Excitement about GDELT and some Personal Intellectual History

GDELT (Global Data on Events, Location and Tone), a new, freely available, global events data base covering the years 1979 til recent is tearing up the interwebs among social science types (e.g., see Foreign Policy here and A Dart Throwing Chimp here).  You can download the data here.  GDELT is exciting in part because it covers all countries in the world and in part because it is very similar to the ICEWS data [gated], which form the input to a “political instability” (in quotes because I despise the term “instability,” but that’s a topic for a different post) forecasting system that to which several thousand government officials have access (info gleaned at ISA last week).

Back in 2008 Steve Shellman and I wrote a piece in Perspectives on Terrorism titled “Conceptualizing a New School of Political Conflict and Terror:  From Attributes to Behavior & Why Policymakers Will Care” (PDF; html here).  I just checked my CV so I could locate the piece, and learned that I never put it on there.  It isn’t the kind of thing that helps one get a raise (i.e., not published in a refereed journal), so I guess I just never got to it.  In any case, we open with this claim:

Some interesting changes in the interchange among US government officials and social science researchers, political scientists in particular, in the nation’s universities are afoot.  One might date the beginning of this new relationship to then Vice President Al Gore’s initiation of the State Failure Project in 1994.

One of the key people who launched that project was my dissertation advisor, Ted Gurr.  Shellman and I then continue and make the following kinds of arguments, explaining why the trend was emerging.

First, this essay takes issue with the dominant structural and systems-oriented theoretical approaches to studying terrorism and political conflict and instead advocates an actors-based process approach to studying such phenomena… Second, this essay calls for more quantitative studies of terrorism to treat terror as a tactical choice within a framework of other violent and cooperative actions. Future work should model the interaction between state and non-state actors’ tactical choices (and vice-versa).

We continue…

This shift is critically important because it means that theory becomes much more useful to policy makers: the emphasis on parties to the conflict leads this research to develop hypotheses about the conflictual behavior of dissidents in response to government behavior and vice versa. By moving away from thinking about the impact of democratic v. autocratic institutions, the size of GNP/capita, and the ethnic composition of society these scholars have begun to ask the following sorts of questions:
* When does repression work? When does it backfire?
* Why are some dissident groups so much more violent than others?
* What explains varying levels of discrimination in targeting across time and space?
* What event sequences lead to conflict escalation?
* What are the effects of government countermeasures on the tactical choices of dissidents?
* What explains the ebb and flow of government-dissident behavioral exchanges?
* What event sequences lead to negotiated peaceful settlements?
* Why can factions spoil a peace process?
Note that information about political institutions, economic output, and ethnic composition are of limited usefulness for answering these questions. Why? Because those characteristics of the country in which these conflicts unfold do not change much over time.

What the heck does this have to do with GDELT?  The enthusiasm noted in the posts linked to above comes from folks who are at the intersection of scholarship and policy, as Shellman and I suggested would continue to grow, and GDELT is the type of data one needs to implement the shift Shellman and I call for (and I have been calling for since 1995; see below).

Shellman, of course, is a PhD student of mine, and he is one of the key personnel that made the ICEWS project successful.  In addition to being a Gurr student I am also one of Mike Ward’s students, and Mike is another one of the key personnel who made ICEWS successful.  The third is Phil Schrodt, whose work with events data spurred automated coding of conflict events (and Phil is a co-author on the GDELT project).

KEDS was the first automated events data project to study conflict.

When I decided to write my dissertation on the termination of the war for national liberation in Zimbabwe I decided to adapt the events data modeling approach that Ward, Schrodt and others had been using to study interstate conflict, such as arms races, and other foreign policy behavior.  I could not understand why conflict researchers were so interested in the structural characteristics of polities, economies, and societies—characteristics that no government nor dissident can change on their own—and disinterested in the behavior of governments and dissidents toward one another (which those actors quite obviously select).  Since I wanted to know what led to a conflict’s end, it seemed to me that I ought to study the interactive behavior of the combatants.  And events data made that possible.

Sean P. O’Brien knew of my dissertation work (reported in a 1995 article in JCR [gated]) and pursued similar work on terror attacks in his dissertation (like me, publishing the main findings as an article in JCR in 1996 [gated]).  Later O’Biren was hired by DARPA and pitched his boss on launching a project he called ICEWS, which was modeled loosely on his dissertation project (as summarized in that JCR article).  ICEWS was the first successful DARPA social science project to be rolled out into regular use by the government.  So, while there are self-hating political scientists who do not think we can forecast, and politicians who want to shut down NSF funding for political science research, there is none the less plenty of reason to be bullish about both conflict research and our ability to impact policy.

But that’s not all.  Two undergraduates at Florida State who took a course of mine before going off to earn PhDs (and on whose dissertation committees I sat as an outside member), Scott Edwards and John Sulik, launched (Edwards with the idea and position at Amnesty International USA to make it possible; Sulick as the grunt doing the initial satellite imagery work) the Eyes on Darfur project, which uses remote imaging to document human rights abuses and has since become a standard tool used not only by AI (see Eyes on Syria, Eyes on Nigeria, etc.), but Human Rights Watch and others (Sulik & Edwards 2010 [gated]).  USAID recently launched an effort to crowd source ideas for how decentralized information might be aggregated to stop mass violence against civilians.  Finally, and this is even more exciting, check out the Human Rights Data Analysis Group’s post on their blog today.

I have not yet dug into GDELT, but I guess it will not surprise you to learn that, like many others, I am pretty damn excited about its public release.