Concepts and scope
What the programme studies
What is Algorithmic Immersion Radicalisation?
AIR is a developing framework for examining whether documented digital curation, sustained online immersion, reinforcement and social interaction form part of pathways towards ideological mobilisation and harm. An AIR code records the strength of public evidence for defined digital-pathway indicators; it is not a direct measurement of an algorithm’s effect or a clinical diagnosis.
What is CIVM?
Coercive Ideological Violence Mobilisation is the project’s operational category for qualifying coercive, rights-infringing or violent ideological mobilisation that does not meet the project’s terrorism threshold. Lawful protest, protected expression and mere ideological disagreement do not qualify.
Are CIVM and terrorism the same thing?
No. The project applies separate conduct and purpose rules. A thwarted plot is also separate from an executed terroristic attack. The categories describe inclusion and classification; they are not a severity ladder.
Why does the dataset begin in 2000?
The period provides a pre-social-media comparison window and covers the changing digital environment. Under the current rule, pre-2005 cases are coded Nil AIR because the project does not project the later social-media environment backwards. From 2005 onwards, AIR level depends on documented evidence—not date alone.
Recorded evidence
What the data represent
Does the case database contain every qualifying event?
It is designed as a comprehensive Five Eyes public-source case database within the stated rules, but completeness cannot be guaranteed. Public disclosure, legal definitions, searchability, language, reporting and recency vary. The database is revised when better evidence or omitted cases are identified.
What does “calculation included” mean?
It is the project lead’s authoritative inclusion decision for the reviewed release. The v4.2.3 public denominator contains the 346 records marked “yes”; the single record marked “no” remains in the master workbook for audit lineage and is not published. Inclusion is an analytical dataset decision, not a finding about guilt, dangerousness or population-level risk.
What does missing or “not recorded” mean?
It means the released public evidence did not support a value under the relevant rule. It is not a zero and must not be interpreted as evidence that a characteristic, behaviour or pathway was absent.
Why do some context indicators have gaps or breaks?
Survey waves, censuses, official series and measurement systems have different cadences. Instruments, populations, definitions or source regimes can change. The explorer preserves gaps and documented breaks rather than interpolating a smooth but unsupported trend.
Analytical safeguards
What comparisons do—and do not—show
Does a chart prove that social media caused violence?
No. A visual association, correlation, shared trend or temporal sequence does not establish causation. Country-level context does not measure an individual actor’s exposure, motive or response. Causal claims would require a separate design capable of addressing confounding, selection, measurement and alternative explanations.
Can demographic characteristics be interpreted as motives?
No. Sex, age, religion, citizenship, immigration status and other descriptive fields cannot be treated as motives without explicit case evidence. Aggregate patterns must not be used to profile individuals or communities.
Does the website measure border-security performance or show that migration causes crime?
No. The current public release contains selected migration, asylum, refugee, crime, justice and confidence indicators, but no harmonised Five Eyes border-system performance dataset. National migration aggregates cannot establish an individual case pathway or a migration effect. The Border Security & Public Safety module separates available context, held case evidence and policy interpretation.
Are AIR levels validated risk scores?
No. They are expert-coded evidence levels under a disclosed project rubric. Formal inter-coder reliability, external replication and predictive or causal validation are not yet complete. The levels must not be used for individual risk assessment.
Does “Your Digital Use” assess whether I am healthy, lonely or at risk?
No. It calculates entered time and a social-media share. An optional 2024 GWI value is a population-level contextual reference, not a healthy-use threshold, percentile, diagnosis, AIR level or assessment of mental health, loneliness, radicalisation, violent intent or public-safety risk. Time alone does not reveal content, purpose, platform design, displacement or individual circumstances.
Why are very small groups sometimes suppressed?
Small-cell protection reduces disclosure and over-interpretation risks when sensitive characteristics are filtered. Suppression is an ethical and statistical safeguard; it does not erase the underlying governed record. The current case corpus remains internal and no row-level case record is publicly released.
Does public availability mean every name or identity field is republished?
No. Public availability is a source condition, not a sufficient ethical justification for republication. The programme should withhold or minimise names and sensitive attributes where youth anonymity, victim protection, legal restriction, disproportionality or weak public interest requires it. The public-interest decision, any pseudonymisation and the correction or removal route should be documented. The explorer must never be used for doxxing, harassment or individual risk assessment.
Open research infrastructure
Access, citation and correction
Is the website free?
Yes. The AIR Research Programme website has no subscriptions, advertising, paywall or premium content.
Can I download the data?
Yes. The reviewed Case Database v4.2.3 and its release metadata are available as public CSV files, as are the governed Social Context v1.5 data, indicator registry, source index and release metadata. The case CSV contains only the 22 fields approved for public release; the internal case dictionary, source register and administrative fields are not published.
Are entries in “Your Digital Use” stored or sent to AIR?
No. The form has no submission route, and its calculation code does not store entries, add them to a URL or send them to AIR. Reset clears the visible entries. The site’s separate theme control may retain only the selected light, dark or device preference.
How should I cite the website?
Sundberg, K.W. & Reynaud, A. (2026). Algorithmic Immersion Radicalisation (AIR): Five Eyes Research Dashboard. fiveeyesair.org. For a figure or data extract, also record its figure identifier or dataset version and access date.
How can I report a correction?
The permanent correction and enquiry service remains a launch dependency. Until a tested fiveeyesair.org mailbox or secure form is published, the verified LinkedIn profiles on the Contact page are the interim professional routes. Do not send sensitive personal, operational or emergency information through social media. For a non-sensitive correction, identify the case, indicator, figure or page; the disputed field; the proposed correction; and a direct authoritative source. A submission is evidence for review, not an automatic change.