The PRIME dataset includes all suitable lone-actor cases for the period 1990 to 2016, in the United States and Western Europe. It stands at 125 individuals, which begs the question of ecological fallacy: should we make predictions about an individual's risk level based on statistical treatment of sparse data gathered from such a heterogeneous population? However, while the data we have aren't good enough for actuarial prediction, few would claim we should dispense with risk assessment.
Our findings suggest that lone actors are not that lone. We found that both physical and online links with other extremists are critical to the adoption and stability of the motivation to engage in terrorism and the acquisition and maintenance of the capability to act.
To deepen our understanding, we studied the sequence in which indicators occurred. We focused on mental health-related proxies often associated with 'individual vulnerability'. We found that the same indicators seemed to signal different processes at different stages. At times, they appeared associated with susceptibility to influence; at others, with exposure to radicalising environments; at others still, with the event's outcome.
Psychological distress sometimes seemed related to increased risk, while at other times it seemed to play a protective role, inasmuch as risk appeared to rise after treatment, suggesting that needs-based interventions, to the extent that they alter future interactions between person and situations, could have unintended consequences. For example, alleviating someone's depression (a good thing) might do much to restore their desire to go out, thereby exposing them to whole new (potentially risky) social environments as a result.
Comparison between actors on the basis of ideology didn't reveal meaningful differences. However, when we stopped looking at types of actors but modelled the interaction between personal and situational indicators, we identified four kinds of person-exposure patterns (PEPs): solitary, susceptible, situational and selected. What the idea of PEPs suggests is that risk doesn't reside in any combination of individual vulnerabilities or in any criminogenic environment, but that it emerges from the interaction between specific configurations of individual propensities and particular contextual features.
Brought together, our findings imply that vulnerability, motivation, capability, even risk itself, which are often treated as personal attributes, are better conceived of as contingent states, which emerge when people with certain characteristics, themselves dynamic, encounter situations with particular features that are also subject to change. This means that vulnerability, motivation, capability, and overall risk are inherently transient and context-dependent.
With that in mind, to not merely assess, but to analyse the conditions under which unwanted behaviour might emerge, we need inference frameworks that can:
- be applied across the ideological spectrum
- guide the identification and interpretation of fundamentally multifinal and equifinal indicators
- clearly articulate how personal factors intersect with contextual factors to generate (or suppress) the problematic behaviour.
The S5 framework
S5 was designed to operationalise the theoretical foundation behind PRIME and antecedent projects (see here, and here). In a field where violent extremism has been described as the product of a kaleidoscope of factors, creating infinite individual combinations, S5 sets out how five categories of effective factors interact, through processes of exposure and emergence, to generate (or suppress) the risk of radicalisation and terrorist action.
S5 shifts the analyst's attention from 'who and why' to 'who and where'.
At the individual level, what matters is susceptibility to moral change, which plays a key role in the acquisition (or not) of a terrorist propensity. Linking the individual to the context in which propensity change takes place is social and self-selection factors – background characteristics and personal preferences that affect the likelihood of finding oneself in any given place at any given time. These matter because radicalisation and terrorist behaviour take place in settings with specific characteristics that enable terrorism-supportive moral development (which matters for radicalisation) or the situational emergence and maintenance of the motivation to act (which matters for terrorism).
Note that vulnerability arises from the combination of susceptibility and selection factors, not susceptibility alone. These risky settings concentrate in environments with a particular social ecology. The processes that matter for the emergence of these settings, such as a terrorism-supportive moral context, are, in turn, influenced by certain factors at the system level.
Four of these five levels of explanations are concerned with context, not the individual. In other words, S5 shifts the analyst's attention from who and why to who and where.
This framework makes sense of some common observations. For example, if selection is the mechanism linking individual and environment, then a lack of stable profiles is to be expected: as risky settings displace as a result of systemic changes (such as counter-terrorism policy), new kinds of (susceptible) people get exposed.
Likewise, if we think of motivation as a self-sustaining personal attribute, we might wonder at the relatively low number of lone-actor terrorist plots. But if the motivation to act emerges from – and, crucially, is maintained by – the interaction of person and context, then any change in situation has the potential to halt the move to action. It may be much harder to keep up a stable apprehension of a situation (the perceived provocations, the belief in sufficient capability, the expected rewards) without co-conspirators, whose shared perceptions would otherwise contribute to situational stability.
Furthermore, while most risk assessment tools and models include a dozen or more ‘factors’, the markers that appear the most reliably associated with involvement in violent extremism are gender, prior criminal history and connections to radical peers. This would support the view that what really matters for the assessment of terrorism risk is evidence of a pre-existing criminal propensity (or a high susceptibility to acquiring one) and of vectors of selection for exposure to terrorism-enabling settings. If propensity, not ideology, is determinant, then it is no surprise that some actors have been known to 'change the t-shirt', simply moving from one kind of extremism to another as they got exposed to new radicalising settings.
Finally, given the prominent role of moral context in the emergence of risky settings, the very low volume of terrorist acts in our societies makes perfect sense. We are resilient to radicalisation and terrorism risk because the vast – vast – majority of people simply do not believe that terrorism is morally legitimate, and the stability of our systemic norms and processes of governance keep it so. Which means the reason why so few so-called susceptible individuals ever radicalise or carry out an act of terrorism, is that a lot more planets than just ‘individual susceptibility’ must align for the process to occur.
As my colleague Paul Gill puts it, "the next big challenges are essentially conceptual." With the (welcome) multiplication of datasets, risk indicators will proliferate. But if the same indicator can be a marker of risk in one context and a marker of protection in another; if potentially infinite combinations of markers can characterise the same process; if most of the so-called risk factors we rely on aren't reliable predictors, then what we really need are risk indications – not so much inventories of what to look for, but how and where to find it.
For that, we need explicit inference frameworks – analytical models built on the best evidence and the most robust theory available – to structure our risk assessments. These frameworks should be formalised and validated, as Paul Taylor suggests, through the systematic evaluation of the decisions they inform.
Yet, we have to recognise that producing more robust explanations will not necessarily improve the prediction of individual behaviour. Prediction requires knowledge of future conditions (the situations an individual may someday encounter), which is beyond our reach. This, I would say, is another reason to shift more of our focus from assessing ‘risky people’ to assessing ‘risky contexts’.
- Noémie Bouhana et al., 2016. “Risk Analysis Framework. FP7 PRIME Project,” http://www.fp7-prime.eu/deliverables/PRIME_D3.1_Risk_Analysis_Framework_Public.pdf
- Noémie Bouhana, E. Corner, P. Gill, and B. Schuurman. 2018. “Background and Preparatory Behaviours of Right-Wing Extremist Lone Actors: A Comparative Study,” Perspectives on Terrorism, vol. 12, no. 6.
- Emily Corner, N. Bouhana, and P. Gill. 2018. “The Multifinality of Vulnerability Indicators in Lone-Actor Terrorism,” Psychology, Crime & Law, vol. 2744, pp. 1–47.
- Bart Schuurman, L. Lindekilde, S. Malthaner, F. O’Connor, P. Gill, and N. Bouhana. 2017. “End of the Lone Wolf: The Typology that Should Not Have Been,” Studies in Conflict & Terrorism, vol. 0731, no. December.
- Bart Schuurman, E. Bakker, P. Gill, and N. Bouhana. 2018. “Lone Actor Terrorist Attack Planning and Preparation: A Data-Driven Analysis,” Journal of Forensic Sciences, vol. 63, no. 4, pp. 1191–1200.
As part of CREST’s commitment to open access research, this text is available under a Creative Commons BY-NC-SA 4.0 licence. Please refer to our Copyright page for full details.