Proceedings:BD1/Notes

From Wikimania
Jump to navigation Jump to search
Transcript from Sunday's presentation by Betsy Devine. Note: These are not exact words - I've tried to keep them verbatim as possible, but I'm trying to get the things that capture the gist of Betsy's speech.

The title of my talk is Schwdinger's wiki. My husband who's a physicist said, "you're calling this WHAT?"

1. Media attention

Betsy spoke briefly about the "quantum impact" of the attention of media. You change the content just by looking at it, and media attention has this kind of effect on any interactive community.

The first type of challenge mentioned was the Vandal Wave: the example given was the "swiftboating" edit war from 2005/11/29. Vandal Waves come from media attention when many people are trying to edit (in conflicting and something factually inaccurate ways) the same article.

How did the conflict start? There are 95 pages in the "Wikipedia controversial topics," and any of them could have generated a vandal wave given the right kind of media attention. Swiftboating is a term that has a vague connection to a type of boat that was used in Vietnam. There was a debate about which definiton (of many wildly varying ones) was correct. The page was created in 2005 August, and between then and Nov 22 2005, there was almost no edits whatsoever. But at 15:23 on Nov. 29, something happened. In the ensuing 10 minutes, the article got more edits than it'd gotten during the first 3 months of its history.

[Displays graph of edits to Swiftboating article]

[Shows examples of Swiftboating edits] which are things like:

  • changing "accurate" to "inaccurate" (completely flipping the meaning of the article)
  • Addition, removal of POV tags
  • deleting entire paragraphs and replacing it with the single word "BUNK"
  • putting juvenile language in the paragraph

Audience question: What is vandalism? Betsy: [Note: I missed the answer, can someone fill in?]

Back to Swiftboating: What happened?

"It began with examples that linked to published sources, as per wikipedia guidelines. WThe article was blogged by Joshua Micah Marshall in "Talking Points Memo" and it got posted on swiftvets.com, where someone encouraged people to go and edit the article."

"What I'm calling a vandal wave is quite different than multiple edits by a single vandal. We have a lot of experience with single vandals [and how to stop them]. The challenge with multiple vandals is that it is set off by a media event. You can't just block one IP, more come in its place. It degrades your experience of wikipedia if you are a new user. You come in and you can't have the pleasure of being able to add a tangible edit of your own."

"In contrast, the "Swiftboated" article page got no media attention, no vandal wave: the conclusion is that this is set off by media."

[Shows graphs of similarly vandlized stories and their posts over time, graphing IP editsside by side by named edits. It's clear that you see a huge spike in IP edits when the vandalism starts.]

[Explains the Adam Curry podcasting scandal, and how the Podcasting and Adam Curry user categories spiked at the time of the vandal wave]

Wikipedia has bots. Betsy talked about her favorite, pgkbot. Bots try to figure out if this is a user that's been banned, vandalizing, etc. to alert people of possible problems. In this instance, somene created a username called "JEWS DID WTC! lol jews. lol internets". pgkbot blocked the IP, sent a "try again" message attempting to dissuade them.

Use code to detect the problem, use code to deflect the problem.

Audience comment: You do realize that people that make usernames like that don't expect to keep that name?

Betsy: Not one person who made their first edit to Wikipedia on Swiftboat during the vandal wave went on to become a regular contributor.

How can we detect vandal wave? Two obvious metrics:

  • time between edits by different users
  • surge in ratio of ips to registered users

Showed a graph with a spike on 10/6 (what happened here?): this is not a vandal wave, because you can see that all of the edits were made by users, this was a response to a major event [didn't catch what].

The second challenge discussed was the Spin Wave. The example given was the congressional edits in early 2006.

"People will pay professional writers to blog or write about their product in a positive light. You can get paid to it - in fact, there is a site called paidtocomment.com that gets people to do just this."

Conclusion:

Jimmy Wales: "It's an encyclopedia, not an experiment in democracy."

BUT

  • It's not just about facts, it's about code (that makes the editing and detection of those facts possible).
  • It's not just about words, it's about numbers (time between edits, ratio of ip edits to named edits, other metrics that you can use).