Do autism spectrum disorders (ASD) increase the risk of terrorism engagement? A literature review of the research evidence, theory and interpretation, and a discussion reframing the research-practice debate

ABSTRACT

A number of publications have recently suggested or claimed that autism spectrum disorders (ASD) do or may increase an individual’s risk of or vulnerability for terrorism engagement. In this paper, we aim to ascertain the extent and nature of this purported relationship between ASD and terrorism engagement as reported in peer-reviewed literature. We analyse the relevant literature by considering research designs and the importance of comparison groups in analytic studies for studying why outcomes occur. This review finds that the evidential and theoretical basis in research for the identified suggestions and claims is lacking. Existing research cannot definitively conclude, nor does it suggest, that individuals with ASD are any more vulnerable to, or any more at risk of, terrorism engagement than other individuals. The findings of this literature review pose questions that arise across the research-practice debate. We discuss and attempt to broaden the research-practice debate in relation to the ongoing ASD-terrorism debate by drawing upon critique from the field of science studies.

KEYWORDS:

• Autism spectrum disorders (ASD)
• terrorism risk assessment
• research designs (interpretation)
• research-practice debate
• science studies

Introduction

Autism spectrum disorders and terrorism engagement

Autism spectrum disorders (ASD) include a ‘wide spectrum of development disorders’ (Muhle, Trentacoste, & Rapin, 2004). ASD are lifelong neurodevelopmental disorders, as distinct from psychopathologies or mental illnesses. However, Im (2016) reports that comorbidities are common in ASD and can include ‘schizophrenia (25%), depression (12.5%), and attention-deficit disorder (18.75%)’ (p. 25) (see also Chown, Beardon, & Cossburn, 2018, p. 59; Inderberg et al., 2019, p. 1; Maenner et al., 2021, p. 1; Murphy, 2010, p. 462). The Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition) criteria for ASD require ‘persistent deficits in social communication and interaction along with restricted patterns of behavior, interests, or activities, beginning in the early developmental period and causing significant functional impairment; intellectual and language impairment may or may not be present’ (Im, 2016, p. 15).

Im (2016) notes that ‘instances of violence committed by individuals with known or suspected ASD’ have been discussed in scholarly literature and in the media, and that, ‘over the last two decades, the interest in this topic has increased, but research has yielded inconsistent or, at best, inconclusive evidence regarding any association between ASD and violence’ (p. 14). While Im’s (2016) review looks at ASD and general violence rather than terrorism, a number of recent publications suggest or claim that ASD does or may increase an individual’s risk of, or vulnerability to, terrorism engagement. These suggestions or claims have appeared in peer-reviewed journal articles, in non-peer reviewed reports and in the media.¹

For example, the relationship between ASD and terrorism is discussed in three Radicalization Awareness Network (RAN) reports.² Krasenberg and Wouterse (2019) report that ‘certain characteristics can make people with ASD more vulnerable to radicalization and violent extremism’ (p. 12). Krasenberg and Wouterse (2019) claim to ‘identify risk factors that make these people [individuals with ASD] vulnerable to [engaging in] violent extremism’ (p. 1). Krasenberg and Wouterse (2019) recommend that it is ‘important [for clinicians] to define the specific aspects of the … disorders that pose a risk and might lead to radicalization’ (p. 11). De Marinis and Boyd-MacMillan (2019) write that ASD and certain other diagnoses are over-represented in individuals involved in violent extremism or at risk of violence, whether as lone-actors or otherwise’, and that, ‘while only a very small percentage of individuals [are] being radicalized, abnormal functioning can increase risk and/or vulnerability to [terrorist or violent extremist] recruitment’ (p. 2). Al-Attar (2019) suggests that ‘a number of features associated with autism may shape [an individual’s] vulnerability and risk … [of] extremism’, and when an individual [with ASD] starts to act out their terrorism-related interests, risk may be heightened’ (p. 11). Al-Attar (2019) describes how ‘the same targets may be pursued repeatedly due to the obsessive repetitive nature of [restricted] interests, and hence, vulnerability/risk may also be repetitive and intense’ (p. 11).

Suggestions or claims linking ASD to an individual’s risk of or vulnerability to terrorism engagement have also recently appeared in a number of peer-reviewed journal articles. For example, Al-Attar (2020) writes that ‘features of ASD may shape experience, functioning and behavior that may in some instances exacerbate an individual’s vulnerability’ (p. 927). Al-Attar (2020) recommends that,

when terrorist acts are planned or executed by individuals with autism, it is important to develop an understanding of the individual’s autistic functioning and how it may contextualise factors that push them towards terrorism and aspects of terrorism that may pull them in, in order to manage and reduce risk. (p. 926)

Faccini and Allely (2017) write that,

although there is no conclusive evidence supporting the notion that individuals with ASD are more violent compared to individuals without ASD, there may exist specific generative and associational risk factors which may increase the risk of offending among individuals with ASD. (p. 77)

Faccini and Allely (2017) present a series of case studies that aim to ‘demonstrate the functional relationship between the ASD and terroristic acts’ (p. 79).

Considering these recent contributions to the debate on the potential association between ASD and an individual’s risk of or vulnerability for engaging in terrorism, it is pertinent to take stock of the current body of research-based evidence. As noted, Im (2016) found the evidence regarding any association between ASD and general violence to be ‘inconsistent or, at best, inconclusive’ (p. 14). In this review of peer-reviewed literature, we look specifically at the relationship between ASD and terrorism engagement. To what extent does contemporary research-based evidence, theory or critique support the aforementioned claims and suggestions? How might existing research help us to make sense of this potential association? And what practical, theoretical and ethical questions arise from the findings of this literature review for future research and clinical practice? Furthermore, we will explore how these questions, and the ongoing debate about the purported intersections between ASD and terrorism engagement arise across the broader debate between modern research and practice. The discussion will attempt to broaden and to reframe the research-practice debate in order to contextualize our findings on the ASD-terrorism question.

Definitions

In order to clarify the scope of this literature review, we must first define and distinguish the terms, ‘ASD’, ‘terrorism’, ‘risk’, ‘vulnerability’ and ‘needs’.

ASD and terrorism

There are no universal definitions of ‘ASD’ or ‘terrorism’. These terms are marked by heterogeneity and are framed by a variety of definitions, definitional issues, interpretations or diagnostic criteria that shift across geopolitical space and historical time. For this reason, meta-analysis is not a feasible method for this review. ASD can include both Kanner’s autistic disorder and Asperger’s disorder, which, as Im (2016) notes ‘were included as discrete entities … in DSM-4’, but which ‘were subsequently collapsed into the single designation of autism spectrum disorder (ASD) in DSM-5’ (p. 14). Im’s (2016) definition broadly captures ASD as it morphs over time and across the various DSM editions up to DSM-5, and which also includes other diagnostic criteria used in practice. For the purposes of this review, we follow Im’s (2016) definition of ASD. For inclusion in the review, we require studies to report a formal ASD diagnosis. Publications that include a suspected, distance or archival diagnosis by researchers or a contested formal diagnosis are excluded from this review.³

Similarly, ‘terrorism’ and ‘violent extremism’ are two overlapping terms with shifting definitions. The term ‘terrorism’ will be used herein but we adopt a broad enough definition that overlaps with other definitions of terrorism and violent extremism in use. For the purposes of this review, we adopt an FBI definition, as cited in Arnold et al. (2003), which defines terrorism as ‘the unlawful use of force or violence against persons or property to intimidate or coerce a government, the civilian population, or any segment thereof in furtherance of political or social objectives’ (p. 48). This review distinguishes between terrorism and non-violent political extremism, and also radicalization, a process of mobilization towards terrorism which can be conceptualized in various ways, but which doesn’t yet meet the threshold of being terrorism as defined above.

Note that in Australia and in some other countries, the unlawful use of violence in terrorism also includes the preparations and planning of terrorist violence. To qualify as terrorism, we require that the support of violence must be behavioral and more than ideological. Simply believing in terrorist action is, by itself, not enough. Terrorism engagement must include a behavioral and a political element that collectively work towards or become political violence.⁴ We understand radicalization as a process (which is also more than merely ideological and which may include behavioral, social and emotional elements) that may or may not lead to terrorism engagement.⁵ This review studies the intersection of ASD and terrorism, which may, but does not necessarily include radicalization. Therefore, research on the intersection between ASD and radicalization, but which doesn’t meet the threshold for terrorism engagement as defined above, is excluded from this review.

Risk and vulnerability (versus needs)

Although various definitions exist, risk is usually defined as a measure of statistical likelihood, based on probability. We adopt Monahan’s (2012) definition of risk. Monahan (2012) explains that,

the definition of risk assessment given by Kraemer et al. (1997) is the one I have found most useful: ‘The process of using risk factors to estimate the likelihood (i.e. probability) of an outcome occurring in a population'. Importantly, Kraemer et al. (1997) define a risk factor simply as any variable that (a) statistically correlates with the outcome (in this case, violence), and also (b) precedes the outcome in time. There is no implication in this definition that the risk factor in any sense causes the occurrence of the outcome. (p. 7)

Kraemer et al. (1997) discuss how ‘terms such as risk, risk factors, and especially the term cause are inconsistently and imprecisely used’, which they find to be problematic because ‘imprecise reports can impede the search for understanding the cause and course of any disease [or more generally, any phenomenon] and also may be a basis of inadequate clinical or policy decision-making’ (p. 337). Note that defining risk as a statistical likelihood, or in terms of probability, requires correlation between terrorism engagement and a risk factor, which in turn, requires a direct comparison between cohorts, or more specifically, findings from analytic studies that employ an appropriate control or comparison group.

The terms vulnerability and risk are often conflated (Knudsen, 2020). However, these are distinct concepts. While risk is a comparative and cohort measure of the likelihood of engagement, vulnerability describes an individual’s openness or susceptibility for engagement. It vitally important to distinguish the concept of needs (the requirements of an individual or a group) from the concepts of risk and vulnerability, as defined above. This review does not question whether or not individuals with ASD may have specific needs that may be distinct from the needs of individuals who do not have ASD. Meeting an individual’s needs is just as important in the context of terrorism engagement as it is in any other context. It is not in question that practitioners should carefully consider the potential interplay of ASD and terrorism engagement in rare cases where these co-occur in order to understand an individual’s needs. Rather, what is in question here is whether or not researchers or practitioners may consider the potential interplay of ASD and terrorism engagement in rare cases where these co-occur in order to understand an individual’s risk or vulnerability for terrorism engagement. This review asks whether research may shed any light on the question of whether or not the potential interplay of ASD and terrorism engagement has an effect, if any, on an individual’s risk or vulnerability for terrorism engagement. Therefore, the three concepts of needs (requirements), vulnerability (openness) and risk (likelihood) should be understood as being distinct.

A note on the ‘D’ in ASD and the concept of neurodiversity

Researchers arguing for a neurodiversity perspective on ASD have recently found that the social and behavioral deficits related to ASD may be misconstrued or misinterpreted by framing ASD in relation to a neurotypical majority (Milton, Gurbuz, & Lopez, 2022; Mitchell, Sheppard, & Cassidy, 2021). Moreover, the assumption that neurodiversity and behavioral differences are always problems, ‘disorders’ (as in the ‘D’ in ASD) or dangers to society ignores the fact that, under certain circumstances, these differences (for example, detailed thinking, affinity for analysing complex patterns, a comfort with rules and guidelines) can also function as cognitive strengths (Robertson, 2010). Baron-Cohen (2017) writes that ‘the notion of neurodiversity is highly compatible with the civil rights plea for minorities to be accepted with respect and dignity, and not be pathologized’ (p. 746). He conceptualizes ‘autism as an example of diversity in the set of all possible brains, none of which is “normal” and all of which are simply different’ (Baron-Cohen, 2017, p. 746).

Although this review continues to use the terms ‘ASD’ and ‘disorder’ in line with the included literature, we acknowledge the neurodiversity debate. Following Robertson (2010), we adopt a neurodiversity perspective of ASD by taking a symmetrical approach to framing our research question, that is, by recognizing that ASD may also be associated with strengths and positive outcomes, rather than only exploring or adopting a deficit or disorder framing. Furthermore, while the concept of the risk of terrorism engagement requires analytic studies to directly compare terrorism engagement outcomes for cohorts of ASD and non-ASD individuals, we don’t interpret these comparison groups to necessarily be a comparison between normative and non-normative groups of people. We also note that, in the case of the absence of supporting research evidence or theory, claims or suggestions that ASD and terrorism engagement are correlated may, perhaps unintentionally, stem from or lead to the stigmatization of people, more often young people, with ASD. This situation may also negatively affect the precision of terrorism risk assessments by potentially focusing on individuals who we do not know to be at risk of terrorism.

Method

An advanced title-abstract-keyword search using Boolean logic was conducted across multiple online search platforms to search a number of relevant scholarly databases. Table 1 reports the Boolean search syntax, databases searched, search limiters and the number of publications identified for each platform.

Table 1. Database search results.

Search platform	Access date	Search limiters	Includes databases or indexes searched	Search syntax	Number of publications identified
Academic Search Elite (EBSCOHost research databases)	22/11/21	Peer Reviewed; English	Academic Search Elite, EBSCO ebook collection, Medline, APA PsychArticles, APA PsychInfo. Includes SocIndex	(terroris* OR extremis* OR radicali* OR ‘political violence' OR ‘far right' OR jihard* OR CVE OR PVE OR XRW OR RWE OR ‘extreme right wing' ) AND (ASD OR autis* OR ‘autism spectrum' OR Asberger# OR PDD-NOS OR ‘PDD NOS')	107
Oxford Clinical Psychology	22/11/21	Peer Reviewed; English	All	(terroris* OR extremis* OR radicali* OR ‘political violence' OR ‘far right' OR jihard* OR CVE OR PVE OR XRW OR RWE OR ‘extreme right wing' ) AND (ASD OR autis* OR ‘autism spectrum' OR Asberger# OR PDD-NOS OR ‘PDD NOS')	3
ProQuest One Academic	24/11/21	Peer Reviewed; English; Academic Journals; Limiter: Last 5 years	All. Includes Criminal Justice Database, Sociological Abstracts, Worldwide Political Abstracts, PAIS index	(terroris* OR extremis* OR radicali* OR ‘political violence' OR ‘far right' OR jihard* OR CVE OR PVE OR XRW OR RWE OR ‘extreme right wing' ) AND (ASD OR autis* OR ‘autism spectrum' OR Asberger* OR PDD-NOS OR ‘PDD NOS')	927
Web of Science (All Databases)	24/11/21	English	All. Includes the Core collection plus Medline	(terroris* OR extremis* OR radicali* OR ‘political violence' OR ‘far right' OR jihard* OR CVE OR PVE OR XRW OR RWE OR ‘extreme right wing' ) AND (ASD OR autis* OR ‘autism spectrum' OR Asberger$ OR PDD-NOS OR ‘PDD NOS')	70
PubMed	24/11/21	English	All	(terroris* OR extremis* OR radicali* OR ‘political violence' OR ‘far right' OR jihard* OR CVE OR PVE OR XRW OR RWE OR ‘extreme right wing' ) AND (ASD OR autis* OR ‘autism spectrum' OR Asberger* OR PDD-NOS OR ‘PDD NOS')	24
Science Direct (Elsevier)	24/11/21	Peer Reviewed; English; Modified Search (advanced search only supports 8 Boolean terms and no wildcards).	All	(terrorism OR extremism OR extremist OR terrorist) AND (ASD OR autism OR autistic OR Asberger OR ‘PDD NOS')	698
Scopus	24/11/21	English	All	(terroris* OR extremis* OR radicali* OR ‘political violence' OR ‘far right' OR jihard* OR CVE OR PVE OR XRW OR RWE OR ‘extreme right wing' ) AND (ASD OR autis* OR ‘autism spectrum' OR Asberger* OR PDD-NOS OR ‘PDD NOS')	88
				Total	1917

1,917 records were identified by the combined database searches. After removing 569 duplicates, 1,348 records were screened for inclusion in the review by title and abstract, based on the following inclusion criteria. For inclusion in the review publications:

• must be peer reviewed

• can include journal articles, books or book chapters

• can include reviews and meta-analysis

• must meet the aforementioned definitions of ‘ASD’ and ‘terrorism’ as stated in the ‘definitions’ section in the introduction

• must relate to the review questions stated in the introduction. We include, and distinguish between, (1) literature that makes a suggestion or claim that ASD changes, or may change, an individual’s risk of, or vulnerability for, terrorist engagement; and (2) literature that includes original research (theory, critique or empirical findings) on the intersection of ASD and terrorism engagement, and/or interpretations of this research.

Following this process, 44 publications were downloaded into an Endnote library for screening by full text. The bibliographies of these 44 publications were searched and additional author searches were conducted in Google Scholar to identify any additional relevant publications that were not identified by the database search. Three additional publications were identified. In addition to this process for screening peer-reviewed academic literature, the three non-peer-reviewed RAN reports already identified in the introduction were also included in the literature review (no further searches were conducted for grey literature). In total, 50 publications were identified for screening by full text. Of these, 16 publications were included in the review for analysis, and 34 publications were excluded.

Results

Table 2 lists the included publications and categories their reasons for inclusion. Table 3 briefly describes the included publications.

Table 2. Included publications and reasons for inclusion.

Citation	Includes original theoretical or empirical research	Includes interpretation of research findings	Makes claim or suggestion linking ASD to risk or vulnerability	Peer-reviewed
Al-Attar (2018)			●	●
Al-Attar (2019)			●
Al-Attar (2020)	●		●	●
Allely and Faccini (2018)		●	●
Chown et al. (2018)		●		●
Corner and Gill (2021)		●		●
Corner et al. (2016)	●	●		●
De Marinis and Boyd-MacMillan (2019)		●	●
Faccini and Allely (2017)	●	●	●	●
Gill et al. (2020)		●		●
Knight et al. (2017)	●			●
Krasenberg and Wouterse (2019)		●	●
Little et al. (2021)	●		●	●
Misiak et al. (2019)		●		●
Walter et al. (2020)		●		●
Weenink (2015)	●	●		●

Table 3. Descriptions of included publications.

Citation	Title & description
Al-Attar (2018)	Interviewing terrorism suspects and offenders with an autism spectrum disorder Journal article that describes a hypothetical case study ‘to illustrate how the different facets of autism may collectively unfold to impact a terrorist offender’s pathway to terrorism’ (Al-Attar, 2018, p. 332).
Al-Attar (2019)	Extremism, radicalization & mental health: handbook for practitioners RAN report identified on p. 3–4 that makes claims or suggestions relevant to the review question. Described as a ‘handbook for practitioners’ (p. 1) that provides a ‘research- and theory-informed aid for clinical forensic practitioners working with individuals who present with risk/vulnerability to violent extremism and mental illness’ (p. 3). Pages 11–16 discuss potentially relevant features of ASD that may shape vulnerability and risk to violent extremism.
Al-Attar (2020)	Autism spectrum disorders and terrorism: how different features of autism can contextualize vulnerability and resilience Journal article that describes (theorizes) a functional link between ASD and terrorism engagement.
Allely and Faccini (2018)	Rare instances of individuals with autism supporting or engaging in terrorism: a reply A reply to Chown et al. (2018).
Chown et al. (2018)	Rare instances of individuals with autism supporting or engaging in terrorism: a response to Lino Faccini and Clare Allely A critique of Faccini and Allely (2017). Comments on the interpretation of other relevant research findings.
Corner and Gill (2021)	Psychopathy and terrorist involvement Book chapter that explores what terrorism studies researchers know about the role of psychopathology and other disorders in terrorist involvement, and how research knowledge on the topic has shifted over time, through four distinct paradigms.
Corner et al. (2016)	Mental health disorders and the terrorist: A research note probing selection effects and disorder prevalence Research note probing disorder prevalence in Corner and Gill’s (2015) earlier study (which reported strikingly different prevalence rates of mental illness and other disorders in two subtypes of terrorist actors: lone-actors and group terrorists).
De Marinis and Boyd-MacMillan (2019)	A mental health approach to understanding violent extremism RAN report identified on p. 3–4 that makes claims or suggestions relevant to the review question.
Faccini and Allely (2017)	Rare instances of individuals with autism supporting or engaging in terrorism Journal article that includes aims to establish ‘a functional connection between autism-based deficits and the terroristic threats, terrorism, and … naïve, empty terroristic threats or acts’ (p. 70).
Gill et al. (2020)	Systematic review of ‘mental health problems and violent extremism’ This systematic review interprets relevant prevalence studies (Weenink, 2015; Corner et al., 2016). It interprets the low and heterogeneous disorder prevalence rates across a range of studies, and across a range of disorders, as reflecting equifinality and multifinality (p. 58–61). Discusses the importance of comparison and control groups.
Knight et al. (2017)	‘Violent versus nonviolent actors: An empirical study of different types of extremism’ This study attempts to find variables that might differentiate violent actors (VAs) and nonviolent actors (NVAs). ‘Asperger Syndrome’ is proposed and tested as one amongst a series of variables that may potentially distinguish VAs from NVAs. This study finds a small difference (4.2%) in the prevalence rates of Asperger syndrome coded from the two groups (VAs and NVAs). However, this difference is not found to be statistically significant (i.e. it is deemed to be too likely to have been produced by chance rather than finding a real effect or being a difference produced by confounding).
Krasenberg and Wouterse (2019)	‘Understanding the mental health disorders pathway leading to violent extremism’ A RAN report identified on p. 3–4 that makes claims or suggestions relevant to the review question.
Little et al. (2021)	‘Online self-radicalization: a case study of cognitive vulnerabilities for radicalization to extremism and single actor terrorism’ Presents a case study of an individual with ASD who engaged in lone-actor terrorism. Data is drawn from clinical and medical records. The authors interpret the findings in terms of cognitive vulnerability.
Misiak et al. (2019)	‘A systematic review on the relationship between mental health, radicalization and mass violence’ Comments on the interpretation of findings in Corner et al. (2016).
Walter et al. (2020)	‘Characteristics of autism spectrum disorder and susceptibility to radicalization among young people: A qualitative study’ This study explores susceptibility to non-violent radicalization, rather than terrorism. Therefore the empirical data (interviews with practitioners) presented here is not included in this review. Also comments on the interpretation of Corner et al. (2016).
Weenink (2015)	‘Behavioral problems and disorders among radicals in police files’ Empirical study of a forensic sample of 140 Dutch foreign fighters using archival data drawn from several police databases. This study presents and interprets prevalence rates on variables (including ASD) and other behavioral problems.

Analysis

Original research findings and their interpretations

Table 4 lists included publications that contain original research, along with details of their research type (empirical, theoretical or critical research) and, for empirical studies, the research design, sample characteristics, data source and data capture methods. No critical research publications were identified. One publication (Al-Attar, 2020) was classified as theoretical research. All five empirical studies were observational studies. There are three descriptive studies: two utilize case studies (Faccini & Allely, 2017; Little, Ford, & Girardi, 2021) and one is a retrospective cohort study without a comparison group (Weenink, 2015). There are two analytical studies (both retrospective cohort studies with a comparison group). One of these studies has an inconclusive finding that is not statistically significant (Knight, Woodward, & Lancaster, 2017). Therefore, Corner, Gill, and Mason (2016) are the only analytical study with a conclusive research finding in the included publications.

Table 4. Included publications containing original research: study designs.

Citation	Research type and design	Sample characteristics	Data capture source and methods	Analysis methods
Corner et al. (2016)	Empirical, analytical study: a research note probing the retrospective cohort study from Corner and Gill (2015). Utilizes a matched comparison group to compare lone-actors and group terrorists (this is not an analytic study across terrorists and non-terrorists).	119 lone-actor terrorists and a matched comparison group of 119 group-actor terrorists. The matched variable is the country of attack.	Coding of archival sources	statistical regression analysis, descriptive statistics
Faccini and Allely (2017)	Empirical, descriptive study: case studies	Presents case studies of individuals with ASD engaging in terrorism acts, terrorism threats and empty threats of terrorism.	Archival data	Interpretation of case studies based on Al-Attar’s (2020) functional theory that connects ASD facets and terrorism engagement.
Knight et al. (2017)	Empirical, analytical study: no relevant, conclusive finding (fails to reject null hypothesis)	This analytic study conducts a quantitative analysis of data gathered from 40 cases of extremist individuals.	Archival data (coded from case studies)	Descriptive statistics
Little et al. (2021)	Empirical, descriptive study: case study	A case study of an individual with ASD who engaged in multiple acts of terrorism through online activity	Draws upon clinical interviews and psychometric testing	Analyses the case through the lens of cognitive vulnerabilities
Weenink (2015)	Empirical, descriptive study: retrospective cohort study (no control or comparison group).	convenience sample of Dutch foreign fighter terrorists	coding of forensic archival sources (police databases)	Descriptive statistics
Al-Attar (2020)	Theory: this article describes a functional link between ASD and terrorism engagement.	n/a	n/a	n/a

We now describe and analyse the research identified in Table 4.

Corner et al. (2016)

Corner et al. (2016)’s research note provides further analysis of data from Corner and Gill (2015). Corner and Gill’s (2015) study analyses a dataset of 119 lone-actor terrorists and a matched comparison group of 119 group-actor terrorists, matched by country of attack. Corner and Gill (2015) calculate the aggregated mental illness (and disorder) prevalence, including ASD, for their two subtypes of terrorists (lone-actors or group terrorists). They find that ‘the odds of a lone-actor terrorist having a mental illness [or a disorder, such as ASD] is 13.49 times higher than the odds of a group-actor having a mental illness’ [or a disorder, such as ASD] (Corner & Gill, 2015, p. 23). Corner and Gill (2015) also find a significant difference in mental illness and disorder prevalence rates for terrorist lone-actors (31.9%) compared to group terrorists (3.4%) (p. 24). 2 note that this finding replicates that of Gruenewald, Chermak, and Freilich (2013), who find a similar difference between diagnoses prevalence rates for far-right lone-actors (40.4%) and far-right group-actor terrorists (7.6%) (p. 562). In these studies lone-actor terrorists have a significantly higher prevalence of diagnoses when compared to group terrorists. Corner and Gill (2015) also compared their sample of ‘lone-actors with a history of mental illness to those without’, and found that ‘those with a history were significantly less likely to have some form of command and control link’ to terrorist groups (p. 30).

Corner and Gill (2015) hypothesize that the mechanism of selection effects may explain these differences. That is, ‘not everyone who wants to become a terrorist gets to be a terrorist’ (Corner & Gill, 2015, p. 30). As Corner and Gill (2015) write,

Horgan (2005) notes that particular individuals may never meet recruitment criteria because of overt psychological characteristics rendering them unsuitable. Spaaij (2010) explains that because of psychological conditions, certain individuals fail to become recruited despite demonstrating willingness, and act independently instead (p. 24).

Corner et al.’s (2016) research note undertakes further analysis by disaggregating Corner and Gill’s (2015) aggregated categories. Firstly, instead of just considering group and lone actors, Corner et al. (2016) disaggregate these categories further, splitting them across five terrorist actor subgroups ordered by ‘loneness’, where ‘loneness’ is the degree to which a terrorist acts alone or else in relation to a group and an ideology. They then compare the prevalence rate of aggregated mental illnesses and other disorders in each these terrorist subgroups to one another and to the population base rate (see Corner et al.’s (2016) Figure 1, p. 562). A positive association is found between loneness, which varies across the five different terrorist sub-types, and the aggregated prevalence of mental illness and disorders. In other words, across all five terrorist subgroups, subgroups that have more ‘loneness’ (or less links to kin in groups and to ideology) report higher prevalence rates of mental illness and disorders.

Corner et al.’s (2016) Figure 1 graphs the changing prevalence rates as they vary across the disaggregated terrorist actor categories (which vary by loneness), revealing a sloped line. This figure also graphs the aggregated rate or base-rate for all terrorist sub-types, before it was disaggregated by subtype, which is, of course, a single value aggregated or flattened across all subtypes, or a horizontal line. Showing the two prevalence relationships on the same axis is potentially confusing (although not technically incorrect). It’s not possible to compare the absolute values of these two lines (which is the way that some other authors incorrectly read this figure), since the two samples (population samples and terrorist samples) are not comparison groups (they are unmatched and not directly comparable).

Corner et al. (2016) do not analyse a direct comparison across the terrorist and base-rate groups. These two groups are not comparison groups. In fact, in their abstract, Corner et al. (2016) note that their research note is descriptive Instead, the difference that Corner et al. (2016) wish to highlight with their first figure is the difference between aggregated categories (which, by their very nature, are horizontal or artificially flattened baselines where we have averaged over many variables by choosing a broad unit of analysis) and disaggregated categories where something happens (in this case, a positive association is found between loneness and diagnoses).

Corner et al. (2016) interpret these findings as evidence for their argument that terrorism and mental illness is not a dichotomy (see also Corner & Gill, 2015). In other words, there are not only two possibilities: that mental illness (and other disorders) entirely explain terrorist engagement or else have nothing to do with it. Corner et al. (2016) conclude that ‘this research note provides a further call for terrorism researchers to be more exact; both in terms of the types of terrorist they are researching and by what exactly they mean by mental health problems’ (p. 565).

It should be noted that Corner et al.’s (2016) study design can say nothing about the risk of terrorism engagement. We can’t say that there is an over representation of diagnoses in lone-actor terrorists compared to the general population (that is not a valid way to interpret Corner et al.’s (2016) Figure 1), but we can say that there is an over representation or skew towards diagnoses in lone-actor terrorists compared to group terrorists. These two kinds of comparisons are different and they require distinct analytic study designs, different matched comparison groups, and their outcomes have very different implications. Only a study design comparing matched comparison groups of terrorists and non-terrorists (not terrorists and population base rates) can potentially make any findings related to the risk of terrorism engagement.

Corner et al.’s (2016) findings that lone-actor terrorists have more prevalent disorder and mental illness diagnoses compared to group terrorists, and that the degree of loneness is positively correlated with these diagnoses isn’t especially surprising since some diagnoses can potentially lead to isolation, and isolation is related to Corner et al.’s (2016) quality of loneness disaggregating terrorist actor subtypes. Corner et al.’s (2016) hypothesis that selection effects is likely to explain the observed differences makes sense. We should expect to find an increased prevalence of ASD among lone-actor terrorists compared to group terrorists since ASD can, in some cases, lead to isolation. Furthermore, we should also expect to find an increased prevalence of ASD in terrorist cohorts compared to randomized cohorts from the general population since terrorists are predominately young and male (Barracosa & March, 2021), ASD diagnoses are four times more prevalent among males (Centre for Disease Control, 2022) and ASD is more commonly diagnosed in young people (83% aged under 25) as the diagnoses rate has increased over time (Australian Institute of Health and Welfare, 2017).

If we aggregate Corner et al.’s (2016) actor subtypes and instead consider the prevalence rate of ASD for all terrorists, then the adjusted prevalence rate of ASD diagnoses for all terrorists is less than 1% (this figure is calculated by averaging the prevalence rates of ASD for group and lone actors (0% and 3.3%) and accounting for the fact that group actor terrorists are more prevalent among terrorists than lone-actors). This prevalence rate for ASD among terrorists from Corner et al.’s (2016) data is similar to the prevalence rate calculated by (Smith et al., 2020) in a non-peer-reviewed, unpublished study of 95 convicted, domestic Australian terrorists to 2020 (excluding foreign fighters). After coding all available legal court documents from these 95 domestic terrorist cases, Smith et al. (2020) report one (n = 1) individual with a formal ASD diagnoses, which gives a prevalence rate for ASD of just over 1% for this terrorist data (which is not so much a sample of domestic terrorists in Australia, but includes all convicted Australian terrorists up to 2020).⁶

These remarkably low prevalence rates for ASD among convicted terrorists from these two studies suggests that ASD is not a major factor in terrorism engagement. Of course, there is always the possibility of missing data, but it would be more likely to expect an over-diagnoses of ASD among convicted terrorists compared to the general population since the convicted terrorist group has come into contact with the criminal justice system and is assessed by psychiatrists at a much higher rate than the general population. In summary, Corner et al. (2016) make no findings that demonstrate or suggest that individuals with ASD are any more likely or open to engage in terrorism than anyone in the general population.

Faccini and Allely (2017)

Faccini and Allely (2017) interpret a series of case studies that describe people with ASD engaging in terrorism acts, terrorism threats and empty threats of terrorism. The case studies are categorized by increasing levels of engagement according to Kruglanski et al. (2014) degree of radicalization scheme (i.e. naïve threats, real threats, passive or active supporters or fighters). Summarizing their findings, Faccini and Allely (2017) write that ‘essential to the analysis was establishing a functional connection’ between ASD and terrorism acts, ASD and terrorism threats, and ASD and empty threats of terrorism (p. 70). Faccini and Allely (2017) claim that their case studies ‘sought to clearly demonstrate’ the functional connection between ASD and terrorism (p.79). However this aim cannot be met by Faccini and Allely’s (2017) research design. A descriptive study, such a case study, can suggest or generate a hypothesis for a functional connection between ASD and terrorism, but only an analytic study with a control or comparison group can ‘clearly demonstrate’ why outcomes occur. Faccini and Allely (2017) do note that ‘only case studies could be presented that suggested how an ASD contributed to individuals affiliating to or engaging in terrorist acts’ and that ‘further studies are needed before more firm conclusions and interventions can be offered and developed’ (p. 79).

Faccini and Allely (2017) explain that Kruglanski et al. (2014) degree of radicalization scheme ‘provides a model for connecting the symptoms of autism and the different means of engaging with terrorists’ (p. 72). Kruglanski et al. (2014) scheme provides a model for stratifying different subtypes of terrorists, based on their degree of radicalization, which applies to all terrorists. Although, Faccini and Allely (2017) describe a hypothetical functional connection between ASD features and Kruglanski’s scheme, they don’t establish why an individual with ASD may fall into one category rather than another, or that this stratification scheme, or terrorism engagement more generally, is necessarily correlated with ASD. Faccini and Allely (2017) draw upon Al-Attar’s (2016) functional theory to link ASD facets (or symptoms) to push and pull factors for terrorism engagement (see also Al-Attar, 2020).

Faccini and Allely (2017) reductively describe how certain facets of ASD create a feedback loop that, in hindsight, can accommodate all possible degrees of radicalization to terrorism in Kruglanski’s stratification. By reductive, we mean that ASD facets alone provide the mechanism through which an individual’s radicalization and increasing engagement happens in the absence of other behavioral indicators or putative risk factors identified by terrorism studies. Such mechanisms are poorly understood in general, even for terrorists who do not have ASD. This interpretation is too simplistic and fails to account for terrorism studies’ account of terrorism, which informs, but which doesn’t entirely describe or explain, pathways to terrorism engagement in general.

The critique of psychological reductionism isn’t original. Bhui, James, and Wessely (2016) argue that ‘we are too ready … to propose mental illnesses as the explanation behind such complex behaviors’ (p. 1). ‘A psychiatric diagnosis, where appropriate is important, but it does not explain motivation – diagnosis will interact with prevailing social and cultural concerns’ (Bhui et al., 2016, p. 1). Similarly, Corner and Gill (2018) write that ‘in the absence of rigorous clinical and empirical procedures, the reductionist view, where terrorists are characterized as suffering from some mental disorder purely on the nature of the attack behavior, ignores … highly complex neurological, psychological and sociological processes’ (p.147). These authors also note that ‘research is yet to empirically determine at which point the experience of psychiatric symptoms is relevant to violent radicalization. Depending on circumstance, it may be a catalyst, an inhibitory factor, and even a consequence’ (Corner & Gill, 2018, p. 147).

Meloy, Habermeyer, and Guldimann (2015), and Erlandsson and Meloy (2018) show how it is possible to frame interpretations of case studies on the intersection of ASD and terrorism engagement in a less reductive way.⁷ Instead of framing their interpretations wholly in terms of why engagement takes place and providing a reductive interpretation where an ASD facet or symptom appears as a radicalization increasing mechanism at a series of transition points, Meloy et al. (2015), and Erlandsson and Meloy (2018) frame the question of terrorism engagement in a case study also involving ASD in a more complex way, by asking how engagement happens, not just in relation to ASD, but through the prism of a terrorism specific psychometric instrument, TRAP-18. The authors consider how ASD symptoms may interact with the TRAP-18 indicators that apply to all terrorists, regardless of their mental health or neurodevelopmental disorder status (Erlandsson & Meloy, 2018; Meloy et al., 2015).

Little et al. (2021)

Little et al. (2021) present a case study, based on a patient’s clinical record, of an individual with ASD who was convicted of terrorism offences in the UK. John (a pseudonym) ‘was convicted of terrorism offences committed between 2016 and 2018 within the context of anti-Muslim hate crimes, including soliciting to murder, inciting violence and hoaxes’ (Little et al., 2021, p. 114). John’s offences involved sending letters (Little et al., 2021).

Little et al. (2021) suggest that ‘impaired cognitive flexibility may potentiate radicalization through concrete thinking, a vulnerability to more essentialist ideologies (such as political extremism) and reduced ability to switch between alternate beliefs due to a reduced capacity to weigh contradictory evidence in one’s mind’ (p.118). They point out that

this is consistent with John’s self-reported repetitive and obsessive computer-based anti-Muslim research, stereotypical behavior and his endorsement of extreme right-wing political beliefs. He described himself as ‘not being able to stop' repeating the behaviour, possibly stimulating his need for routine and repetitiveness, characteristic of ASD. (Little et al., 2021, p. 118)

Little et al. (2021) describe how ‘the [radicalization] process continued in a repetitive cycle, with John becoming increasingly radicalized until he took action’ (p.118).

Little et al.’s (2021) conceptualization of radicalization as a cognitive or ideological process is at odds with the terrorism studies and critical terrorism studies critiques of radicalization that understand the process of becoming a terrorist as being more than ideological and including an interplay of psychological, social, emotional and behavioral factors (Youngblood, 2020, p. 2). Again, like Faccini and Allely (2017), Little et al. (2021) propose that ASD reductively explains the complex mechanism of radicalization to terrorism engagement; in hindsight, ASD can neatly and entirely account for the entire radicalization to terrorism pathway. However, like many accounts of radicalization, Little et al. (2021) also fail to explain when, why and how radicalization becomes terrorism in only a vast minority of cases. In terrorism studies, this lack of explanation is called the specificity problem.

Little et al. (2021) claim that ‘the present study illustrates the case of lone-actor terrorism where the interplay between motivational (Kruglanski et al., 2014), cognitive and behavioral characteristics (Decety et al., 2017) associated with ASD made the individual susceptible to radicalization and terrorism’ (p. 114). However, it’s not possible to know whether or not this illustration is the right one because it is not possible to know why an outcome occurs using a descriptive or case study design. There is no existing evidence with which to support Little et al.’s (2021) illustration; this remains one of many possible interpretations or hypotheses.

Weenink (2015)

Weenink (2015) reports the prevalence rate of individuals formally diagnosed with ASD in a sample of 140 Dutch foreign fighters (n = 1 or 0.7%) (p. 24). Archival data is drawn from several police databases. Weenink (2015) interprets his prevalence findings in general (or by aggregating different diagnoses) by combining prevalence rates for mental illnesses and other disorders (including ASD) and comparing the combined prevalence rate (6%) to that of the general population, concluding that ‘6% in our sample people with a diagnose[s] … are overrepresented’ (p. 27). Weenink (2015) interprets this result by claiming that ‘the results are at odds with the consensus view on terrorists alleged “normality”’ (p. 17).

However, as we argued in our analysis of Corner et al. (2016) above, it isn’t possible to find an overrepresentation from a descriptive study by comparing unmatched base-rates to a forensic terrorist sample. A whole series of likely confounding factors, many of which relate to the criminological nature of Weenink’s archival database, mean that we shouldn’t expect forensic cohorts to represent or to match base-rates from the general population. Reviews of general (or non-terrorist) violence and ASD show a marked heterogeneity in findings across studies, which suggests that confounding factors are likely to be important.

For example, Im (2016) notes that ‘sample-selection issues must be considered (e.g. forensic psychiatry samples often produce higher rates of ASD, as these subjects are likely to have mental health needs of some kind)’ (p. 25). Im (2016) writes that ‘regarding prevalence studies [for general violence], those conducted in forensic settings generally found an overrepresentation of ASD, but selection biases undercut any conclusions that this overrepresentation reflects greater violence among individuals with ASD’ (p.29). ‘On the whole, prevalence studies have provided no persuasive evidence that individuals with ASD are more violent than those without ASD’ (Im, 2016, p. 29).

Al-Attar (2020)

Al-Attar (2020) notes that ‘there is no empirical evidence to link autism and terrorism, in the general population’. In other words, there is no evidence of a correlation between ASD and terrorism engagement. Al-Attar (2020) then writes,

however, when terrorist acts are planned or executed by individuals with autism, it is important to develop an understanding of the individual’s autistic functioning and how it may contextualise factors that push them towards terrorism and aspects of terrorism that may pull them in, in order to manage and reduce risk. (p. 926)

But these two statements cannot co-exist; measuring risk requires a correlation to be known. It is unknown whether or not it is important, or even necessary, ‘to develop an understanding of the individual’s autistic functioning … in order to manage and reduce risk’ (Al-Attar, 2020, p. 926). However, we note that it may be necessary to do this in order to meet an individual’s needs.

Al-Attar’s (2020) aims to consider how ASD ‘may contextualize’, which means to frame or to shape risk. And yet, Al-Attar’s (2020) functional theory describes how ASD hypothetically explains and how it functionally effects (and increases) the risk of terrorism engagement. Al-Attar (2020) describes ASD as a risk factor, not as a contextual factor or situation within which terrorism risk plays out. As we noted in our discussion of Faccini and Allely (2017), Meloy et al. (2015), Erlandsson and Meloy (2018) do the latter.

Al-Attar (2020) defines her unusual semantic usage of the term contextualizes by explaining that

ASD is not causal to risk and all efforts should be made to avoid stigmatising ASD, a lifelong neurodevelopmental condition, and to conceptualise its role as ‘contextual’ rather than causal to both risk and protection. In other words, features of ASD may shape experience, functioning and behaviour that may in some instances exacerbate an individual’s vulnerability, as opposed to cause the vulnerability. (p. 927)

We strongly agree with Al-Attar (2020) that ‘all efforts should be made to avoid stigmatizing ASD’. And Al-Attar (2020) also usefully notes that ‘caution is encouraged against drawing conclusions of causality and over-simplification of autism-terrorism links, especially in individuals with co-existing mental health problems and psychosocial adversities’ (p. 926). But Al-Attar (2020) also appears to misunderstand the difference between causality, correlation and context here. As Monahan (2012) also points out in defining risk, the concept of risk doesn’t require causes, but rather, correlations (p. 7). The question of risk is not whether or not ASD causes terrorism engagement, but rather, whether there is a correlation or an association between the two variables of ASD status (whether someone is formally diagnosed with ASD or not) and terrorism engagement (whether an individual engages in terrorism or not). But a correlation is also more than a context or the wider situation in and through which something, such as terrorism risk, takes place.

Context frames scientific variables and effects and the questions we pose about them, in a particular way. Correlation is an association between two variables, pointing to a relationship or an effect, even if an indirect one or one where causes may be unknown. Correlation shows that two variables have a relationally, even if one may not cause the other. Correlation and causation take place and are measured in context, but like risk, vulnerability and needs, these three concepts of context, correlation and causation are all distinct. When Al-Attar (2020) uses the term contextualizes as in shapes, she doesn’t mean situationally frames, she is really referring to a correlation where the risk of terrorism engagement changes (she means increases), even if indirectly, even if only in some circumstances, and even if for only some people.

In Al-Attar’s (2020) functional theory, ASD is not the context for terrorism (as, for example, in Meloy et al.’s (2015), and Erlandsson and Meloy’s (2018) framing of ASD through TRAP-18, as discussed above). But rather, Al-Attar (2020) theorizes how ASD may function to heighten the risk of engagement, at least for some individuals with ASD in some cases. In Al-Attar’s (2020) functional theory, ASD entirely or reductively accounts for how and why some individuals engage in terrorism (but not the specificity problem, or how, when and why the vast majority of individuals with ASD don’t follow this functionalism). This account also poses another specificity problem: that is, it fails to specify why the ASD pathway into terrorism should be specific for individuals with ASD.

Youngblood (2020) notes that reductive and individual accounts of terrorism explained only in terms of psychopathology or other disorders fail to capture the social processes of radicalization studied by terrorism studies scholars. Youngblood (2020) explains that,

there is little evidence that radicalisation is primarily driven by psychopathology [or disorders] (Misiak et al., 2019; Post, 2015; Webber and Kruglanski, 2017). Rather, radicalisation appears to be a process in which individuals are destabilised by various environmental factors, exposed to extremist ideology, and subsequently reinforced by members of their community (Becker, 2019; Jasko et al., 2017; Jensen et al., 2018; Mills et al., 2019; Webber and Kruglanski, 2017). Even ‘lone wolves', or solo actors, often interact with extremist communities online (Holt et al., 2019; Kaplan et al., 2014; Post, 2015) (p. 2).

Al-Attar’s (2020) functional theory is not only reductive, it is also partial or biased. That is, it only considers how ASD hypothetically functions to increase risk, but not how ASD hypothetically functions to reduce risk or perhaps not to change it. Al-Attar (2020) claims to explore ‘how different features of autism may contextualize risk and resilience and the implications for support and diversion approaches’ (p. 926). But, while Al-Attar’s (2020) functional theory theorizes how and why features of ASD may hypothetically increase an individual’s risk of engaging in terrorism by describing how facets of ASD function as ‘push’ and ‘pull’ factors for terrorism engagement, it only considers resilience in connection with risk and as being the opposite of risk. This is not how purported protective and risk factors are conceptualized in terrorism studies. Resilience is not a lack of risk, and vice versa. Rather, purported protective and risk factors may often be different factors, not decreases or increases in a single factor.

For example, in their review of ASD and general offending, Rutten, Vermeiren, and Nieuwenhuizen (2017) note that, ‘specific symptoms of ASD can [in theory] predispose individuals to offending behaviour’, while ‘on the other hand, it has also been argued that some symptoms of autism protect people with ASD against involvement in criminal behaviour’ (p. 2). There is no theoretical or empirical reason to suspect that ‘push and pull factors’ that mobilize individuals into terrorism or purported risk factors are more relevant than purported protective factors. We do not know how or when protective and risk factors for terrorism engagement may interact in practice.

Rutten et al. (2017) write that ‘many people with Asperger’s syndrome have an overactive sense of right and wrong and are usually conscientious and unwilling to break the law’ (p. 2). Reviews of the role of ASD in non-terrorism offences consider the possibility that ‘the abnormal or restricted interests that are typical of ASD can play a role in delinquent behavior’ (Rutten et al., 2017, p. 1). However, these reviews also consider the possibility that restricted interests and law abiding behavior could also theoretically protect individuals from general violence. Al-Attar’s (2020) functional theory should also consider the possibility that ASD may also be protective against, or may perhaps have no measurable effect on terrorism engagement. Or perhaps Al-Attar’s (2020) theory could more clearly state its scope and limitations as a partial theory at the intersection of ASD and terrorism.

Al-Attar (2020, p. 928) describes seven ‘facets’ of ASD, which are not ASD symptoms, but are ‘a mixture of diagnostic and research evidenced correlates of ASD’, and then describes (or rather, theorizes hypothetical) functional links between these facets of ASD and ‘push and pull’ (i.e. engagement) factors for terrorism. Al-Attar (2020) writes that these ASD facets ‘have been postulated to contextualize risk and protection in offenders with ASD’ (this statement is followed by a list of self-citations) and that, ‘when it comes to contextualise risk, its [each facet’s] role in risk needs to be reduced and its role in resilience enhanced’ (p. 928). However, since Al-Attar’s (2020) functional theory is hypothetical, it is hasn’t been postulated. A postulate is an axiomatic or obvious truth that everyone agrees upon, and which is deduced analytically, through deductive logic. Of course, in scientific fields where postulates are common, such as pure mathematics and theoretical physics, theory, which includes postulates, must also be tested empirically. However, in this case it is not obvious, and it does not follow theoretically from analytic logic, that ASD effects risk or protective factors for terrorism engagement. While Al-Attar’s (2020) description may be useful as a description of a hypothesis of how ASD facets may function to increase terrorism engagement, this hypothesis remains untested.

Al-Attar (2020) concludes by writing,

ASD does not constitute a risk factor for terrorism in the general population and even in terrorist offenders with an ASD diagnosis, there is no evidence to suggest that ASD plays a causal role. Instead, different aspects of ASD may interact to contextualise push and pull factors in a nuanced way. Such a contextual role may differ from individual to individual and it is imperative that an individualised case formulation approach is used to identify if and how facets of ASD contribute to push and pull factors. (Al-Attar, 2018, p. 945)

But ‘an individualized case formulation approach’ or in research terms, a descriptive case-study, cannot identify if ‘facets of ASD contribute to push and pull factors’ (Al-Attar, 2020, p. 945). A case study may identify how ‘facets of ASD contribute to push and pull factors’, but only if the former turns out to be true. Only an analytic study design can answer questions about why individuals engage in terrorism. Only an analytic study design can answer the question of whether or not ASD and terrorism risk are correlated or not. In the absence of empirical evidence correlating ASD with terrorism engagement, it is not possible to ‘identify’; one can merely hypothesize, such an association.

Al-Attar (2020) notes that terrorism ‘risk assessment frameworks advocate a structured professional judgement [SPJ] approach that uses individual case formulation to take into account how factors, including ASD and mental illness, can contribute to vulnerability and risk (Heide et al., 2019)’. This is true. However, we note that mental illness items are not always included as risk factors or as part of the risk assessment in all SPJ assessment frameworks. In some frameworks, a needs analysis is separated from a prioritization or threat analysis, and mental illness items inform the former, but not the latter. We also agree with Al-Attar (2020) that terrorism risk assessment frameworks advocate an SPJ approach as best practice. But where the S meets the P in structured professional judgement doesn’t mean that clinicians are free to judge why something happens to an individual in the absence of any analytic or correlational evidence that there exists an association between a putative risk factor and an effect. The indicators that give an SPJ judgement structure are themselves structured by another s: science.

Structured professional judgement calls for clinicians to make judgements that are both evidence-based (which, in this case crosses the research fields of psychology and terrorism studies) and contextually situated (or framed by their clinical practice experience) – not for one to displace the other. In the discussion, we will argue that, although there are modern tensions that arise here between research and practice, that neither clinical context nor research evidence should drop out of the question.

Interpretations of findings by other authors

De Marinis and Boyd-MacMillan (2019) claim that ‘certain diagnoses, such as the neuropsychiatric disorders of schizophrenia or autism spectrum disorder (ASD), are over-represented in individuals involved in violent extremism or at risk of violence, whether as lone-actors or otherwise’ (p.2). Similarly, Allely and Faccini (2018) write that, ‘ASD does appear to be present in more than 1 per cent [a historical ASD base-rate] of the cases (Corner et al., 2016). In Corner et al.’s (2016) sample of 153 lone-actor terrorists, they found that 3.3 per cent had ASD’ (p. 64). Walter, Leonard, Miah, and Shaw (2020) also report that ‘previous studies suggest an overrepresentation of autistic people among terrorist offenders, specifically lone-actors, as compared to the general population’ (p. 408).

But as our analysis in the previous section finds that Corner et al.’s (2016) findings do not show that ASD is overrepresented amongst terrorist cohorts to non-terrorist cohorts. There are no existing analytic studies that find a correlation between ASD and terrorism engagement, or that lead to the interpretation that individuals with ASD are more open to terrorism engagement or any more likely to engage in terrorism than anyone else. Walter et al. (2020) note that ‘it is impossible to draw conclusions, and suggesting a potential link between ASD and the susceptibility to radicalization or terrorism would be inappropriate, risking further stigmatization of individuals with ASD’ (Walter et al., 2020, p. 411).

De Marinis and Boyd-MacMillan (2019) claim that, ‘while only a very small percentage of individuals [are] being radicalized, abnormal functioning can increase risk and/or vulnerability to recruitment’ (p. 2). Faccini and Allely (2017) discuss the case of an individual with Asberger’s syndrome who demonstrated a restricted interest involving trains (‘he had a lifelong history of taking vehicles and impersonating train personnel’) (p. 73). They write that, ‘Mr. G poignantly made the case, that because of his specialized knowledge of various train systems he was at risk for being recruited or abducted by terrorists’ (Faccini & Allely, 2017, p. 73). However, there is no empirical evidence with which to support these claims.

In their review of non-terrorist outcomes and ASD, Rutten et al. (2017) review studies on the prevalence of ASD in forensic populations (in this case delinquents), as well as reviewing studies on the prevalence of offending in people with ASD. They found that prevalence rates varied widely between studies for both of these questions, and suggested that this heterogeneity ‘might be due to the use of different diagnostic instruments, the diversity of the samples, the high rate of comorbid psychiatric disorders and the various types of offending behavior’ (Rutten et al., 2017, p. 1). This heterogeneity suggests that confounding factors are likely to be important.

Rutten et al. (2017) find that, on the one hand, ‘the prevalence of ASD diagnoses, particularly Asperger’s syndrome, in forensic settings is remarkable because it is much higher than the prevalence of ASD diagnoses in the general population’ – but, on the other hand, ‘contrary to our expectations, the prevalence of delinquency was lower in all the samples of people with ASD than in the general population’ (p. 7-8). But perhaps, by these author’s own logic, this result isn’t remarkable at all, since these two samples (ASD in delinquents versus delinquency in ASD) are different, the archives from which data is captured may be different and they may have different sets of confounding factors. In line with our findings from this review on ASD and terrorism, in their review of ASD and delinquency, Rutten et al. (2017) ultimately find that, ‘we cannot conclude from our analysis that people with ASD are more likely to offend than the general population’ (p. 1).

Further claims or suggestions linking ASD to terrorism risk or vulnerability

A number of authors link ASD status to the risk of terrorism engagement. Faccini and Allely (2017), Allely and Faccini (2018), Al-Attar (2020) (as discussed in earlier) and Krasenberg and Wouterse (2019) all explicitly associate ASD with the risk of terrorism engagement. Faccini and Allely (2017) note that their ‘analyses were completed to understand the rare occurrence when someone with an ASD does support or engage in terrorism, so to better understand the risk factors and prevent or minimize the extent that these individuals are exploited or victimized’ (p. 72). They posit that, ‘although there is no conclusive evidence supporting the notion that individuals with ASD are more violent compared to individuals without ASD, there may exist specific generative and associational risk factors which may increase the risk of offending among individuals with ASD’ (p. 77). However, this hypothesis appears to be derived from a misinterpretation of the prevalence rates from Corner et al.’s (2016) study, or perhaps Al-Attar’s (2020) functional theory, which is less a well-developed theory and more a description of a hypothesis.

Similarly, Krasenberg and Wouterse (2019) ‘identify risk factors that make these people [individuals with ASD] vulnerable to [engaging in] violent extremism’ (p. 1). Krasenberg and Wouterse (2019) also conclude that it is ‘important [for clinicians] to define the specific aspects of the … disorders that pose a risk and might lead to radicalization’ (Krasenberg & Wouterse, 2019, p. 11). And Allely and Faccini (2018) note that ‘we made a number of points in our paper about why ASD is actually more likely to be “the condition to blame”’ [original emphasis] for terrorism engagement and consequently, that ‘it is difficult to not conclude that some individuals with ASD might be at high risk of being involved in these terrorist acts’ (p. 65). Our analysis doesn’t run into such a difficulty. Our review finds that it is entirely unknown and not very well theorized what effect, if any, ASD may have on an individual’s risk or vulnerability of terrorism engagement.

Discussion: reframing the research-practice debate

Walter et al. (2020) note that the debate surrounding the prevalence of ASD among terrorists is ongoing (p. 410). Krasenberg and Wouterse (2019) describe this debate in some detail, from the clinical practitioner viewpoint, writing,

‘in recent years, practitioners working in the field of P/CVE have pointed to a potential link between mental health disorders on one side, and radicalisation on the other. However, academic research on the prevalence of mental health disorders among terrorists has failed to produce clear evidence supporting this hypothesis. In fact, studies conducted since 2012 seem to repeatedly reach the conclusion that there is no clear connection between mental health disorders and terrorism … . According to practitioner experiences, we need to be cautious about the literature we rely upon. In autism and mental health work, there are publications by passionate and excellent academics, but they don’t always have experience of clinical trials (CT) with patients and, in some cases, autistic patients. This can lead to a gap between academic literature and real life practice and experiences. Some of the evidence that practitioners have gathered is not publishable, which could be explained by difficulties encountered with sharing and publishing confidential data, due to medical confidentiality agreements, legislation (GDPR) or governmental disclosure limitations.

We argue that Krasenberg and Wouterse (2019) express valid concerns about the framing and translation of academic research into practitioner or clinical contexts. These concerns are echoed by the critique of modern science in the field of field of science studies, and are perhaps best described by what that Donna Haraway calls ‘situated knowledges’ (Haraway, 1988). Haraway (1988) addresses the research-practice debate described by Krasenberg and Wouterse (2019) above, and also the science/society or human/non-human binaries that framed an earlier and related debate between feminist humanities scholars, on the one hand, and scientists, on the other hand. These kinds of debates can be understood as the undesirable or paradoxical effects of being ‘modern’ which, as Latour’s (1993) brand of science studies describes, is an operation that purifies a whole series of categories or binaries, such as research and practice, evidence and experience, nature and culture, science and society, … and so on, and then attempts to translate across them. But modern translations (which tend to involve mixtures or hyphens or bridges) invariably fail because you can’t put back together two categories that never came apart like that.

Haraway (1988) reframes modern debates over evidence and experience, which is related to the research practice problem, by pointing out that no one gets to win the argument. Haraway (1988) points out that that the two sides of a modern debate (feminists and scientists, or in this case, by extension, researchers and practitioners) are both right and are both wrong, but about distinct, although related, things that (although they appear to us now as divided) never really came apart like that. Haraway (1988) critiques the way in which modern science attempts to ‘de-locate’ or ‘de-situate’ research evidence and knowledge. Latour’s (1993) critique of modern science, that science ‘has never been modern’, similarly argues that science never really managed to purify the hermetically sealed laboratory out of the world, or to properly leave out the confounding social, political, philosophical and historical conditions in and through which science is produced.

Can science studies help us to understand, but also to reframe, Krasenberg and Wouterse’s (2019) concerns that the lack of evidence linking ASD and terrorism engagement doesn’t quite fit their clinical experience? Haraway’s (1988) situated knowledges didn’t just critique scientists and the way scientific evidence is framed in translation. There is a second distinct, but not unrelated, translation problem at play here too. Haraway herself was originally trained as a scientist, as well as a feminist science studies scholar. As well as critiquing scientists from the point of view of feminist humanities in the scientist/feminist version of the debate, Haraway (1988) also critiques feminist humanities scholars from the point of view of science. No one gets to win this debate. And no one gets to entirely lose it either. Haraway (1988) pointed out that feminists were guilty of underutilizing or even openly rejecting scientific evidence because it didn’t address their needs or capture their experience. Haraway’s version of the feminist/scientist science/humanities debate from thirty years ago sounds remarkably similar to the researcher/practitioner version of the ASD-terrorism debate described by Krasenberg and Wouterse (2019) today.

We are not suggesting that Krasenberg and Wouterse (2019) are explicitly calling upon practitioners to reject research in favor of experience (it is possible that they are merely illustrating science’s version of the situated knowledges problem: that modern science isn’t the view from nowhere). Rather, we argue that both parts of Haraway’s (1988) double-provocation must be put into action here in order to reframe these kinds of modern debates. That is, we must open the two distinct, but related, questions of knowing situated science or situated research evidence and the question of doing evidence informed practice onto one another. Haraway’s (1988) reframing of modern debates as double-sided doesn’t merely involve building a bridge or placing a hyphen between the categories of research and practice or evidence and experience. Instead, Haraway (1988) reframes the modern debates between science and society, and research and practice, from both sides, changing both of them.

Modern clinical practice must be framed by research evidence, while modern knowledge must be framed by clinical practice experience; one does not trump, and nor can it be truly separated from the other. While Krasenberg and Wouterse (2019) are right to point out that modern researchers have a job to do that entails framing and situating research questions and findings in relation to practice, context and situation, this isn’t the same as saying that clinical experience can ever displace research evidence (or vice versa). This argument calls for more, not less, collaboration between researchers and practitioners.

Shiela Jasanoff (2006) once remarked that ‘science and secrecy do not sit comfortably together’ (p. 21). Research transparency, which is one of the hallmarks of modern science, is just as important as patient confidentiality (Hales, Wesselmann, & Hilgard, 2019; Klein et al., 2018; The PLOS Medicine Editors, 2014). We contend that it is possible to design transdisciplinary collaborations and co-design research in clinical and security settings in such a way that transparency and confidentiality, which are distinct concepts, can be both be achieved. Evidence and experience are also not the same thing; we need both. Evidence and experience have never been modern, which is to say that no one will ever entirely win or lose a debate that has been purified and translated across them. Researchers and practitioners must find novel ways to open these categories, and what they each know and do, onto one another.

Conclusion

Our analysis highlights the importance of understanding the role of comparison or control groups in analytical research designs, and the limitations inherent in the interpretation of descriptive research designs. It is important for researchers, and for practitioners and clinicians interpreting research, to understand that descriptive research can only generate, rather than test hypotheses, and cannot find out why outcomes occur.

Our findings resonate with Im’s (2016) review on non-terrorism offences and ASD. ASD does not appear to be, as Im’s (2016) title puts it, ‘a template to perpetrate’. However, this is not the same as saying that ASD and terrorism engagement are entirely unrelated (Corner et al., 2016; Corner & Gill, 2015). This review finds that there is a lack of research evidence with which to test claims or hypotheses that ASD increases an individual’s risk of, or vulnerability for, terrorism engagement. Existing research cannot conclude that individuals with ASD are any more vulnerable to, or any more at risk of, terrorism engagement than individuals in the general population.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 For an example of recent media coverage the topic, see, for example, Grierson (2021), and Moseley’s (2021) response to this.

2 See also Mental Health Europe’s response to these reports (Mental Health Europe, 2019).

3 Im (2016) notes that symptoms of ASD can sometimes be confused with other conditions and that it is important to distinguish between these (p. 16). For example, Anders Breivik presents a case of a terrorist whose mental illness or disorder status (including his ASD status) is disputed by professionals (see Fargnoli, Trevisan, Fargnoli, & Belli, 2013). For this reason, Faccini and Allely (2016) is not included in this literature review.

4 Freckelton (2013) mentions a case study were a terrorism conviction was legally overturned because an individual with ASD was found not to have the cognitive ability to form the political intent required for a terrorism offence. Although the legal definition of terrorism varies across jurisdictions and individuals with an ASD diagnoses can have a wide range of cognitive ability, and therefore Freckelton’s (2013) observation will not necessarily be true of each case, we note the possibility that an individual’s ASD status may perhaps affect the cognitive ability required to form an intent to engage in and to use political violence or terrorism.

5 There is a double-meaning here and we wish to evoke both possibilities. Firstly, a radicalisation process leading to terrorist engagement is not always evident for all terrorists, and therefore may not be present. And secondly, we acknowledge that the very concept of radicalisation and its various conceptualisations, is a contested concept in terrorism studies. There is a widely held consensus in most of the research literature on terrorism that radical ideology, in and of itself, isn’t the only driver of terrorism engagement. Social, political and emotional elements are required in order to describe and explain terrorism engagement.

6 Three of the authors of this literature review, Smith, Spaaij & Druitt, were also authors of this study, which was conducted through the Australian Security Sciences Partnership (ASSP) for Victoria Police. This study is not included in the review since it is not peer-reviewed or published in the public domain. This study did not focus specifically on ASD, but mental illness and disorders were coded from court documents, allowing a prevalence rate for ASD to be calculated for the purposes of this review.

7 These studies are excluded from this literature review because there is no formal ASD diagnoses in these cases.