Charismatic Influencers, Certainty, and their effect on Polarisation Online

Charismatic influencers are particularly effective at polarisation through their use of narratives designed to appeal to a broad audience. This is typically achieved by projecting certainty and simplifying issues in social domains where uncertainty is unavoidable. Increased online activity further accelerates this process by closing down debate and repelling those with moderate views.

Charismatic Influencers and Polarisation Online

Online polarisation can be accelerated by charismatic political influencers, that is, influencers typified by a broad social reach and active engagement online. Using an agent-based model (Betts et al., 2025), we show why and how messaging with narratives that reward certainty over complexity, and thus having a broad reach, amplifies the effect of an extremist influencer on social polarisation, creating new challenges for digital security. 
Although online polarisation is increasingly recognised as a concern by governments, mitigation strategies and law enforcement have often focussed on harmful content rather than on the structural dynamics of the interaction between influencer and their audience. This approach tends to overlook the online dynamics driving both the spread of division and the extremisation of beliefs and attitudes. In these cases, influencers matter not simply because of what they say, but because they can rapidly reshape opinion dynamics and push communities toward opposing camps faster than institutional responses can adapt (Bliuc et al., 2024). From a security perspective, this means that polarisation risk is not only about extreme views being shared online, but about how quickly and widely social divisions become solidified. Therefore, a systematic, evidence-based understanding of the role of influencers is essential if we want to be able to predict the way issues could divide societal beliefs and opinions before that happens (rather than reacting after social fragmentation has occurred).

Polarisation risk is not only about extreme views being shared online, but about how quickly and widely social divisions become solidified.

Using an agent-based model simulating an artificial society with members holding opposing views on a given issue, our research shows that polarisation emerges even in the absence of influencers, driven by everyday, social interaction and opinion adjustment. This process can often be observed in the real world or online when people’s beliefs may start off as moderate but evolve through the course of an argument to become markedly more extreme. Indeed, this “commonplace” process of opinion extremisation online now represents the norm. By contrast earlier views of online extremism saw radicalisation mobilised by exposure to extremist propaganda and calls to action (as seen in the mid 2010s with Western recruits joining Islamic terrorist organisations). 

In our model, we observe that the effects of social interaction are amplified once charismatic influencers are introduced. This results in the dynamics of polarisation accelerating, with moderate positions disappearing rapidly. Importantly, our model shows non-linear effects, where small increases in an influencer’s appeal or activity can produce disproportionately large system-level consequences. This helps explain why a few highly visible figures can reshape entire digital communities. Viewed this way, polarisation can be seen not just as a problem of ideological change, but the rate of that change.

Certainty Increases Influencer Effect

In Betts et al. (2026), we show that influence depends on appeal, but we do not specify how appeal manifests in real digital environments. We argue here that one potentially powerful mechanism is the projection of certainty, particularly in complex social and political domains where ambiguity is uncomfortable, and disagreement costly. In these environments, certainty can function as a proxy for appeal to the message recipients by reducing interpretive effort, signalling authority, and accelerating group consensus.

This means that opinion leaders or influencers who present their views as inevitable or self-evident (in strong, non-ambiguous terms) are likely to lower the cognitive and social costs of (belief) alignment. In the terms of our model, this process can increase effective influence and help explain why particular voices dominate within the public space and drive faster polarisation. For example, scientific, or (fact laden) quasi-scientific language or reasoning is one commonly used technique by far-right political actors to increase their authority and provide an apparently inclusive context where hate is not explicitly expressed (Åkerlund, 2020; Ylä-Anttila, 2018).

Neutrality and Intellectual Humility to Reduce Polarisation

Our model shows that influencers can delay polarisation when they have a wide reach, but maintain a neutral position. Examples of institutions in the United States of America that may potentially act in this way include the Environmental Protection Agency and the Centre for Disease Control. Intellectual humility, understood as recognising the limits of models of thinking, tolerating uncertainty, and distinguishing facts from values (Krumrei-Mancuso et al., 2025), serves this function when presented in an inclusive way. In online debate, intellectual humility is visible when participants explicitly acknowledge the limits of their evidential and interpretive frameworks, tolerate unresolved uncertainty, and separate factual disagreement from value conflict. For instance, a participant in a contentious discussion on climate policy might concede that current models are uncertain at the regional level while still defending mitigation on normative grounds. Presented inclusively, this posture does not require self-effacement but rather, it helps participants identify whether they are contesting evidence, inference, or priorities.

Intellectual humility can reduce polarisation by slowing down convergence around single voices...

From a complex systems perspective, we argue that intellectual humility can reduce polarisation by slowing down convergence around single voices and keeping disagreement from hardening into factional identity. That is, it helps to maintain a healthy, civil disagreement in contexts where people hold opposing beliefs (Bliuc et al., 2024). In environments dominated by charismatic influencers, they act as a stabilising force by weakening the feedback loop between influence concentration and rapid polarisation. This helps maintain a core of moderate individuals “holding the middle ground” on a key issue, as our agent-based model results show.

Conclusion and Implications for Policy and Practice

Our findings (Betts et al. 2025) suggest that digital polarisation is not just a problem of divisive content, but also a structural one, as the dynamics of interaction online follow socio-psychological drivers such as attention, influence and popularity. Addressing the security risks posed by charismatic influencers requires reducing the rewards for “merchants of certainty”, slowing influence concentration, and treating tolerance of complexity as a core component of digital resilience.

Implications for practice are:

• Monitoring and early warning. 
  Our modelling suggests that an effective way to prevent polarisation and in particular radicalisation is for practitioners to focus on monitoring the speed and concentration of influence, not just the presence of extreme content being circulated in online social networks. Rapid convergence around a small number of strong, highly confident voices should be treated as a higher-risk signal than ideological extremity alone.
• Prevention and education.
  Preventive approaches should complement media literacy with complexity literacy to explicitly address the reasons why reductionist reasoning fails to capture the complete picture of a debate or issue. Complexity literacy should complement existing pre-bunking approaches and promote of critical thinking. This may include illustrating why simplistic explanations of complex social problems fail when queried from multiple perspectives. We propose that these types of approaches can reduce susceptibility to charismatic certainty without undermining trust in science.
• Platform governance/regulation issues.
  Because small changes in influence parameters can potentially have large downstream effects, platform design choices that reward performative certainty or amplify already dominant voices are security-relevant decisions. Governance that discourages/slows down/regulates unchecked influence can reduce polarisation without relying on censorship.