PROTOCOL: Counter-narratives for the prevention of violent radicalisation: A systematic review of targeted interventions

The problem, condition or issue

Since the start of the 21^st century, academics and counter-terrorism practitioners have increasingly sought to better understand the process(es) through which individuals transition from non-violence into terrorism (Dawson, 2017; Europol, 2016; Horgan, 2008; European Parliament, 2017). This progression, believed to be characterised by the movement from an unfocused state of uncertainty towards a narrower, unambiguous state of clarity (Horgan, 2008), is often referred to as ‘violent radicalisation’. At least in part, research into violent radicalisation has been motivated by a desire to identify ‘phase specific intervention strategies’ in the radicalisation process (Horgan, Gill, Bouhana, James Silver, & Corner, 2016).; Identification of these “pinch-points” may provide opportunity for intervention, helping redirect individuals onto a trajectory towards non-violence (p. 53).

One core focus of enquiry by social scientists has been the role of narratives in the process of violent radicalisation (e.g. terrorist organisations’ use of narratives for propagandistic purposes (Braddock, 2015)). Narratives here refer to recollections of events which happen in sequence (Barthes &Duisit, 1975; Genette, 1982) with characters that can cause changes (Richardson, 2002), neatly contained within an identifiable beginning, middle and end (Hinyard&Kreuter, 2007, p. 778). The following is a popular religious narrative found in numerous religious texts (including the Qur'an and The Torah);

Moses, having pleaded with The Pharaoh of Egypt to release the Hebrews and accept the One True God, threatened Pharaoh with divine retribution. The Pharaoh was arrogant and ignored Moses’ warning. As promised God punished Pharaoh with several disasters such as drought, famine, disease, locusts, lice and frogs brought upon his own people.

The objectives of such narratives are to present a social construction of the world which serves the interest of those legitimizing violence. In this instance, one can justify the murder of innocent civilians as retribution for not accepting a particular worldview.

The intervention

To tackle these violent messages, researchers have considered the potential for counter-narratives messaging. Briggs and Reve (2013) differentiate between three approaches to counter-messaging, including alternative narratives, government strategic communications and counter-narratives. Whilst the former attempt to promote positive stories about social values such as openness and tolerance (alternative narratives) or raise awareness about positive government initiatives (government strategic communications), counter-narratives are conceptually different. By deconstructing, discrediting and demystify violent extremist messaging, counter-narratives present people with alternative social constructions to those presented by the propagandist. They are described as tailored responses which challenge the themes intrinsic to extremist narratives (Braddock & Horgan, 2016, p. 386). Taking the narrative above, for instance, a counter-narrative could state that;

In the divine retribution, it was God, not Moses, who rained terror on his people. In this way, the murder of innocent civilians by man is nothing more than the very behaviour for which Pharaoh, a man, was punished. Only God can offer divine retribution.

Through challenging such narratives, counter-narratives may help individuals make more informed decisions through deeper consideration of, for example, the validity of certain arguments, the rationality of hatred or the legitimacy of violent action (Goodall Jr, 2010).

In applied research, counter-narratives have attempted to challenge prevailing narratives in a number of areas, including discrediting inaccurate historical narratives (Luckhurst, 2012) and news coverage (Lee, 2012). In the field of violent extremism, the concept of a counter-narrative is not new. Following the 9/11 attacks, counter-narratives were used to target attitudes about the treatment of Muslims in the United States as part of the US State Department's ‘Shared Values’ initiative (Fullerton & Kendrick, 2006). The campaign included five mini-documentaries which were broadcast in Muslim majority countries, showing the happy lives of American Muslims, living freely in America without persecution.

More recently, the Organization for Security and Co-operation in Europe (OSCE) incorporates counter-narratives to promote interfaith dialogue and for young people at risk of violent radicalisation. The campaign also uses documentary style videos promoting themes of acceptance and tolerance (targeting polarised attitudes) under the #UnitedCVE hashtag.

Finally, a more direct example of the use of counter-narratives to discredit terrorist propaganda is the Anti-ISIL Global Coalition (funded by the European Union's Strategic Communication Task Force). These initiatives use counter-narratives to degrade, and ultimately defeat, Daesh by targeting romanticised attitudes towards the group with online articles such as ‘The Truth About Life Under Daesh in Raqqa’ (The Global Coalition Against Daesh, January 11, 2018).

How the intervention might work

Although the evidence-base for the development of counter-narratives is limited, there have been several contributions to our general understanding of attitude and behaviour change in respect to violent radicalisation and terrorism. These contributions have come from psychology (Ginges, 2009; Taylor & Horgan, 2006), psychiatry (Bhui, Everitt, & Jones, 2014), sociology (Bouhana&Wikstrom, 2011) and political science (Sageman, 2004).

Psychological research would suggest that narratives can influence both attitudes and behaviour in a number of ways. Theories informed by aspects of communication such as the credibility of the source, message repetition, individual differences in the “receiver” or “target” (Anderson &Hovland, 1957; Hovland, Janis, & Kelley, 1953), the presence of dissonance or ‘mental discomfort’ (Aronson, Turner, &&Carlsmith, 1963; Bochner&Insko, 1966, p. 614; Festinger, 1957) as well as the target's initial, attitudinal position (Sherif&Sherif, 1967; Sherif, Sherif, &Nebergall, 1965) have contributed to our understanding of the persuasive power of narratives in several contexts. However, our understanding of persuasive communication can also inform our attempts at countering it.

In line with discrepancy models, Brock (1967) found that discrepancy causes reduced belief change through the creation of counter-arguments. Discrepancy, Brock argues, causes more elaborate processing through a central (rather than peripheral) route which naturally leads to more counter-arguments and inevitably reduced attitude change. In other words, by encouraging more elaborate processing, the likelihood of considering several sides of a narrative is increased and the rational (Halverson, Corman, & Goodall Jr, 2011), simplistic (Kesterson in Cabayan, V., &Yandura, 2013) and singular narrative approach to persuasion (which has enjoyed immense success in violent radicalisation) is rendered less effective.

Alternatively, the construction of a counter-narrative could be more heavily informed Berlo's aspects of communication whereby differentiations are made between persuasive communication as response shaping, response reinforcing or response changing. A response reinforcing process, whereby the message seeks to reinforce currently held beliefs and make them more resistant to change (Miller &Burgoon, 1973) rather than a response-changing process (whereby the message seeks to actively challenge an existing message) is more likely to be effective (Bettinghaus, 1973; Brembeck& Howell, 1952; Cronkhite, 1969; Dillard & Shen, 2013; Scheidel, 1967 as cited in Dillard and Shen).

Why it is important to do the review

In conjunction with the Partnership for Conflict, Crime & Security Research, Ferguson (2016) published a “horizon scan” (p. 5) of the research landscape in relation to the research question “how can media and communications be used to counter identity-based violence (IBV) or Violent Extremism (VE)?” (p. 2); the report details the state of research in counter-narratives and, although it provides insight into the area, it does not synthesise the experimental evidence on the effects of counter-narrative interventions but, rather, cites government-led and grassroots initiatives akin to those mentioned earlier in this paper. These initiatives, as mentioned, starkly lack an evaluative component, hence, Ferguson's conclusion that there is insufficient evidence to demonstrate the efficacy and effectiveness of counter-narrative strategies for countering violent extremism (CVE). The current authors, as mentioned, do not believe this is the case, thus, prompting the current review.

Briggs and Reves (2013) authored a similar report which offered an overview of different “counter-messaging” initiatives, including counter-narratives. Rather than assessing their effectiveness, Briggs and Reve drew on theory in persuasive communication to highlight the potential of these attempts. For example, message credibility was discussed in relation to message dissemination by former violent extremists, government departments or survivors of violent extremism (pp. 17-18). Again, many of the cited attempts lacked an evaluative component so it is unsurprising that this theoretical approach was used.

In the second half of Ferguson's report, she discusses alternative approaches for which there is a stronger evidence base such as inoculation (the cited study attempted to make participants ‘immune’ to hate speech, for example) and alternative narratives (offering participants an alternative world view, but not directly referencing an established narrative). In relation to these alternative approaches, Ferguson assesses that this literature base is also limited and, within these studies, the measured outcomes do not directly address violence prevention (p. 17), thus excluding many from the current review. Furthermore, by the definition used in this review, many of the cited alternative approaches are not counter-narratives.

For this reason, although these papers will be included in Stage 3 of the search strategy, the scope and rigour of the current review will attempt to expand upon these “horizon scan” syntheses of counter-narrative studies. Acknowledging that it is important to discuss all grassroots, government-led, empirical and non-empirical counter-narrative attempts, this review aims to go further by providing a synthesis of the effectiveness of targeted counter-narrative interventions. In this way, the quality of the evidence, rather than the types of attempts, can be quantified.

In terms of evidence, syntheses have been made of the effectiveness of similar approaches in other fields. For example, (Stice& Shaw, 2004) provided meta-analytic evidence on the use of a proximal approach called dissonance-based interventions (DBI) which encourages individuals to adopt a way of thinking which contradicts their current way of thinking (e.g. challenging social constructions of ‘beauty’ or ‘thinness’). Furthermore, Chan, Jones, Hall Jamieson, and Albarracin (2017) provided meta-analytic evidence on the factors underlying effective counter-arguing or ‘debunking’ of misinformation (for example, conspiracy theories or ‘fake news’). However, to date, there has been no synthesis of the effectiveness of counter-narrative interventions for the prevention of violent radicalisation. This review seeks to address this.

The review will contribute to existing theory and evidence on counter-narrative interventions and allow researchers and practitioners to better understand message style and content, psychological fulcra for change targeted and, importantly, the effectiveness of this approach in reducing outcomes related to violent radicalisation. In doing so, the review may aid those tasked with the design of counter-narratives and, ultimately, help those at risk of violent radicalisation to more critically consider the validity of the messages being communicated through extremist narratives.

Criteria for including and excluding studies

Search strategy

Potentially relevant literature will be identified through a four-stage search strategy, which will comprise:

Stage 1) Targeted keyword searches on a list of relevant databases (Table 2).

In conjunction with a specialist librarian at the National University of Ireland Galway, a comprehensive list of search terms has been developed (see Table 1). The search strategy will be replicated by a co-author upon selection of the studies (controlling for date changes). If there is more than 10% discrepancy between selections, both authors will apply the search strategy together and resolve any differences.

Description of screening process

All identified literature will undergo a three-stage screening process. The titles of all literature will be screened, and papers excluded based, on the exclusion criteria. Studies will be excluded if they do not: 1.

Deliver an intervention which challenges an existing or experimentally introduced narrative with a counter-narrative to reduce propensity towards violent extremism.

During this stage, papers will be rejected if, based on the title, they are clearly not eligible. The abstracts of the remaining literature will then be screened, again with studies excluded according to the exclusion criteria. Finally, the full texts of remaining studies will be screened according to the inclusion criteria, producing the final set of studies to be included in the review.

The above will be performed by the primary researcher (SC). A second screener (CD) will repeat the three-stage screening process. As mentioned, if there is more than 10 percent discrepancy between both screeners’ included studies, both screeners will repeat the third stage of the screening process with both screeners’ final studies (based on the inclusion criteria) before reaching a consensus. See Coding Lists a) and b) (Appendix 2) for title and abstract exclusion criteria and full title inclusion criteria.

Criteria for determination of independent findings

In relation to effectiveness, it is expected that some studies will report impact estimates in relation to various outcomes related to violent radicalisation. In order to ensure that the impact for each intervention type and outcome are derived from statistically independent findings, the following will apply: 1.

In the case of multiple comparable outcomes and effect sizes within the same study (e.g. a study examining the effects of a counter-narrative on explicit attitudes and explicit stereotyping), the outcome which; a) uses the measure with the highest validity as well as; b) is most proximal to the desired outcome (primary or secondary risk factors for violent radicalisation) will be used.

Details of study coding categories

For data extraction, a coding tool has been developed to isolate; a) descriptive information (including intervention design) as well as; b) information that allow the effects of the intervention on the outcome variables of interest to be represented. A full coding scheme is provided as Appendix 3 with a singular example. A risk of bias analysis will be conducted according to the Cochrane Effective Practice and Organisation of Care Review Group (EPOC) Data Collection Checklist. A quality analysis will be conducted according to the GRADE assessment of study limitations.

The primary researcher (SC) will code all of the studies (including risk of bias and quality assessment), and a second coder will double code all studies for risk of bias and quality (CD). If there are more than 10 percent discrepancy in critical fields, both coders will assess the discrepant studies and ensure the differences are resolved.

Statistical procedures and conventions/data synthesis

Although we are limiting our criteria to experimental research (alongside the specified quasi-experimental designs), it may be the case that certain outcomes cannot be synthesised in a meta-analysis, but must be synthesised narratively. In line with the Campbell MECCIR standards, if a quantitative synthesis is not planned, or if it is not possible, authors must plan the specific methods to narratively synthesize the results of the included studies. We intend to: 1.

Present a detailed summary of the results of each study, discussing relevant features of the intervention and research design and presenting effect sizes and 95% confidence intervals;

Assessment of heterogeneity

Heterogeneity will be tested using the I-squared statistic. The tau-squared statistic for random-effects meta-analysis will also be reported.

Assessment of publication bias

Using sub-group analysis of published and unpublished studies, we will examine the possible effects of publication bias. Although it is unlikely we will find 10 or more studies for statistical testing, funnel graphs will be used if possible (Egger, 1997).