The Five Eyes Algorithmic Immersion Radicalisation (AIR) Dashboard is an exploratory research dashboard for examining publicly documented cases of ideologically motivated violence, thwarted plots, and coercive ideological mobilisation across Australia, Canada, New Zealand, the United Kingdom, and the United States.
The dashboard combines three elements: a structured AIR case dataset, a national-level context dataset, and interactive comparison tools. The case layer remains the core research focus. The context layer helps users compare public narratives against evidence on digital exposure, crime and public safety, health, demography, housing, economic stress, institutional conditions, and social change.
The site is designed for researchers, students, public safety professionals, journalists, and informed public users who want a transparent way to examine patterns without collapsing complex events into simple causal claims.
Algorithmic Immersion Radicalisation (AIR) describes how engagement-driven digital platforms can intensify pathways toward ideological rigidity, coercion, or violence through a cycle of curation, immersion, reinforcement, and mobilisation.
AIR is not a simple “rabbit-hole” theory. It is not a claim that algorithms mechanically produce terrorism. It is a socio-technical accelerant framework: it helps examine how platform architecture, identity, grievance, peer validation, influencer ecosystems, offline conditions, and political shocks may interact in ways that compress escalation timelines or intensify mobilisation. AIR is trans-ideological; the mechanism may appear in different ideological environments even where the content differs substantially.
The current working dashboard dataset contains publicly documented cases across the Five Eyes countries. It includes terroristic attacks, thwarted or foiled terroristic plots, and CIVM incidents. Inclusion is based on publicly available documentation, and case coding is intentionally conservative.
AIR levels reflect the degree to which public sources document online immersion or digital radicalisation indicators. They should not be read as proof that a platform caused a case.
Coercive Ideological Violence and Mobilisation (CIVM) is a working analytical category used in this project to capture conduct that appears ideologically motivated and coercive, intimidating, rights-infringing, violent, or mobilisation-oriented, but may not meet legal terrorism thresholds.
CIVM is not a criminal offence and is not a legal classification. It is used for analytical comparison. Lawful protest, dissent, and democratic participation are not CIVM unless coercive, intimidating, violent, or rights-infringing conduct is present.
The dashboard includes national-level context indicators that support comparison across digital exposure, crime and public safety, health and wellbeing, demography, economic and housing stress, institutions, and social conditions. Context data are descriptive, not causal. Some indicators are trend-ready, while others are benchmark, sparse, country-specific, subnational, or limited-coverage indicators.
Context matters because it helps users test public narratives against evidence, distinguish AIR case patterns from broader social conditions, and avoid simplistic claims about social media, crime, extremism, or social decline.
Z-score (standardised): Each displayed series is standardised using its own valid values, so values show how far that series is above or below its own average. This supports pattern comparison but does not show original units.
Actual values: show the original measurement units where direct interpretation matters. Indexed comparison: shows relative change from the first valid value. Benchmark cards: are used where an indicator has one source year or limited coverage. Country-specific and subnational cards: are used where values should not be treated as Five Eyes-comparable national measures.
Missing values are not treated as zero. The dashboard blocks aggregation where the context metadata does not permit aggregation, and it avoids continuous trend lines for indicators that are sparse, benchmark-only, subnational, or not comparable.
The timeline below is computed from the current AIR case dataset and is descriptive only. It does not imply a causal relationship between digital exposure and case occurrence.
Conceptual foundation: Algorithmic Immersion Radicalisation (AIR), developed by Kelly W. Sundberg, Alannah, and collaborators as a socio-technical framework for understanding ideological escalation in Five Eyes security environments. No public manuscript URL is added here unless it is hosted in the site package or otherwise authorised.
This dashboard draws on a structured Five Eyes AIR case dataset and a national-level context layer. It is designed for disciplined description, comparison, and interpretation of publicly documented cases, not prediction or causal proof. The current working dataset contains 224 publicly documented cases: 93 Terroristic Attacks, 52 Thwarted / Foiled Terroristic Plots, and 79 Coercive Ideological Violence and Mobilisation (CIVM) incidents.
The AIR case layer uses publicly available reporting, official records, court documents, government sources, credible investigative material, and related open-source documentation where available. Public documentation varies by country, case type, and time period. Absence of evidence in public sources is not evidence that a factor was absent. The dataset does not include non-public intelligence, sealed investigations, unreported cases, or cases where the public record is insufficient for conservative coding.
The incident dataset covers Canada, the United States, the United Kingdom, Australia, and New Zealand. It distinguishes among Terroristic Attacks, Thwarted / Foiled Terroristic Plots, and CIVM incidents. The term “Terroristic Attacks” is used descriptively because Five Eyes countries do not use identical legal definitions or recording practices.
Calculated from data.json at runtime so the methodology summary remains aligned with the working dataset.
Coercive Ideological Violence and Mobilisation (CIVM) is used here as an analytical category, not a legal classification. It helps separate lawful protest and democratic engagement from incidents where ideological mobilisation appears to involve coercion, intimidation, violence, rights-infringing conduct, or mobilisation-oriented disruption.
CIVM is not a criminal offence, is not a statutory category, and is not a recommendation for policing or prosecution classification. Lawful protest, dissent, and democratic participation are not CIVM unless coercive, intimidating, violent, rights-infringing, or mobilisation-oriented conduct is present. CIVM cases also vary substantially in severity, mechanism, and legal character.
Each case is assessed for the apparent strength of publicly documented evidence connecting it to Algorithmic Immersion Radicalisation. AIR levels reflect how strongly public sources document digital immersion, online radicalisation, platform exposure, online community reinforcement, or related indicators. Higher ratings require stronger public-record support; lower or nil ratings reflect weaker, absent, or indeterminate public evidence.
AIR coding does not prove that algorithmic curation caused a case. AIR is treated as a socio-technical accelerant framework, not a monocausal explanation. Offline networks, personal grievance, ideology, organisational influence, local conditions, mental health, opportunity, and broader social context may also matter. Where public sources do not document online immersion, that absence should not be read as proof that online influence was absent.
The dashboard also integrates the Five Eyes AIR Context Master Version 3.5.0. The current context layer contains 135 active indicators, 7,370 context rows, and 7,239 non-null values. These indicators support descriptive comparison across digital exposure, crime and public safety, health and wellbeing, demography, housing, economic stress, institutions, and social conditions.
Context indicators are not offender-level measures and cannot, on their own, explain individual cases. They vary in comparability, country coverage, year coverage, population scope, source definition, and quality. Some are trend-ready; others are benchmark, sparse, partial-coverage, country-specific, subnational, or limited-coverage indicators. A value-bearing indicator may therefore appear as a line, card, dot plot, matrix row, country-specific note, or caveated context item depending on coverage and suitability.
The dashboard uses a display-governance model so that indicators are visualised according to their structure and comparability. Trend-ready indicators may appear as line charts. Benchmark or source-year indicators may appear as cards or dot plots. Country-specific or subnational indicators appear only in country-specific contexts. Indicators unsuitable for relationship analysis are not used for correlation or relationship outputs.
Five Eyes averages or aggregates are shown only where the metadata and data structure permit them. Aggregation is blocked when indicators are not comparable, when metadata prohibits aggregation, or when coverage is too limited for responsible summary. Partial coverage is caveated, and country lines may be shown where aggregate values would be inappropriate.
The Online Time and Social Context Explorer is exploratory. It compares daily social media use, daily internet use, AIR case patterns, and selected national context indicators in order to support descriptive comparison, not causal inference. Each displayed series is standardised using its own valid values, so values show how far that series is above or below its own average. This supports pattern comparison but does not show original units.
Actual values and Indexed comparison remain available where appropriate. Relationship outputs use only valid overlapping country-year observations and are blocked where data are insufficient, benchmark-only, country-specific, subnational, or otherwise unsuitable. Descriptive only. Not causal. Correlation, visual alignment, similarity, divergence, or co-movement does not establish causation.
Missing values are not converted to zero. Rows marked as unavailable, scaffold, placeholder, no-data, missing, or source-gap are excluded from visualisation. Invalid zero placeholders are not plotted, while true source-backed zero values may be preserved where metadata supports them. These safeguards are intended to prevent artificial final-year drops caused by missing or provisional data.
Recent years require caution. Case records from 2025 or 2026 may be valid where documented, but context indicators may not yet have comparable source coverage. Users should read partial years, provisional source periods, and sparse final-year observations carefully.
These caveats guide the dashboard's interpretation of indicators. They explain why some value-bearing measures appear as lines, cards, dot plots, matrix rows, or country-specific notes rather than as ordinary trend charts.
The dashboard does not claim that social media use causes radicalisation, that online time causes violence, that AIR explains every case, or that context indicators explain individual cases. It also does not claim to capture every relevant case, fully harmonise every Five Eyes comparison, or generalise country-specific and subnational indicators across all countries.
The strongest findings supported by the dashboard are descriptive: documented case patterns, distributions, timelines, category comparisons, and carefully caveated contextual comparisons. Its purpose is to support evidence-informed interpretation, teaching, and research discussion while keeping causal claims outside what the data can responsibly support.