Could language analysis tools detect lone wolf terrorists before they act?

Nidal Hasan, the US army psychiatrist turned lone wolf terrorist

By Alex Fradera

By the time a terrorist attack has begun, the security services have already failed. But the challenge they face in detecting potential attacks is substantial, especially since the tactic of terrorism has increasingly been taken up by individual attackers inspired by, but not directly beholden to, formal movements. Spotting a lone wolf among the flock is no easy task, especially when it relies on a bottleneck of human analysis. A new paper in the journal Aggression and Violent Behavior uses a test case of a real lone wolf attack to explore ways we may be able to deal with this in the future. Using online language analysis tools, it hunts within blocks of text for the warning signs we might otherwise miss, with the hope of helping us to more effectively detect the predator.

Giti Zahedzadeh of the Centre for Neuroeconomic Studies at Claremont Graduate University in California wanted to see what insights could be gleaned about violent tendencies by analysing text using the open-source web application Voyant. The application has a number of features such as creating word clouds, word frequency lists, and word association chains to pull out patterns from blocks of text.

The raw data Zahedzadeh used were writings by Nidal Hasan, the US military psychiatrist who turned traitor and went on an ideologically motivated mass shooting in 2009 in Fort Hood, Texas, killing 13 and injuring many more. Hasan, who is now incarcerated awaiting execution, received some attention from authorities before his crime but he was never classified as “being involved in terrorist activities,” leaving him a free hand to commit his attack. Were warning signs missed?

The Voyant programme was first applied on a powerpoint presentation Hasan gave two years before the attack, as part of the final stage of his medical training. He chose the topic “The Qur’anic World View As It Relates to Muslims in the U.S. Military.” Its purpose was ostensibly to explore the topic to inform policy and treatment, but with the benefit of hindsight, one can see an individual working out in public his own misgivings and sounding out justifications for a course of action. Using Voyant, Zahedzadeh was able to visually zoom through key words and explore their associations, such as “fight”, “cause”, and “important”, and generate a word cloud highlighting the prominence of the words “god” and “submission” in one of the sections.

Perhaps more interesting is the email exchange between Hasan and a radical cleric, Anwar al-Awlaki (who was later killed by a US drone strike in Yemen in 2011). Al-Awlaki presided over the funeral of Hasan’s mother some years prior to Hasan’s atrocity, but had since moved to Yemen to continue his activities with less interference. The exchange shows Hasan testing out ideas and probing whether certain violent actions could be consistent with Islamic law. These messages were picked up at the time by the Joint Terrorism Task Force but ultimately not acted on. However, the Voyant analysis pulled out a number of red flags, such as a high coincidence of the term “permissible” in the context of the word “suicide”, and the pairing of “intention”, “kill,” and “soldiers” – the latter precisely reflecting the act Hasan was to go on to conduct.

Again, reading the text with the benefit of hindsight, the warning signs feel like they jump out at you without the need to collate word patterns, but of course, it’s easy to spot signs of something once you know about it. The purpose of using a technology approach is to foreground patterns that may seem otherwise innocuous, objectively presenting frequencies and interconnections in black and white. Moreover, gathering this kind of data into a larger corpus, based upon many cases, could help us understand “the process in which an individual moves from radical opinion to violent action,” Zahedzadeh said.

Overt attacks and covert thoughts

Alex Fradera (@alexfradera) is Staff Writer at BPS Research Digest

2 thoughts on “Could language analysis tools detect lone wolf terrorists before they act?”

  1. It is also fascinating to see the role of psychological under-reaction to threats play out in these scenarios. I work for a large company and declarations of aggression and violence against non-believers are common place from some radicalized colleagues. One colleague who went on to be convicted of stabbing someone, vocally expressed violent and openly pro-islamizing views to anyone who would listen. He had a typical UK attacker profile, petty criminal background, reformed convert after troubled family life, history of drug dealing to non-adherents and mental health issues. There were several others who openly celebrated IS attacks, one even on social media(the company monitors social media). However the company was petrified of accusations of racism, in the same way that was observed in the Rotherham case, so they preferred to ignore it. The majority, knew this and even expressed horror at this toleration of clearly unacceptable views within a corporate setting, but still feared the possible backlash, if they took it further. This shows that lone-wolves don’t just spring from no-where, they are visible for a long period, but it is our interpretation of those signs and what we do with that information, that determines how effective we are at spotting these individuals.

    Even in cases where managers have first hand knowledge of extremist leanings and declarations of cognitive alignment with terrorist ideologies, political correctness has a huge part to play, especially within large organisations, where corporate fear of weaponised labelling is considerable. It would be interesting to study in the case mentioned in the article, how many of Hasan’s colleagues had privy to key information, but would never of dreamed of relaying that material, for fear of damage to their career or personal safety.

    There is also a geographic element to this issue. Signs tend to get ignored to a greater extent within sympathetic communities and consequently, it is harder to find and indeed voice these lone wolf signs, even if you find them. In many UK towns/cities violence can result from any disagreement with the religious sub-group narrative. In/out group preference(Tajfel et al 79) filters these signs. Hence spotting lone wolf signs, must take account of people’s readiness to speak and the consequences for them, of that reporting behaviour. This is even more relevant today with the Las Vegas shooting.

Comments are closed.