Virtual Worlds Require Virtual-HUMINT (VHUMINT)

Via CT Blog -

The excellent Dark Web project at Arizona Universities Artificial Intelligence Lab has recently completed research into the use of Web 2.0 media by International jihadi groups. While fascinating in some respects it also clearly demonstrates how traditional text-mining attempts to collect data can be applied to some Web 2.0 applications, but miss the mark with virtual worlds.

The leader of the lab Dr. Chen kindly forwarded their research paper to me and it can be linked to here (Cyber Extremism in Web 2.0: An Exploratory Study of International Jihadist Groups or here). In essence the Dark Web project’s methodology is to search for material with extremist or terrorist style language (but please read the paper for a better description of methodology). Interestingly, they concluded that sites such as Facebook and MySpace, which have been big components of the Web 2.0 milieu are not suited to the propagation of extremist views,

“We did not consider social networking sites, such as MySpace and Facebook, although they are a major component of Web 2.0. Based on our preliminary exploration, we found that the prevalent amount of personal data on these sites, the tight social linkages, and the potential issue of “guilt by association” (for site owners and “friends”) may have discouraged extremists from using such a medium.“

With regard to virtual worlds the Dark Web project found nuanced evidence of extremist ‘activity’ within Second Life, by using Second Life’s internal search system to look for text containing extremist language. More than anything this highlights the difficultly in researching extremist movements within virtual worlds as the worlds themselves do not easily allow themselves to be searched or data-mined in this fashion. The issue of search within virtual worlds has been grappled with by a number of commentators, the problem being what to search for, people, content, land, buildings, events, etc. But more importantly language used within virtual worlds doesn’t hold the same meaning when pulled out of its in-world context. For example the dark-web project cites a number of groups it discovered in Second Life using extremist language, one of who (Terrorists of SL) has a small virtual headquarters called ‘Taliban Towers’. An examination of this site and associated group would tend to suggest they are a role-playing collective with little real-world application. The same goes for other in-world organizations such as, Elite Jihadi Terrorist group.

Therefore, what this research does is point to something fundamental about how global intelligence and law-enforcement agencies need to approach the examination of virtual worlds, and that is that raw data-crunching is likely to prove unsatisfactory. Ironically, virtual worlds require a uniquely human approach. The only sure way to gather information on extremist or criminal groups operating in virtual worlds is to enter the environment and interact with the suspected groups. The United States Intelligence community is not short of computing power but what this new environment needs is the human touch or to put it in the language of the Beltway -- layer Virtual-HUMINT over the SIGINT mission.