AI has grown quite adept at spotting trends in large quantities of data, and what’s more, doing so slightly faster than more traditional methods.  One of the more interesting use cases of this approach recently emerged via a new paper from Oxford University, which used AI to better understand the causes of religious violence.

The researchers look for two specific periods of xenophobic social anxiety that are then commonly followed by periods of extreme physical violence.  The researchers use multi-agent AI to construct a psychologically realistic model of a human, including how we think and how we identify with specific groups.

These agents were created using realistic theories of cognitive psychology to understand how we think and process information.  It’s one of the first times that this information has been coded into an algorithm for use in this way.

Processing information

The researchers attempted to understand how we process information against our personal experiences.  They were then able to combine several AI models, some of which aimed to mimic people who had positive experiences with people of different faiths, and some which mimicked people who had not.  The aim was to simulate the escalation and de-escalation of tensions over time, and to explore whether it’s a process that can be managed.

Each simulated scenario contained millions of human model agents, with each agent designed to have slightly different variables in terms of age, ethnicity and so on.

As a result of the simulations, the team discovered that the most combustible conditions tend to involve so called social hazards, which include things such as outgroup members denying the core beliefs of the ingroup.  This begins to overwhelm people to the extent that they can’t be dealt with any longer.  This challenge to the core beliefs of the majority raises anxiety, but even then, in only 20% of such scenarios did violence result.  In these instances, there was always a trigger caused by someone outside of the group going against the core beliefs of the group.

Understanding the escalation of tensions

The team believe that their work represents the first time that AI and machine learning have been able to provide an accurate reflection of how religious tensions emerge, especially in terms of taking account of the biases against minority communities.

“Ultimately, to use AI to study religion or culture, we have to look at modelling human psychology because our psychology is the foundation for religion and culture, so the root causes of things like religious violence rest in how our minds process the information that our world presents it,” they explain.

The team hope that by providing a higher degree of understanding of the root causes of religious violence, it will help communities to better minimize the risk of conflicts emerging, whilst also supporting the integration of different communities.

Their next step is to build upon this work and study the various attempts to integrate communities such as the Roma in Slovakia or Syrian refugees across Europe in order to help governments understand how best to optimize the integration process.

Given the ingroup/outgroup tensions that seem to be arising around the world at the moment, it’s work that can’t come soon enough.