How Online Radicalisation Works: What the New X Study Reveals About Cambridge Analytica, the IRA, and the Dark Triad Machine
The debate is over. A major new experiment shows how effective Cambridge Analytica and Russian troll farms were in shifting political emotions with hostile, anti-democratic content online
Some more background for the central story of information warfare in Trojan Horses
For nearly a decade, sceptics have dismissed allegations that Steve Bannon’s Cambridge Analytica and Evgeny Prigozhin’s troll farms successfully intervened in the Brexit referendum and Trump’s first presidential election in 2016. But a new study shows how effective their innovations in digital psychological warfare really were .
I. The Experiment That Finally Measured Online Radicalisation
The new X (formerly Twitter) field experiment reported by The Guardian —“Partisan X posts increase political polarisation among users, research finds” (27 Nov 2025) — finally gives us causal proof of online radicalisation:
The study itself, by Piccardi, Saveski et al., was published in Science (2024):
What they did:
Recruited ~1,300 consenting X users.
Used an AI classifier to detect posts expressing anti-democratic attitudes and partisan animosity.
Nudged some users’ feeds to show slightly more of this content, and others slightly less.
What happened:
Users exposed to more hostile content became more polarised within days, measurable on the 0–100 “feeling thermometer”.
Hostility toward the opposite party rose by just over 2 points — the same amount U.S. politics polarised between 1978 and 2020.
Reducing hostile content reversed some of that effect.
For the first time, we have a controlled, causal measurement of the psychological mechanism that CA and the IRA exploited.
II. Cambridge Analytica: Engineering Hostility by Design
Cambridge Analytica’s methods were publicly exposed by Christopher Wylie. His memoir Mindfck* describes exactly how CA built self-contained radicalisation ecosystems using stolen Facebook data:
The British parliamentary inquiry into CA, published by the DCMS Committee, documented the firm’s claims to possess “5,000 data points on every American voter”:
CA’s core techniques mirrored the X experiment:
Fake Facebook groups (“Smith County Patriots”, “I Love My Country”).
Inflammatory content stressing grievances, paranoia, racial tension.
Psychometric profiling based on the Facebook data taken through Aleksandr Kogan’s app.
Targeted A/B testing to find the most emotionally activating messages.
Offline activation — coffee shop meetups designed to turn online grievance into embodied camaraderie.
It was an analogue version of what the X experiment did algorithmically.
III. Aleksandr Kogan’s Dark Triad Pipeline
Dr Aleksandr Kogan, the Cambridge academic who harvested the Facebook data for CA, was simultaneously conducting Russian-funded research at St Petersburg State University on:
dark-triad personality traits (narcissism, Machiavellianism, psychopathy),
online trolling and cyberbullying,
political expression among authoritarian-leaning users.
The Guardian exposed this link in 2018: Kogan lectured in Russia on“New Methods of Communication as an Effective Political Instrument” — effectively, how psychometric profiling could be used as political influence.
His Russian research team even used:
“data of Facebook users from Russia and the USA … by means of a special web-application”.
Inside Cambridge Analytica, SCL’s psychologists then replicated this authoritarian-susceptibility research to identify groups most reactive to hostile messaging.
The X study shows that increasing exposure to hostile, anti-democratic cues produces measurable affective polarisation in days. Kogan’s work shows who is most vulnerable to those cues.
IV. The IRA: AAPA (Anti-Democratic & Partisan Animosity) at Full Scale
Everything CA built manually, the Internet Research Agency (IRA) deployed at industrial scale.
The Mueller Report (Vol. I) lays out how the IRA created entire ecosystems of false US personas:
They included:
Blacktivist
Being Patriotic
Army of Jesus
Heart of Texas
Each page was tailored with the same kind of content the 2024 study tested:
hostility toward political opponents,
anti-democratic rhetoric (“the system is rigged”),
identity conflict,
emotional escalation (fear, rage, betrayal).
And the IRA went further:
Offline activation of rallies and counter-rallies in the U.S. (Houston 2016).
Amplification of GRU hack-and-leak operations, feeding their content pipeline with “evidence”.
The GRU indictments confirmed how stolen DNC/Podesta material was weaponised:
The new X study gives us the missing scientific overlay:
The IRA was not only spreading misinformation — it was administering a continuous AAPA-rich stimulus that we now know reliably increases polarisation.
V. Sam Patten: A Bridge Between CA and the Eastern Networks
Sam Patten — consultant, former Kazakhstan and Ukraine political operative, and longtime associate of Konstantin Kilimnik — joined CA after working for pro-Russian political parties in Ukraine.
The DOJ details his background here:
Under Patten’s watch in 2014, Cambridge Analytica’s Oregon field operation began running focus groups on Vladimir Putin, Crimea, and “strong leader” imagery — a startling early piece of evidence that CA was testing authoritarian cues on American voters.
Wylie’s account confirms these focus groups were real and unauthorised inside the company.
Patten and Kilimnik — the same Kilimnik later identified as a Russian intelligence officer by the Senate Intelligence Committee — had met years earlier in Moscow and worked together frequently.
The Senate Intelligence Committee, Vol. 5, describes Kilimnik as a “Russian intelligence officer” and calls the Manafort–Kilimnik relationship a “grave counterintelligence threat.”
VI. Nix, LUKOIL and SCL’s Covert Memo
Whistleblower Christopher Wylie revealed that CA CEO Alexander Nix personally sent LUKOIL CEO Vagit Alekperov a white paper outlining CA’s U.S. political microtargeting systems. He then told Wylie:
“They understand behavioural micro-targeting in elections … but they are failing to make the connection between voters and their consumers.”
SCL subsequently drafted a 2014 internal memo for Nix describing:
“special intelligence services”,
contracted ex-officers from Russian and other security services,
large-scale fake social media persona networks,
the ability to infiltrate opposition campaigns via “intelligence nets”.
That memo places CA/SCL squarely in the psy-ops / disinformation contracting market — not simply digital marketing.
VII. Security Exposure: Russian IP Addresses and Kogan’s Laptop
The UK Information Commissioner’s Office later concluded:
“Some of the systems linked to the investigation were accessed from IP addresses that resolve to Russia and other areas of the CIS.”
This is not proof of coordinated access.
But it confirms technical exposure.
Combined with the fact that:
Nix told LUKOIL the Facebook dataset was with Kogan in Russia,
Kogan travelled repeatedly to St Petersburg with his laptop,
and Russian-state-linked grants funded some of his research,
then even a simple keylogger on Kogan’s device could have yielded the data CA used to model American voters. No hack of CA required.
VIII. The Scientific Picture: AAPA + Dark Triad = Radicalisation Engine
Put the pieces together:
The X experiment proves that small increases in anti-democratic attitudes and partisan animosity (AAPA) content produce measurable polarisation in days.
Cambridge Analytica built tailored AAPA micro-environments, targeted at high-susceptibility dark-triad and authoritarian profiles.
Kogan’s Russian-funded research showed how to identify precisely those users most reactive to authoritarian cues.
The IRA created a full-spectrum AAPA ecosystem, integrated with GRU disinformation.
Nix’s outreach to LUKOIL and SCL’s internal memo show CA was selling something far beyond data analytics — it was selling a psychological warfare service.
The ICO’s finding shows the technical systems were exposed to Russian IP access.
In short:
We now have the causal science to explain why Cambridge Analytica’s and the IRA’s techniques were so effective — and why the 2016 information environment spiralled into unprecedented levels of hostility and distrust.
The radicalisation pathway is now fully visible:
Behavioural profiling → Hostile-content exposure → Emotional shift → Identity hardening → Offline mobilisation.
IX. Conclusion
The debate around Cambridge Analytica and Russian influence once relied heavily on inference and circumstantial evidence.
But the new X experiment changes the landscape.
It shows that:
hostile content exposure
is not merely correlated with radicalisation —
it causes it.
And CA and the IRA were not just flooding the zone.
They were doing it strategically, using psychometric profiling, dark-triad research, and identity warfare.
For the first time, we can say:
The psychological weapon used in 2016 is no longer mysterious.
We can now measure it — and we know it works.