William Beutler on Wikipedia

Posts Tagged ‘Aaron Halfaker’

What You Missed at Wikimania 2017

Tagged as , , , , , , , , , , , , , , , , , , , , , , , ,
on August 18, 2017 at 4:39 pm

N.B. At the end of this post I’ve embedded a Spotify playlist for the delightful 2006 album “Trompe-l’oeil” by the Francophone Montreal indie rock band Malajube. It’s what I was listening to as I arrived at Montréal–Pierre Elliott Trudeau International Airport last week, and I think it would make a nice soundtrack for reading this post.

♦     ♦     ♦

Wikimania 2017, the thirteenth annual global meeting of Wikipedia editors and the larger Wikimedia movement, was held in Montreal last weekend. For the fifth time overall, and the first time in two years, I was there. I’ve covered previously attended Wikimanias, sometimes glancingly, and sometimes day-by-day, and this time I’ll do something a little different as well.

One nice thing about a conference for a project focused on the internet: many of the presentations can be found on the internet! Some but not all were recorded and streamed; some but not all have slides available to revisit. The second half of this post is a roundup of presentations I attended, or wished I attended, with media available so you can follow up at your own pace.

But first, a note on a major theme of the conference: implicitly if not specifically called “Wikimedia 2030”, and a draft of a “strategic direction” document circulating by stapled printout from the conference start, later addressed specifically in a presentation by Wikimedia Foundation executive director Katherine Maher and board chair Christophe Henner. It’s available to read here, and I recommend it as a straightforward and clearly-described (if detail-deficient) summary of how Wikimedians understand their project, and where its most dedicated members want to take it.

Draft strategic direction at Wikimania 2017As one would expect, the memo acknowledges the many types of contributors and contributions, brought together by a belief in the power of freely shared knowledge, and a committment to helping organize it. It also focuses on developing infrastructure, building relationships, and strengthening networks. One thing it doesn’t talk much about is Wikipedia, which might be surprising to some. After all, Wikipedia is arguably more important to the movement than the iPhone is to Apple: Wikipedia receives 97.5% of all WMF site traffic, while the iPhone accounts for “only” 70% of Apple’s revenues.

I don’t wish to belabor the Apple analogy much, because there are too many divergences to be useful in a global analysis, but both were revolutionary within their markets, upset competitors, created a whole new participatory ecosystem in their wake, and each grew exponentially until they didn’t. Now the stewards of each are looking beyond the cash cow for new areas of growth. For Apple, it’s cloud-based Services revenue. For the WMF, it’s not quite as easily summarized. But the answer is also partly about building in the cloud, at least figuratively. Although both Wikipedia and the iPhone will remain the most publicly visible manifestations of each organization for the foreseeable future, the leadership of each is focused on what other services they enable, and how they can even make the core product more valuable.

I see two main themes in the memo, about how the Wikimedia movement can better develop that broad ecosystem beyond Wikimedia’s existing base, and how it can improve its underlying systems within movement technology and governance. The former is too big a subject to grapple with here, and I’ll share just a single thought about the latter.

One thing the document concerns itself with at least as much as with Wikipedia is “data structures”—and this nods to Wikidata, which has been the new hotness for awhile, but whose centrality to the larger project is becoming clearer all the time. Take just one easily overlooked line, about how most Wikimedia content is “long-text, unstructured articles”. You know, those lo-fi Wikipedia entries that remain so enduringly popular. They lack structure now, but they might not always. Imagine a future where Wikidata provides information not just to infoboxes (although that is a tricky subject) but also to boring old Wikipedia itself. Forget “red links”: every plain text noun in the whole project may be connected to its “Q number”. Using AI and machine learning, entire concepts can be quickly linked in a way that once required many lifetimes.

At present, Wikipedia is the closest thing we have to the “sum of all human knowledge” but in the future, it may only be the default user interface. Now more than ever, the real action is happening behind the scenes.

♦     ♦     ♦

Birth of Bias: implicit bias’ permanence on Wikipedia

Wikipedia is a project by and for human beings, and necessarily carries the implicit biases of those human beings, whether they’re mindful of the fact or not. This presentation, offered by San Francisco State visiting scholar Jackie Koerner, focused on how to recognize this and think about what to do about it. Slides are accessible by clicking on the image below, and notes from the presentation are here.

Koerner Implicit Bias Wikimania 2017

♦     ♦     ♦

Readership metrics: Trends and stories from our global traffic data

How much do people around the world look at Wikipedia? How much do they look at it on desktop vs. mobile device? How have things changed over time? All of this and more is found in this presentation from Tilman Bayer, accessible by clicking through the image below.

Readership metrics. Trends and stories from our global traffic data (Wikimania 2017 presentation)

♦     ♦     ♦

The Internet Archive and Wikimedia – Common Knowledge Goals

The Internet Archive is not a Wikimedia project, but it is a fellow nonprofit with a similar outlook, complementary mission and, over time, increasing synergy between the two institutions. Every serious Wikimedian should know about the Internet Archive. I didn’t attend the presentation by Wendy Hanamura and Mark Graham, but there’s a lot to be gleaned from the slides embedded below, and session notes here.

♦     ♦     ♦

State of Video in the Wikimedia Movement

You don’t watch a lot of video on Wikipedia, do you? It’s not for lack of interest on the part of Wikipedians. It’s for lack of media availability under appropriate licenses, technology and infrastructure to deliver it, and even community agreement about what kinds of videos would help Wikipedia’s mission. It’s an issue Andrew Lih has focused on for several years, and his slides are highly readable on the subject.

♦     ♦     ♦

The Keilana Effect: Visualizing the closing coverage gaps with ORES

As covered in this blog’s roundup of 2016’s biggest Wikipedia stories, one of Wikipedia’s more recent mini-celebrities is a twentysomething medical student named Emily Temple-Wood, who goes by the nom-de-wiki Keilana. Her response to each experienced instance of gender-based harassment on the internet was to create a new biographical article about another woman scientist on Wikipedia. But it’s not just an inspiring story greenlit by countless news editors in the last couple years: WikiProject Women Scientists, founded by Temple-Wood and Rosie Stephenson-Goodknight, dramatically transformed the number and quality of articles within this subject area, taking them from a slight lag relative to the average article to dramatically outpacing them. Aaron Halfaker, a research scientist at the Wikimedia Foundation, crunched the numbers using the new-ish machine learning article quality evaluation tool ORES. Halfaker presented his findings, with Temple-Wood onstage to add context, on Wikimania’s final day. More than just a victory lap, the question they asked: can it be done again? Only Wikipedia’s contributors can answer that question.

The slides can be accessed by clicking through the image below, notes taken live can be found here, and for the academically inclined, you can also read Halfaker’s research paper: Interpolating Quality Dynamics in Wikipedia and Demonstrating the Keilana Effect.

Keilana Effect (Wikimania 2017)

That was fun! Let’s do this again next year.

Update: Looking for more slides and notes? There’s an “All Session Notes” page on the Wikimania site for your edification.

♦     ♦     ♦

From the Annals of Bad Wikipedia Commentary

Tagged as , , , , ,
on January 4, 2016 at 9:30 am

Wikipedia’s ubiquity in Internet culture is matched by its inscrutability to those looking from the outside in. This makes it an attractive topic for occasional visitation by journalists and public intellectuals alike, but it is not the easiest subject to write about. Bad Wikipedia journalism is abundant, although in my experience most journalists will try to learn something about Wikipedia before covering it.

Alas, the same does not seem to be true for Internet pundits.

This brings us to the long holiday weekend, when two widely-followed Internet writers found a real study—“The Rise and Decline of an Open Collaboration System: How Wikipedia’s reaction to popularity is causing its decline”—by very wiki-knowledgeable academic researchers, linked to it for readers, added some well-meaning commentary, and curiosity-provoking headlines, and actually caused their readership to become less informed about the current state of Wikipedia as a result.

First up, “The rise and decline of Wikipedia?” by economist Tyler Cowen at his long-running blog, Marginal Revolution:

Halfaker, Geiger, Morgan, and Riedl have a new paper on this topic … This is an interesting paper, but I think it undervalues the hypothesis that potential contributors simply prefer to be in on things which are both new and cool. Wikipedia, which is no longer new, cannot be so cool. That is why Beethoven’s 5th does not top the pop charts, though if it were new it might.

And here’s The New Republic’s Jeet Heer, writing for the site’s blog, under the matter-of-fact header “Wikipedia is dying”:

And its treatment of volunteer editors is the culprit. The genius of Wikipedia has been its use of such editors, who do the grunt work that allows the site to maintain a consistent quality. Yet these very volunteers might be the undoing of the site. A new academic paper, flagged by economist Tyler Cowen, reveals that the number of volunteers peaked in March 2007 and has been in steady decline ever since.

So what’s wrong here? Simple: this study isn’t new! As one can verify quite easily on Aaron Halfaker’s website (and Stu Geiger’s as well) this study was published in 2013 (and circulated in late 2012).

Moreover, the question of what’s happening to Wikipedia’s community, particularly the overall number of active editors, has since then become less clear and perhaps more interesting. As first identified by editor WereSpielChequers, later examined by the community’s Wikipedia Signpost, and also by Halfaker himself in a post at the Wikimedia Foundation’s blog, Wikipedia’s “decline” is less obvious than it once was:

The English Wikipedia’s population of very active editors—registered contributors with more than 100 edits per month—appears to have stabilized after a period of decline. We’re seeing some of the same trends globally on other language Wikipedias. … Broadly speaking, it appears the number of very active editors has recovered from a mid-2013 drop and, for the moment, is continuing upward aseasonally.

You would not know this by reading Cowen and Heer! The former post has 76 comments, none of which (that I can tell) point out this survey isn’t new. And while TNR does not allow comments on posts, Heer’s tweet announcing the blog post has been liked, retweeted or replied to about fifty times, without anyone pointing this out, either.

To be sure, Wikipedia still has many problems that cannot alone be addressed by a modest uptick in active participation. That still doesn’t make it OK to pass off outdated scholarship as a new development, and without a considered appreciation of the topic—Wikipedia in 2015 had fewer editors than 2007, a new paper “reveals”!

It’s not hard to see how a dominant storyline about an interesting but little-understood phenomenon (like Wikipedia) can become an entrenched meme, easily passed along from writer to reader, reinforced by feedback and becoming resistant to new information. And we need public intellectuals to help correct this kind of misinformation. Cowen and Heer should update their blog posts, and I’ll update this one if and when they do.