All research

Polarization, radicalization and violent extremism

What moves people toward political violence — and what holds them back.

Keep scrolling

This strand asks how ordinary grievance hardens into the justification of political violence. The work spans 58 countries, several ideological milieus, and the full arc from the conditions that make violence thinkable to the interventions that make it less so.

Violence is not one thing

A preregistered study across 58 countries separates offensive from defensive intergroup violence and finds they do not share a psychology. Support for striking first and support for striking back rest on different motives, which matters because prevention policy has largely been built as though a single 'propensity for violence' existed. The same distinction reappears in work measuring extremist archetypes, where people drawn to violence sort into recognisably different types rather than along one continuum.

Threat, deprivation, and the need for certainty

Across studies, perceived cultural threat is among the most reliable predictors of support for violence — but it works through mechanisms that can be specified. Cultural threat predicts violent extremism via the need for cognitive closure: the discomfort of ambiguity, not the ideology itself, does much of the work. Relative deprivation has re-emerged as central, and narratives like the 'great replacement' show how a perceived loss of group status converts into extreme anti-immigrant sentiment.

Fusion: when the group becomes the self

A meta-analysis of identity fusion establishes how strongly a visceral sense of oneness with a group predicts extreme pro-group action. Related work shows fusion with a political leader predicts willingness to persecute immigrants and political opponents, and that fusion can also form with groups one does not belong to — driving extreme activism on behalf of others' struggles.

The new problem: AI in the pathway

Recent work sets out a framework for radicalization in the age of AI, arguing that recommender systems and conversational agents change not just the volume of extremist content but the shape of the pathway into it. This connects directly to the misinformation strand, where the same infrastructure is examined from the supply side.

What follows for prevention

If entry is social, exit is too. Work on resilience against radicalization in schools develops measurement that treats resistance as something buildable rather than merely absent, while research on lay perceptions of terrorism shows how public beliefs about what causes it shape which counter-terrorism policies people will accept. A machine-learning analysis of more than 34,000 articles maps where the field has concentrated — and what it has left alone.

  • 26Papers in this strand
  • 15Journals & publishers
  • 2018–Active since

Papers in this strand

Also relevant to

All publications

Related press coverage

Press Coverage

FAQ

Frequently asked questions

Who believes what, why — and what artificial intelligence changes.

Next strand

Next strand

Misinformation, conspiracy beliefs, and AI

Read more →