Research and analysis

The role of social media in the motivation and capability of violence-fixated individuals

Published 24 September 2026

This is not a statement of government policy.

Rapid projects support government departments to understand the scientific evidence underpinning a policy issue or area by convening academic, industry and government experts at a single roundtable. These summary meeting notes seek to provide accessible science advice for policymakers. They represent the combined views of roundtable participants at the time of the discussion and are not statements of government policy.

This publication considers the following question, taken from a meeting note of a roundtable chaired by Professor Dame Angela McLean (Government Chief Scientific Adviser) facilitated by the Government Office for Science (GO-Science) on 22 May 2025.

What evidence is there on the role of social media activity and content in the motivation and capability of Violence-Fixated Individuals (VFIs) to commit violent acts?

In current His Majesty’s Government (HMG) work, violence-fixated individuals (VFIs) are defined as: individuals presenting a fascination with extreme violence and the intent to inflict extreme violence, where ideology is not the primary driver.

For this meeting, extreme violent content (EVC) is defined as content that: depicts the apparently real, deliberate infliction of death, severe physical pain or suffering on an animal or human.

Summary of roundtable findings

There is a huge volume of EVC online. Consumption of EVC has become mainstream. However, evidence for the impacts of exposure to online violence is sparse and of low quality.

Known VFIs comprise a very small cohort, making it unfeasible to generate statistically meaningful conclusions from their actions. While some associative links have been identified through case study analysis, the rarity of violent acts again makes statistically meaningful conclusions difficult.  

It is highly challenging to design an ethical and statistically valid study that could establish causation and effect sizes between online and offline behaviours for any kind of violent outcome.

Research that HMG might commission in this space will depend on which harms it wants to reduce and why.

Evidence for the role of social media activity in VFIs’ motivations and capabilities to commit violent acts

1. Social media enables individuals to pursue interests in EVC cheaply, at scale, anonymously and without shame. It can fast-track individuals to extremely violent stimuli compared to the period before social media existed, including through new vehicles such as the creation and sharing of memes and applying elements of game play to acts of violence (Lakhani and Wiedlitzka, 2022; Lakhani and others, 2022). 

2. However, data on online and social media activities provide only limited insights into individuals’ lives (Smith and others, 2020). Establishing connections between online behaviours and real-world actions presents a very substantial challenge, even when an individual’s entire online behaviour can be accessed (Brown and others, 2024; Smith and others, 2020).

3. Repeated exposure to violent content can lead to desensitisation, where users seek more extreme material. However, for many individuals, it is pre-existing susceptibility to desensitisation, habituation and repetitive pursuit of rewarding stimuli that leads to them seeking repeated exposure to and escalation of violent content. 

4. There is more robust evidence for desensitisation to violence from exposure to media such as television and video games (such as Greitemeyer and Mügge, 2014) than for social media activity. Desensitisation, though, has not been established as a causal factor in violence.

5. Evidence for the impacts of exposure to violence online features fewer meta-analyses. The evidence is generally of low quality. Most studies focus on “legal but harmful” content.  

6. One recent review (Wolfowitz and others, 2022) looked at 23 social media activities as risk factors for radicalisation, exploring cognitive (motivational) and behavioural (capability) impacts. It found that exposure to radical content online appears to have a larger relationship with radicalisation than other media-related risk factors. However, the quality of the evidence is low and more robust study designs are needed to draw firm conclusions.

7. The algorithms behind mainstream online platforms can promote content which incites emotive responses. However, a recent meta-review demonstrates that it is mostly so-called “legal but harmful” content which can be amplified, rather than illegal material (Whittaker and others, 2025), including legal but misogynistic content promoted to male youths (Baker, Ging and Brandt Andreasen, 2024; Regehr and others, 2024).   

8. Some types of EVC (especially terrorism content) are typically less accessible on mainstream online platforms now than in previous years, but others are easily available (such as footage from war zones, Fisher and Bradley, 2025).

9. EVC is also available on alternative platforms or niche forums, requiring users to either actively seek out such content or follow website links linking to it from these alternative platforms (such as on pornography platforms).

10. Individuals with neurodivergent traits such as Autism Spectrum Disorder and Attention Deficit Hyperactivity Disorder (ADHD) may be predisposed to habituation and to seeking increasing levels of stimulation, including from violent content (Al-Attar, 2020; Worthington and others, 2022; Al-Attar and Worthington, 2024). Neurodivergence interacts with many other individual bio-psycho-social predispositions and situational factors to shape how an individual chooses, interacts with and is impacted by harmful online content.

11. Evidence of individuals engaging with social media content concerning how to commit violence (in other words, practical aspects) is more predictive of those individuals proceeding to act violently than engagement with ideological or hate speech content (Brown and others, 2024). This evidence comes from research into drivers of radicalisation.

12. While only a small number of consumers of EVC are likely to carry out acts like the Southport murders, there is a vast amount of EVC online and consumption has become mainstream (Hardy and Stewart, 2023; Fisher and Bradley, 2025). For example, one prominent, non-ideological EVC website has millions of members, 8,000 active videos, a system for sending ‘kudos’ to active members and ranking systems, effectively gamifying engagement with EVC (BBC News, 2025; see also Lakhani and Wiedlitzka, 2022; Lakhani and others., 2022). Traffic to ‘gore’ content has increased considerably since the October 7 attacks in Israel.

13. Ease of access to violent content has lowered barriers to youth engagement (Ybarra and others, 2022), and we may be seeing a faster escalation of interest and other behaviours than in the past. 

14. Violent content is often embedded in subcultures (such as ’com networks’, incels, gore, and far-right gaming communities, Smith and Smith, 2025), where it is shared for status or humour, for example, and does not always reflect ideological commitment (Lee, 2022; Lee, 2023).

The challenges of linking exposure to EVC online and intent to commit violent acts

15. There are multiple types of violent acts, such as, violence against women and girls, animal cruelty, group violence. It is challenging to design a study that could establish causation between online and offline behaviours for any kind of violent outcome. This is because:

• establishing the strength of relationship between exposure to violent content and acts of violence would require the predictor and outcome to be specified precisely and would need sufficient data
• online activities represent a fraction of an individual’s lived experience and offer limited insight into one’s psychology and behaviour
• it is difficult and costly to gather data on both online and offline behaviours. Building a robust evidence base would require collaboration from multiple stakeholders, such as the police, schools, tech platforms (Smith and others, 2020)
• longitudinal, cross-lagged data that included multiple online and offline factors against a particular outcome would be required (Smith and others, 2020), but it would be difficult to obtain a meaningful sample
• experimental conditions that could test causal mechanisms would be unethical in the context of violent acts

16. Research that HMG might commission in this space will depend on which harms it wants to reduce and why.

17. Known VFIs comprise a small cohort, making it unfeasible to generate statistically meaningful conclusions from their actions, even if extensive data were gathered on those individuals.

18. While some associative links have been found through case study analysis, the rarity of violent acts makes statistically meaningful conclusions difficult here too. Most people exposed to EVC do not commit violence. Normalisation of exposure to EVC significantly reduces the predictive value of tracking who consumes EVC. 

19. One systematic review (Purba and others, 2023) found associative evidence linking social media use with increased anti-social and other risk-taking behaviours in adolescents, but the underlying evidence was largely correlational and often methodologically weak.

20. Research on exposure to violent media more broadly suggests that any effects are unlikely to be direct. Instead, they are likely mediated or moderated by person-based factors (such as personality traits or mental health status) and environmental variables (such as family environment or peer influences). To address these complex pathways, causal epidemiology approaches are needed, which can better explain how and when social media exposure may contribute to violent behaviours through indirect or interacting pathways, rather than a simplistic direct effect.

21. Qualitative and emerging digital trace data indicate some recognisable patterns, such as the active dissemination or glorification of violent content by VFIs (Brown and others, 2024). However, there is no clear causal evidence that increased engagement with or escalation in consuming violent content directly leads to violent acts, despite some anecdotal and case-study reports suggesting a temporal association shortly before such acts.

22. Evidence is needed on the speed of pathways to violence, from developing a fascination through to action. Establishing causality would require longitudinal studies that track individuals’ exposure to violent content and subsequent behaviour over time.  

23. Individuals who commit violent acts often show cognitive fixation on violence which need not be driven by ideology (Gartenstein-Ross and others, 2023). Their capability is enhanced by online resources, but their motivation is often intrinsic and longstanding. There is a task for researchers to better understand these rare individuals who develop fascinations with specific previous attackers, potential victims and methods of attack. 

24. Literature on terrorist actors focuses more on their fixation towards particular individuals than their fixation on violence in itself.

Meeting participants

These participants attended the meeting:

• Professor Dame Angela McLean (GCSA, chair)
• Joe Whittaker (Swansea University)
• Kaitlyn Regehr (UCL)
• Laura Smith (University of Bath)
• Lewys Brace (University of Exeter)
• Maura Conway (Dublin City University)
• Olivia Brown (University of Bath)
• Suraj Lakhani (Sussex University)
• Amrit-Kaur Purba (London School of Hygiene and Tropical Medicine, formerly University of Cambridge – contributions in writing)

References

Al-Attar Z. (2020). Autism spectrum disorders and terrorism: how different features of autism can contextualise vulnerability and resilience.

Al-Attar Z and Worthington R (2024). “Online terrorism offenders with ADHD: in what ways can ADHD create contextual vulnerabilities and risk?” Chapter in: S Macdonald (Editor). Routledge Handbook of Disability, Crime, and Justice.

Baker C, Ging D, and Brandt Andreasen M (2024). Recommending toxicity: the role of algorithmic recommender functions on YouTube shorts and TikTok in promoting male supremacist influencers.

BBC News (2025, 17 March) Shut vile death video site, families say, as Ofcom gets new powers.

Brown O, Smith LGE, Davidson BI, Racek D and Joinson A. (2024). Online signals of extremist mobilization.

Fisher, A and Bradley, A (2025). Gore and violent extremism: an explorative analysis of the use of gore websites for hosting and sharing extremist and terrorist content.

Gartenstein-Ross D, Zammit A, Chace-Donahue E and Urban M (2023). Composite violent extremism: conceptualizing attackers who increasingly challenge traditional categories of terrorism.

Greitemeyer T and Mügge DO (2014). Video games do affect social outcomes: a meta-analytic review of the effects of violent and prosocial video game play.

Hardy J and Stewart C (2023) Gore and violent extremism: how extremist groups exploit ‘gore’ sites to view and share terrorist material.

Lakhani S, White J and Wallner C (2022). The gamification of (violent) extremism: an exploration of emerging trends, future threat scenarios, and potential P/CVE solutions.

Lakhani S and Wiedlitzka S (2022). “Press F to pay respects”: an empirical exploration of the mechanics of gamification in relation to the Christchurch attack.

Lee B (2022) “Lethal subcultures”, Centre for Research and Evidence on Security Threats (CREST).

Lee B (2023) Risky business? Siege, subcultures, risk, and unintended consequences, CREST.

Purba AK, Thomson RM, and others (2023). Social media use and adolescent health risk behaviours: systematic review and meta-analysis.

Regehr K, Shaughnessy MZ, Shaughnessy C (2024). Safer scrolling: how algorithms popularise and gamify online hate and misogyny for young people.

Smith LGE, Wakeford L, Cribbin TF, Barnett J and Hou WK (2020). Detecting psychological change through mobilizing interactions and changes in extremist linguistic style.

Smith W and Smith LGE (2025). Youth radicalisation and harm in online networks.

Whittaker J, Rogers E, Micallef N and Correia S (2025). What are the links between social media algorithms, generative AI and the spread of harmful content online?

Wolfowicz M, Hasisi B and Weisburd D (2022). What are the effects of different elements of media on radicalization outcomes?

Worthington R, Al-Attar Z, Lewis A and Psyzora N (2022). Rapid evidence assessment on neurodiversity and violent extremism.

Ybarra M, Mitchell K and Oppenheim J (2022). Violent media in childhood and seriously violent behaviour in adolescence and young adulthood.