This project is dedicated to the men and women who lost their lives on September 11, 2001, and the many more whose lives were forever changed — including all those who work tirelessly to keep the public safe and secure. In particular, we acknowledge the incredible work of Maureen Basnicki and the Canadian Coalition Against Terror (C-CAT), which she founded following the loss of her husband on 9/11.
Social media has fundamentally reshaped how people across the Five Eyes nations — Canada, the United States, the United Kingdom, Australia, and New Zealand — form political identities, interpret current events, and coordinate collective action. At the same time, criminal justice, public safety, and national security practitioners have observed that the pathways through which individuals move from grievance to ideologically motivated violence are changing — and accelerating. This project was developed to better understand that shift.
This project has its roots in Alannah Reynaud's honours thesis, which examined the phenomenon of incel-motivated violence — cases that are included in this dataset. Alannah has since gone on to pursue her doctorate, and the questions raised by that early work helped shape the broader research programme that became AIR.
Sparked by mounting reports on how social media is negatively impacting both individuals and society, and noting the dearth of criminological and criminal justice research examining the intersection of platform architecture and ideological violence, Kelly and Alannah introduced the AIR framework at the International Association for Intelligence Education (IAFIE) 2025 annual conference in Aranjuez-Madrid, Spain.
Algorithmic Immersion Radicalisation (AIR) is a socio-technical framework that describes how engagement-driven social media platforms can intensify pathways toward ideological violence through a four-stage cycle — curation, immersion, reinforcement, and mobilisation — as illustrated in Figure 1 below. Critically, AIR is driven by platform architecture — not by any specific ideology. The same algorithmic mechanisms operate across jihadist, far-right, far-left, incel, conspiratorial, and other extremist ecosystems. The process is cyclical — it can stop at any stage, or recur at increasing intensity.
Figure 1. The Algorithmic Immersion Radicalisation (AIR) Cycle. Mobilisation outcomes range from CIVM to violent extremism and terrorism.
We plan to submit a number of studies drawing on this dataset and the AIR framework. We are committed to keeping this dataset as current and accurate as possible and welcome feedback — including identification of any cases we may have missed. The References tab provides links to publicly available scholarship underpinning this work, and we will do our best to keep that reference list current as new research emerges.
This project has benefited greatly from Kelly's ongoing collaboration with colleagues from the International Association for Intelligence Education (IAFIE) and the Australian Institute of Professional Intelligence Officers (AIPIO) — both organisations host incredible conferences that bring together amazingly insightful and experienced professionals and academics whose perspectives continue to inform this work.
Special thanks to Keith Cozine and the faculty and staff in the Homeland Security Studies program at St. John's University; Emmanuel Brunet-Jailly and his team at the Borders in Globalization (BIG) Lab at the University of Victoria; Svetlana Yanushkevich and her team at the Biometric Technologies Laboratory at the University of Calgary; and the faculty and staff in the Bachelor of Criminal Justice program at Mount Royal University. We also gratefully acknowledge the amazing and brilliant people at the SAFE Design Council, whose work has noticeably advanced public safety and security through the mindful application of policy, practice, technology, and design. We are also grateful for the ongoing support of Paul Babie and his colleagues at Adelaide University, and John Coxhead, Ruwan Uduwerage-Perera, and Ahmad Al-Hiari at the Solution Oriented Policing Group at De Montfort University.
This interactive research dashboard accompanies an original Five Eyes dataset covering the period from September 2001 to March 2026. The dataset is organised into three analytically distinct tiers: executed terrorist attacks, thwarted (foiled) terrorist plots, and Coercive Ideological Violence Mobilisation (CIVM) incidents. CIVM is a research category introduced in this study to capture ideologically driven collective violence, coercion, intimidation, and disorder that crosses beyond lawful protest but does not meet the conventional threshold for terrorism.
Each incident in the dataset has been coded across multiple dimensions — including target location, target group, harm vector, offender ideology, immigration status, religious affiliation, and an AIR causation assessment — using publicly available court records, official reports, and verified journalism. The individual data tables also include links to relevant news articles for each case. The dashboard is designed to allow police, security professionals, intelligence practitioners, and researchers to filter, explore, and analyse these cases interactively.
This dashboard presents a Five Eyes dataset covering Canada, the United States, the United Kingdom, Australia, and New Zealand from September 2001 to March 2026. The dataset is organised into three analytically distinct tiers: executed terrorist attacks, thwarted terrorist plots, and coercive ideological violence mobilisation (CIVM).
The underlying dataset was assembled through structured open-source collection and verification using public records, court material, police and government statements, major news reporting, and other credible archival sources. The researchers made every reasonable effort to improve completeness, accuracy, consistency, and analytical rigour across all three tiers. Even so, this dashboard is not intended to function as an exhaustive census of every possible incident. It is best understood as a carefully constructed research dataset based on the public record available at the time of compilation.
The terrorism tier is the strongest from a data-quality standpoint because completed attacks are typically widely reported and often generate detailed public records. The thwarted-plots tier is necessarily less complete because disrupted investigations are often only partially disclosed, sealed, or reported unevenly across jurisdictions. The CIVM tier is an analytical category rather than a legal one and therefore involves more interpretive judgement than the terrorism tier.
CIVM is used here to capture conduct that crosses from lawful expression into coercion, intimidation, rights infringement, occupation, targeted public disorder, or violence in service of an ideological cause without necessarily meeting the legal threshold for terrorism. In other words, the term is intended to capture situations in which ideology mobilises violence or coercive collective action. Peaceful protest, lawful assembly, and non-violent political expression are not intended to be included.
The AIR causation field reflects an analytic assessment rather than a claim of mechanistic proof. Coding was designed conservatively. In particular, early-era incidents were not treated as though mature recommender-driven platform ecosystems existed before they did. As a result, cases from the pre-2009 period, and especially the 2001–2006 period, should be interpreted with caution and are less likely to attract AIR coding unless there is a clear and defensible basis for doing so.
This dashboard relies on publicly available information, and public reporting can be incomplete, inconsistent, delayed, or later corrected. Some incident attributes — including motive, ideological affiliation, immigration status, religious affiliation, target classification, and AIR causation — may evolve as investigations proceed or as better source material becomes available. Foiled plots almost certainly undercount the true number of disrupted cases, and CIVM remains a provisional analytical category that should continue to be refined through future scholarship.
This dashboard is provided for research, educational, and informational purposes only. It is supplied on an "as is" basis without any representation, warranty, or guarantee of completeness, accuracy, fitness for purpose, or ongoing currency. Users remain responsible for independently verifying any information before relying on it in research, policy, legal, operational, journalistic, or other consequential settings. To the fullest extent permitted by applicable law, the authors and copyright holder disclaim liability for any loss, damage, error, omission, or consequence arising from the use of, or reliance upon, this dashboard or its contents.