This ‘Psychological' AI System Mimics Religious Violence to Understand It

An international team of scientists has studied the causes of religious conflict using a modeling system based on cognitive psychology.

religious violence, religious conflict, St. Bartholomew's Day massacre
Detail of a painting by François Dubois of the St. Bartholomew's Day massacre, in which thousands of French Protestants were killed over weeks of Catholic mob violence. | Musée cantonal des Beaux-Arts
Detail of a painting by François Dubois of the St. Bartholomew's Day massacre, in which thousands of French Protestants were killed over weeks of Catholic mob violence. | Musée cantonal des Beaux-Arts

Religion is often said to be at the root of the world’s conflicts, but what if mankind were simply naturally violent? How might we begin to understand the circumstances that trigger spasms of religious violence in the first place?

Artificial intelligence has emerged as one method for doing so.

In a first-of-its-kind research study published this week, scientists unveiled a newly developed AI system that combines computer modeling and cognitive psychology to simulate the phenomenon of human religiosity on a large scale. The goal of the system is to help social scientists better understand the conditions and triggers that lead to religious violence — and perhaps even prevent it.

Appearing in The Journal for Artificial Societies and Social Stimulation, the study was designed by an international team of computer scientists and sociologists from the University of Oxford, Boston University, and the University of Agder in Norway.

The research team created a virtual society populated with millions of sims — human model agents that were programmed to react and behave as real humans, using core principles of cognitive psychology. The sims were assigned different ages, races, and ethnicities, along with sets of core beliefs, identity characteristics, and “sacred values.”

RELATED: How Does Religion Manifest in Your Brain?

Previous studies have taken a similar route to modeling human behavior and ethics on a large scale, but this AI system introduces a new element. Rather than employ the usual machine-learning algorithms that power typical AI systems, the study incorporates “psychological AI” systems to mimic how human beings actually think and process information.

Gustave Doré, religious violence, Granada
An engraving by Gustave Doré depicting the crusade against the Moors of Granada. | Wikimedia Commons

“Ultimately, to use AI to study religion or culture, we have to look at modeling human psychology, because our psychology is the foundation for religion and culture,” said researcher Justin E. Laned in a statement. “The root causes of things like religious violence rest in how our minds process the information that our world presents it.”

The team programmed rules for cognitive interaction within the AI simulation to show how an individual’s beliefs match up with a group situation. Some of the AI human models were programmed to have had positive experiences with people from other faiths, while others had negative or neutral encounters.

The researchers then set about running the simulation through various scenarios using groups of human sims in various group sizes involving hundreds, thousands, and even millions of people. The simulated environments generated probabilities that the sims would encounter various environmental hazards, such as natural disasters and disease — and at some point, each other.

RELATED: Artificial Intelligence Is Mapping the Pacific Ocean’s Secret Soundscape

The study found that people are generally peaceful, working together when faced with disasters, but also revealed that long periods of mutually escalating xenophobic tension occur when outgroup members deny another group’s core beliefs or sacred values.

“Religious violence is not our default behavior — in fact, it is pretty rare in our history,” said study co-author Justin Lane.

“It is only when people’s core belief systems are challenged, or they feel that their commitment to their own beliefs is questioned, that anxiety and agitations occur,” says the study’s accompanying materials. “However, this anxiety only led to violence in 20 percent of the scenarios created — all of which were triggered by people from either outside of the group, or within, going against the group’s core beliefs and identity.”

The AI system was informed by two historical periods of religious violence: A Northern Ireland conflict that claimed the lives of more than 3,500 people, which is commonly known as “the Troubles”; and the 2002 Gujurat riots in India in which Hindus clashed with Muslims, killing at least 1,000 people, most of them Muslims.

The ultimate goal of the AI system, researchers said, is to provide a digital model of society — one in which social scientists can study the escalation of violence over time, and how to best manage such crises. This would potentially equip governments with a tool that could offer guidance on how to respond to problems before they escalate.