns put forward by Heilman about the cost and scope of this long-term project, and the WMF's qualifications for undertaking it.
One way to look at this situation is that Wales has essentially been launching vigorous attacks on a strawman – the idea that the Foundation might be intending to build a search engine that does all the things Google does: crawling and indexing everything from books, journals and newspapers to social media sites, online shops and cinema schedules. But his apparent single-mindedness in pursuing this strawman cannot make up for the fact that this is not something Heilman has ever claimed. What Heilman did claim was that the Foundation was planning to build a search engine that would cost tens of millions of dollars. In that, he was undoubtedly correct.
The complete October email exchange
The passages Wales quoted on his talkpage are in green. Salient parts Wales omitted from his summary are in bold red.
This statement alone, omitted by Wales in his summary, seems ample justification for what Heilman wrote on Facebook. The exchange continued:
"Our entire fundraising future is at stake"
A comment Wales made in November 2015 in a three-way email discussion between Wales, Heilman and a WMF staffer sheds further light on his thinking. Wales responded as follows to the assertion that there clearly had been an attempt to fund a massive project to build a search engine that was then "scoped down to a $250k exploration for a fully developed plan":
No such blog post was ever published by Wales, to the Signpost's knowledge. But the Knowledge Engine grant agreement – originally withheld by the board, ostensibly because of "donor privacy" issues, and only released after the Signpost confirmed with the Knight Foundation that there were no privacy issues on the donor's side – is more suggestive of the notion that there was indeed a plan, one on which the Wikimedia Foundation's "entire fundraising future" hinged, according to Wales.
This is hard to reconcile with what Wales told the community in February:
We see that when Heilman said in the above email conversation that this was "not a search tool for Wikimedia properties", Wales readily agreed, stressing the importance of answer engine functions in attracting users that today find their answers on Google. But to the community, Wales has been keen to convey the opposite impression, narrowly focusing on the project's first phase only:
- "The project presented to the board at the Wikimania board meeting in Mexico was about improving internal search and discovery, with some very reasonable and modest first steps outlined."
- "I don't think of improving internal search and discovery to be primarily about revenue and page views."
- "Perhaps you have a typo, or perhaps just a continued misunderstanding. It was clear to everyone (on the board) involved with the grant that this (the work to be funded by the grant) was just an internal search tool. No code, no architecture, no nothing other than a vague idea that maybe someday Wikipedia could also include other "non-commercial" results someday."
Wales specifically objected to the portrayal of the Knowledge Engine as something that would compete with Google. But in the exchange above, he himself twice emphasises that Wikipedia is failing to offer users the answers that Google is providing to them:
Referring to the Knowledge Engine grant agreement, Wales says, "I don't agree that there's a serious gulf between what we have been told and what funders are being told." Yet what funders were told was that "Knowledge Engine by Wikipedia will be the Internet's first transparent search engine, and the first one originated by the Wikimedia Foundation ... a system for discovery of reliable and trustworthy public information on the Internet ... a unique search experience that will go beyond what Google and Bing are already providing their users".
In the above email exchange, Wales also alludes to the possibility that "in due course", the Knowledge Engine project "might expand beyond just internal search (across all our properties)". In recent months, he has multiple times referred to the possibility that "non-WMF resources might be included in a revamped discovery experience" or that "some important scholarly/academic and open access resources could be crawled and indexed in some useful way relating to Wikipedia entries" while insisting that any suggestions "that this is some kind of broad Google competitor remain completely and utterly false."
In the "gaslighting" email, Wales also objects to the fact that Heilman included Wikia Search in a timeline published in the February 3 Signpost issue. But a key element of Wikia Search was "public curation of relevance" – volunteers determining how high up in Wikia's search results Internet pages should be ranked (a process that at times led to hilarious results). And public curation of relevance is also a key element of the latter stages of the Knowledge Engine project, as outlined to the Knight Foundation and described in the official project documentation.
To be sure, the Knowledge Engine is not conceived as a full-fledged Google competitor, complete with shopping results, opening hours of shops and restaurants, cinema times, search results from Twitter and Facebook, and so forth (and Heilman never claimed it was).
But judging from the documentation available, it was – or is – conceived at the very least as a niche competitor to Google, crawling and indexing both Wikimedia properties and selected other Internet content and replicating Google's answer engine and Knowledge Graph functionality. When Jimmy Wales says that the Wikimedia Foundation's entire fundraising future depends on the idea, the hope surely is to draw a significant number of eyeballs to Wikipedia.org by providing answers to natural-language questions, following the lead of other AI assistants, and providing search result listings that take users to relevant pages anywhere in the Wikimedia universe, complemented by a broad range of open access and/or academic sources.
It is an ambitious idea, but not in any way objectionable in itself. What is clear however is that building such a search engine will cost tens of millions of dollars. Heilman's concern was that
- this was a major decision about the Wikimedia Foundation's long-term strategic direction that the community should be involved in,
- this was something that should be openly disclosed rather than kept secret,
- the financial investment required to undertake this ambitious project would lead to other projects being underfunded,
- if the project should fail to gain traction with users, this could result in tens of millions of donor dollars being wasted.
These were not idle concerns. And the fact that Heilman expressed them in no way justifies the repeated vilifications he has had to endure.
Discuss this story
This is the best-grounded look at the whole Heilman affair since it began, aided of course by the digging you folks at the Signpost have done and by the addition of the actual email chain between Wales and Heilman.
What a tale of technical overreach, fiduciary irresponsibility, behind-the-scenes machinations, treachery and duplicity!
Magnificent wordsmithing by Andreas Kolbe. →StaniStani 00:10, 25 April 2016 (UTC)[reply]
My compliments on another excellent piece of work, Andreas. You should really try to get these articles more widely distributed. -- Seth Finkelstein (talk) 01:28, 25 April 2016 (UTC)[reply]
Sorry but this is just a bunch of misconceptions. A query dialog engine is not a Google competitor, it is not even close. (Why do I waste time on reading this?;/) Jeblad (talk) 06:21, 25 April 2016 (UTC)[reply]
Some thoughts
Hi. Sorry if this is a daft question, but this piece is marked as an op-ed. What opinion is being expressed?
Does anyone disagree that our internal search needs improvement? I would think that Andreas and others would be supportive of efforts to have free, open, and independent search functionality. Below other mission-critical services such as providing SQL and XML data dumps, search is pretty important infrastructure, especially as the Wikimedia projects grow.
If we took an input string such as "How old is Tom Cruise?" and broke it up into pieces, I think we could, with some effort, program this and similar queries to return specific data points. We could look at the most relevant Wikidata item (d:Q37079) to extract the "date of birth" field's value ("3 July 1962") and then do a simple date calculation to show that Tom Cruise is currently 53 years old. Or, if we can get the search results to be better, we can pull out and highlight specific data points alongside the search results.
After we solve "How old is [famous person]?"-type queries, we can add support for alternate phrases such as "What age is [famous person]?" Once we solve that, we can move on to programmatically answering other "easy" queries. I don't think what's being described here requires artificial intelligence or IBM's Watson.
You want a concrete opinion? The search results at Special:Search/How old is Tom Cruise? are currently terrible. Tom Cruise bafflingly doesn't appear in the top 100 results. If Tom Cruise did appear in these results, we could look at the search input, see that it uses a known keyword ("age" or "old"), and then extract that information programmatically to serve our reader/researcher more quickly. Who opposes doing this?
Let's talk about how we can improve search and what that will require. Does an organization similar to the Wikimedia Foundation (or the Knight Foundation, for that matter) need to be involved? What value do these organizations provide? I think there's plenty of room for intelligent and thoughtful discussion about priorities and functionality and serving our readers. Can we start now? --MZMcBride (talk) 03:23, 26 April 2016 (UTC)[reply]
About Jimmy's behaviour and character
Was the related email, from Jimbo to Doc James of 30 December 2015 ever shared publicly? HolidayInGibraltar (talk) 19:10, 27 April 2016 (UTC)[reply]
Kudos to Andreas for an incisive and revealing exposé. Is Wales any longer appropriate as a WMF board member and self-appointed WP figurehead? Given the long, damaging record of evasions, obfuscations, manipulations, lies, misdirections, misrepresentations, distortions, and self-serving personal attacks, the answer couldn’t be more obvious. Writegeist (talk) 19:11, 29 April 2016 (UTC)[reply]
Data sources
"In recent months, [Jimmy] has multiple times referred to the possibility that 'non-WMF resources might be included in a revamped discovery experience' or that 'some important scholarly/academic and open access resources could be crawled and indexed in some useful way relating to Wikipedia entries' while insisting that any suggestions 'that this is some kind of broad Google competitor remain completely and utterly false.'"
Please don't forget Fox News appearing in a sample search result. --NaBUru38 (talk) 16:11, 29 April 2016 (UTC)[reply]