A better understanding of radicalisation through AI
Phd Defence Mijke van den Hurk at National Policelab AI
Radicalisation is a complex process, influenced by many variables that interact with each other to varying degrees. AI scientist Mijke van den Hurk investigated whether artificial intelligence could help unravel this intricate interplay. She defended her PhD thesis cum laude on 29 June.
In the spring of 2019, Mijke van den Hurk was on holiday in Sri Lanka when a terrorist attack took place. Over 250 people were killed in suicide bombings targeting churches and luxury hotels. The attacks were claimed by a group affiliated with the radical Islamist organisation IS.
“It was shocking to witness. There was so much fear and uncertainty. The attackers were wealthy, highly educated, and some were women. They did not fit the stereotypical image many people have of terrorists. It left me wondering: what pushes people to do this to one another?” says Van den Hurk.
She started reading books, listening to podcasts, and watching documentaries on the subject. “It became clear that this question is not easily answered. There are many reasons why people join extremist movements. It depends, among other things, on a person’s need for recognition, social status, peer pressure, and external circumstances.”
National Policelab AI
Understanding radicalisation is crucial, given the significant impact of terrorist attacks. Van den Hurk, who graduated cum laude from the MSc Artificial Intelligence programme and had just started as a data scientist at the police, wondered if artificial intelligence could help.
Thanks to the collaboration between Utrecht University and the Dutch police, she was able to focus on the topic long-term within the National Police Lab AI as part of her PhD research. “The beauty of the collaboration between science and practice is that you create a shared representation of a complex subject and expand collective knowledge about it.”
Despite years of research, there is still no single theory that adequately explains why - and, more importantly, how - people radicalise, Van den Hurk explains. “Feeling attracted to an ideology that incites violence is still many steps away from actually committing violence. The multitude of factors and their interplay makes it difficult to intervene in the right way and at the right time.”
Feeling attracted to an ideology that incites violence is still many steps away from actually committing violence
Van den Hurk explored whether Agent-Based Modelling (ABM) could provide answers. ABM is a modelling technique that maps the complexity of human behaviour. It allows you to simulate how individual entities interact over time and what effects this has on their behaviour. “You can think of it as a kind of virtual environment where thousands of people walk around with different personality traits, reacting to each other and their surroundings according to certain behavioural rules. By observing what happens when these individuals influence each other, you can study how radicalisation develops.”
Van den Hurk developed a model that combines different theories on how individuals radicalise, where a person’s agency increases or decreases through the interaction of various social and psychological mechanisms. Through simulation, it became clear how factors such as social influence, feelings of exclusion, identity formation, or exposure to extremist ideas could reinforce or, conversely, inhibit each other.
Within the model, radicalisation emerged from normal social processes that only coincide and reinforce each other under unique circumstances, leading to radical behaviour. Van den Hurk notes: “The direction it takes is not determined by a person’s fixed personal traits, but by which mechanisms have the greatest influence at a given moment. For example, we saw that some small groups radicalised as a counter-reaction to a large majority, but only if the members of that group had no alternative way to meet their social needs.”
Interactive dashboard
Van den Hurk’s model can help gain insight into the processes involved in radicalisation: why does one person radicalise under similar circumstances while another does not? “The model cannot predict when an individual will radicalise or when someone will carry out an attack. However, in the future, it may be possible to create an interactive dashboard where you can adjust different parameters to analyse potential radicalisation scenarios. This could contribute to better policy and appropriate interventions.”
The question that stayed with Van den Hurk after the terrorist attacks during her holiday in Sri Lanka has not been answered by her thesis. “But the research, in collaboration with the police, has helped me better understand radicalisation, both as a societal phenomenon and from my own experience with it.”
Vaccinweigeraar Bob
In Mijke van den Hurk’s thesis, we follow the fictional character Bob, who joins an anti-vaccination group during the coronavirus pandemic while also being a member of a sports club where most members support vaccination. He thus finds himself in two social environments with different norms, values, and expectations.
The anti-vaccination group is tightly knit: members distribute pamphlets, take action, and find recognition in each other. At the same time, they feel threatened by the majority, which strengthens their us-versus-them sentiment. The sports team has a looser structure and approaches vaccination pragmatically: as a condition to continue playing sports together.
The model shows that Bob’s development stems from an interplay of social and individual processes, such as his need for recognition and belonging, the dynamics between groups, his expectations of others, and the choices he makes based on these. If Bob feels a stronger connection to the anti-vaccination group, he may distance himself from his sports team. This changes his social network, and alternative perspectives may diminish.
The model thus demonstrates how social processes can reinforce each other, without predicting whether someone will actually radicalise.