- Language, identity, and threat framing are not by-products of violence: they are its early indicators.
- Machine-learning analysis of ISIS's al-Naba' newsletter demonstrated forecast capability of up to six months on terrorist activity.
- The result was achieved with minimal resources and no institutional backing: the method, not the budget, was decisive.
- The same methodology generalises: from terrorism to conflict prevention, atrocity early warning, political risk, and legitimacy analysis.
A counter-terrorism question turned inside out
The research that became BrainBridge began with an unusual orientation. Most counter-terrorism analysis asks: what have the attackers done, and what might they do next? The doctoral project at the University of Bath asked a different question: what does the violence look like from the perspective of the communities it targets, and could their protection be the organising goal of analysis?
That minority-centred framing led somewhere unexpected. Studying terrorist and extremist rhetoric (how groups talk about themselves, their enemies, and their justifications), it became increasingly clear that rhetoric is not commentary on violence. It is part of the production process of violence. Groups form through language. Enemies are constructed through language. Violence becomes justified through language, long before it is enacted.
If that is true, then a group's own media is not propaganda noise to be dismissed. It is a leading indicator.
Why al-Naba' was the right corpus
Testing the hypothesis required a body of extremist text that was large, sustained, and internally consistent. ISIS's official weekly newsletter, al-Naba', fit precisely: a dated, regular publication in the organisation's own voice, spanning years of operational highs and lows. A corpus like that allows longitudinal analysis: tracking how framing, threat construction, and mobilisation language change over time, and testing whether those changes precede changes in operational behaviour.
Midway through the doctorate, the theoretical insight was paired with technical capability: an AI-powered methodology, developed as part of the PhD itself, applying machine-learning methods to the corpus at a scale qualitative reading cannot reach, with the explicit goal of predicting ISIS attacks from the organisation's own rhetoric.
The finding, and why it mattered
The results were decisive: modelling rhetorical patterns and narrative shifts made it possible to forecast terrorist activity up to six months in advance.
Three aspects of the finding deserve emphasis:
- The signal was in the words, not the metadata. The forecast power came from rhetorical and narrative dynamics: the kind of "soft" data conventional risk models exclude as unmeasurable.
- The horizon was strategic, not tactical. Six months is enough time for prevention to be more than evacuation. It is resource allocation, diplomatic engagement, protective deployment: upstream options that disappear once violence begins.
- It was done with minimal resources and no institutional backing. A doctoral project, demonstrating predictive power that institutional early-warning systems lacked. The method was the breakthrough, not the budget.
The research behind this work has since been published in a number of the world's leading scientific journals: peer-reviewed validation of the methodology that became BrainBridge's foundation.
From a thesis result to a company
After the PhD, the implications were impossible to ignore. If rhetoric forecasts terrorism, the same logic should extend to every domain where escalation is narrated before it is enacted: communal violence, atrocity risk, political instability, legitimacy collapse, even market-relevant dynamics like boycott campaigns and regulatory backlash.
The environment had also changed. Social media had turned rhetoric, identity formation, and grievance mobilisation into observable data streams at unprecedented scale, and AI systems had matured enough to interpret them with depth. The convergence is recent: a decade ago, neither the open data nor the analytical capability existed in usable form.
BrainBridge was founded to institutionalise the methodology: to treat rhetoric and meaning as measurable, predictive forces, and to direct AI and big data toward understanding risk early and protecting vulnerable groups. The framework that emerged from this work, the Third Intelligence, formalises the lesson the thesis taught in practice: the forecast came neither from the expert nor the model, but from the deliberate bridge between them.
The most complete public demonstration since is the Syrian coastal massacres analysis, the same intellectual DNA, applied to a live conflict at ecosystem scale: 100,000 posts, a 3,087-node knowledge graph, and decision-ready intelligence in ten days.
The lesson for early warning
The al-Naba' research is one data point in a larger argument that runs through everything BrainBridge does: the world's most dangerous dynamics announce themselves in language first. Institutions that treat narrative as noise will keep being surprised by crises that were, in fact, well advertised. Institutions that learn to read trajectory (the slow hardening of grievance into justification into dehumanisation) buy themselves the one resource no crisis response can recover: time.
Frequently asked questions
Does forecasting from rhetoric risk criminalising speech?
No. The methodology analyses public organisational media at the level of patterns and trajectories to anticipate harm against vulnerable groups. BrainBridge's values are explicit on this point: early warning, not surveillance; decision support, not decision replacement; human accountability central throughout.
Would this work on groups other than ISIS?
The underlying mechanism, narrative preparation preceding action, is not ISIS-specific. Any actor that must mobilise an audience leaves rhetorical traces. The corpus and models differ per context; the methodological logic transfers, as the Syria analysis demonstrated in a multi-actor conflict ecosystem.
Where can I learn more about the methodology?
Start with Measuring Information Warfare for the knowledge-graph approach, or contact us to discuss research collaboration and access to the analytical infrastructure.
