Betsy Devine has a master's degree in engineering from Princeton and many years of immersion in geek sociology, including both Slashdot and Wikipedia flame wars. Her blog and her Wikipedia watchlist reflect wide interests in science, politics, and reputation theory. Dead-tree publications include Longing for the Harmonies (a popular-science introduction published by WW Norton) and Absolute Zero Gravity (a science-joke collection published by Simon and Schuster.)
Slight expansion of the Abstract for that talk, "Schrödinger's Wiki: The Quantum Challenge of Media Attention":
In quantum mechanics, you can't observe a phenomenon without affecting it. When Wikipedia makes headline news, inbound waves of new visitors challenge the project. I will describe examples of two different challenges.
First, the "Swiftboating" edit war (November/December, 2005) began when political bloggers linked to this article, criticizing its point of view from both left and right. Within minutes, many new visitors, identified only by their IP addresses, started trying to edit the article, some expressing frustration with acts of vandalism.
Second, Wikipedia's increasing use as a media source motivated anonymous edits by Congressional staffers, as investigated by WikiNews in January/February, 2006.
The "Swiftboating" war typifies what I would call a "vandal wave," set off by negative coverage of Wikipedia, a pointer to a specific article, and new-user frustration with editing tools in conditions of heavy use. The Congressional edits could be described as a "spin wave," where highly-motivated professional writers attempted to shift the spin of important articles. In both cases, Wikipedia gets hit by what looks like a wave of undesirable contributors. I'll discuss some tools and metrics for each case, including ways to detect/recruit productive contributors. I hope the audience will also contribute new insights.
The main novelty of my presentation in 2006, novel to me anyway, was the discovery that one could detect an incoming "edit wave" to a page in the second-derivative-wrt-time of edit attempts. That is, some pages get a lot of edits normally and some get only a few, so edits-per-minute (i.e. the first derivative wrt time of edit attempts) measures something but it does not reliably signal attacks on an article. But it would be easy and simple to allocate some rough d-edit/dt to each page, and then set code in the background to flag any sudden departure as something that might be worth closer looks.
///////////////
My academic background is science/engineering, but I've spent much of the past four years working and/or blogging in the area people now call Web 2.0 -- one of my favorite parts of which is Wikipedia! New toy I just found -- you can see my edit count here). It's about 4,000 right now but will probably go up a bit by the time anybody but me reads this. Another techy toy ... images I uploaded to Commons.
