Quite a few people have been working to make sense of the tens of thousands of classified reports published on Jul. 25 by Wikileaks.  --  On Monday, Noah Schachtman, who writes the Danger Room blog on the website of Wired, commented on some of the best attempts to illustrate what the data reveal.[1]  --  Schachtman highlights the work of Drew Conway, an NYU grad student (political science) who used R (an open source statistical programming language) and a graphical plotting software tool to show the spread of combat from 2004 to 2009.  --  It looks "like stop-motion photography of a freeway wreck," Schachtman said.  --  See the original for many additional links....


Danger Room

By Noah Schachtman

August 9, 2010


[IMAGE Six maps of Afghanistan, 2004-2009]

It’s one thing to read about individual Taliban attacks in WikiLeaks’ trove of war logs.  It’s something quite different to see the bombings and the shootings mount, and watch the insurgency metastasize.

NYU political science grad student (and occasional Danger Room contributor) Drew Conway has done just that, using an open source statistical programming language called R and a graphical plotting software tool.  The results are unnerving, like stop-motion photography of a freeway wreck.  Above is the latest example:  a graph showing the spread of combat from 2004 to 2009.  It’s exactly what you wouldn’t want to see as a war drags on.

“The sheer volume of observations [in the WikiLeaks database] inhibit the majority of consumers from being able to gain knowledge from it.  By providing graphical summaries of the data people can draw inferences quickly, which would have been very difficult to do by serially reading through the files,” Conway e-mails Danger Room.  “For instance, in the most recent graph I posted [see above], many people were noticing the increasing number of attacks around Afghanistan’s ‘ring road,’ over time, and seeing that as an indication of the Taliban’s attempt to undermine the Afghanistan government by cutting off villages from one another.”

Conway’s work largely mirrors what the U.S. military’s internal teams of intelligence analysts found.  But Conway and Columbia University post-doc Mike Dewar did all this work themselves, relying solely on free tools and the WikiLeaked logs.  Applying statistical analysis, they found little evidence of tampering in the reports.  Next month, Conway hopes, a group of New York-based R users will be able to tease out more insights from the data.

Obviously, the logs don’t tell the whole story of the war, as Danger Room has noted before.  And the stats may be unduly influenced by the spread of NATO forces into different parts of the country.  But for now, the most striking point to Conway was how bad things turned in 2006 and 2007.  In Afghanistan’s south, for instance, there was only minimal fighting in the start of ‘06.  By the end of the next year…  well, see for yourself.


UPDATE: Conway’s work is just one of dozens of international efforts to turn the WikiLeaks war logs into something graphic.  Visualising Data has links to some of the best, including this one: