feature | information and intelligence
Joelle Burbank and Mark A. Loucas
Identifying
Precursors of
Conflict
Using Open-Source
Intelligence to Determine
Who’s Hot and Who’s Not
19
Top scorers: Somalia. Photo: Milton Grant/U.N.
F
OR the last decade, the social three countries do not relate to the and limitations to consider when using
journal of international peace operations | volume 5 | number 3 | november-december 2009
science community has moved internal dynamics of any single country. content analysis. First, content analysis
ever closer to applying scientific Similarly, articles that mention “weapons does not measure “ground truth” directly.
models to the world’s at-risk nations. The of mass destruction” are generally about Instead, it measures how ground truth is
Fund for Peace (FfP)’s Conflict Assess- relations between countries rather than reflected in the reports coming from a
ment System Tool (CAST) approach is within one. Filter words and phrases can variety of sources, including news media,
one such attempt. then be customized to fit the goals of government reports, think-tanks,
each research project. international bodies, activist materials,
CAST is a computerized content analysis interviews, essays, etc. Second, despite the
tool that uses Boolean logic to identify After the selected collection of articles has accuracy that can be achieved by fine-
and quantify open-source digital data been filtered, it is scanned using search- tuning phrases and adding filter words, a
covering a variety of issues relevant to phrases. Rather than attempting to count computer does not read the same way as a
early warning and assessment of internal events, CAST quantifies information human does. It lacks the ability to identify
conflicts. Starting with a conceptual using the “saliency”, or prominence, of nuances and context that we take for
framework of internal conflict risk that certain issues in the information land- granted. However, all methods of data
includes twelve social, economic and scape. The saliency is calculated by taking collection are subject to limitations, and
political indicators (each of which is made the number of “hits” for a particular when these are recognized and controlled
up of several sub-indicators) CAST is able measure as a percentage of the total as much as possible, content analysis is a
to categorize vast amounts of informa- number of articles in the sample for the powerful research tool. For instance,
tion, then quantify and track changes in pertinent time period. This is more when a large sample of articles is used, it
each variable over time. The technology accurate than simply counting the number helps to balance out the effects of a few
can be adapted to track virtually any issue of hits because it allows one to control for articles that are read incorrectly by the
that is reported in the media or other fluctuations in total media coverage software. It also can reduce human bias,
electronically available content. between countries or time periods. For fill gaps not accounted for by other data
example, if 2 percent of the articles about collection methods and be very timely in
The process begins with downloading Nigeria talk about corruption, and 0.2 following trends.
articles on a particular country or other percent of Sweden-related articles talk
geographic area of interest. This group of about corruption in the