Throughout the years, experts have tried to stem the global, perpetual problem of violence using a variety of tactics, from targeted policies to amped-up policing. But for Hans Breiter, ’88 MD, the solution is a matter of math.
Breiter, a professor of Psychiatry and Behavioral Sciences, is leading the development of a radical, proactive approach to stopping violence using advanced mathematical models of human emotion. Used to pinpoint individuals with the potential to commit violence or terrorism, these models could inform positive interventions to prevent violent acts from occurring in the first place.
Though the concept may sound like the stuff of science fiction, Breiter says it could be implemented in the relatively near future.
“We have now shown proof of concept for critical components of a software system that predicts who has a higher likelihood of producing violence, before they do so,” he says. “With further development, the current prototype could possibly reduce violence in cities like Chicago.”
No doubt, the practice of predicting violence raises a number of deep ethical questions — but Breiter is intent that the intelligence be used for good. If at-risk individuals could be identified early, he proposes, community organizations and local non-profits could step in with positive measures to produce a different future.
“Bringing people who have not committed violence yet into contact with individuals who can help, whether it’s with education, a job or dealing with life stresses, costs much less than increased policing, jail or what happens to a family when one of their members is murdered,” says Breiter, who is also director of Northwestern Medicine’s Warren Wright Adolescent Center. “With a little intervention, we can increase the chance someone becomes a positive member of society — and doesn’t take away someone else’s kid.”
Currently, there are few metrics to accurately predict which individuals might turn to violence, and most focus on past offenders who might strike again. But to accurately zero in on adolescents who show no history of violence — a group where early intervention might have the most impact — is a much more formidable task. It’s one that requires applying engineering principles to violence and mathematical models to messy human emotions, both areas that seem to defy any sense of lawfulness.
But for Breiter, it’s an endeavor that has marked much of his career.
Finding Lawfulness in Behavior
After racing through four years of mathematics classes at a local college while still in high school, Breiter earned a medical degree through the Honors Program in Medical Education at Northwestern and completed a program in logic and metaphysics at St. Andrew’s University in Scotland. He then embarked on a medical career at Massachusetts General Hospital (MGH) in psychiatry and neuroimaging, before delving into neuroeconomics and mathematical psychology.
“I wanted to understand why people make irrational decisions. But I realized we needed to move past older behaviorist models of stimulus-response action and actually use the mathematics of information theory,” Breiter says.
At MGH, he dove into quantitative brain morphometry, applying magnetic resonance imaging (MRI) to the study of psychiatric illness. He pioneered multiple analysis techniques using functional MRIs and applied their first use in psychiatry and the emerging field of neuroeconomics. He spent four years working with Daniel Kahneman, PhD, who later won the Nobel Prize in Economics in 2002 for prospect theory, which explores how people make decisions involving risk.
Breiter went on to serve as principal investigator on the Phenotype Genotype Project in Addiction and Mood Disorder from 2003 to 2009, a major scientific collaboration to link brain reward circuitry with genes involved in depression and addiction. In that data, he and his team stumbled across a key discovery that set him on his current path: law-like patterns of human reward-based decisions, a concept the scientists have since termed “relative preference theory.” One of the few descriptions of human behavior that meets the criteria for physics-based lawfulness, the breakthrough provided the team with the beginning of a mathematical framework for human emotion.
“We could start getting at deep traits of approach and avoidance with just two minutes of experimental data,” Breiter says. For example, during one of the project’s experiments, subjects were asked to rate how much they liked 48 randomized pictures, on a scale of negative three to positive three. The scientists then analyzed the results using information theory mathematics. “With a simple set of picture ratings, we could quantify patterns in people’s responses and detail their emotional traits.”
The framework has applications beyond violence prevention, to any field in which emotion may be important, including medicine. For example, measures of emotional traits might reveal a patient is prone to addiction — information physicians could take into account when prescribing pain medications.
Between 2008 and 2010, Breiter published a series of neuroimaging and behavior articles describing relative preference theory, and he has continued to expand the concept since joining the Feinberg faculty in 2011.
“Ultimately, I want to help move behavioral science to the same level of mechanism that goes on with chemistry, molecular biology and genetics,” says Breiter, noting a paper he published recently in Frontiers of Psychology on the topic. “Although much more work is needed, this has the potential to affect neuroscience and how we develop pharmacology.”
Seeing Patterns in Social Media
But while such science could have major implications in medicine and other areas as it evolves, Breiter has been intent on applying his discoveries to a more immediate problem: the epidemic of violence.
The project is deeply personal to him. “My father was in a concentration camp. My Eastern-European family was obliterated by the Nazis. And every time I hear about some kid being killed in Chicago, I get very upset. To me, working against violence is a sacred endeavor,” Breiter says.
His team has made important practical advancements in the use of relative preference theory. While experiments like picture ratings require active data acquisition from subjects, the scientists have realized the same revealing patterns about behavior can be pulled from more accessible outlets, such as social media posts. That’s critical to adapting relative preference theory to predict the potential for violent behavior across large populations not actively involved in experiments.
In short, the software runs complex mathematical models on material publicly available on sites like Facebook or Twitter to measure relative preference traits. After compiling a trait map for the population, the scientists can annotate whom in that emotional space has committed violence in the past. “And the people who are the nearest neighbors in this 13-dimensonal emotional trait space have a very strong likelihood of doing similar actions,” Breiter says.
The scientists have already proven the basic efficacy of such a software system in predicting unethical or violent behavior among people in private organizations. Breiter now hopes to bring it to larger use.
“I hate to use the analogy, but it might remind people of Minority Report, without the science fiction or Tom Cruise,” Breiter says, referring to the 2002 film set in a future where psychics help police departments apprehend criminals before they commit their crime.
The system departs from the movie in another key aspect: The endgame is prevention, not punishment. “I want an anti-violence system, unaligned with the police, that works with foundations to intervene with these youth in a positive, constructive way — not in a punitive way,” Breiter says. “They have done nothing wrong and the potential for violence is not a crime.”
He cites organizations like BUILD Chicago, which specializes in youth violence prevention, as non-profits that might play a role in making such a system a reality. To do so, he is seeking to build a team with a wide range of collaborators, including neuroscientists, artificial intelligence experts, software engineers, ethicists and forensic neuropsychologists.
Robert Hanlon, PhD, associate professor of Psychiatry and Behavioral Sciences and of Neurology, is one such expert who might serve as a consultant on the anti-violence system. “Our research has identified behaviors, tendencies and personality traits that are associated with future violence, so we can help people understand these factors and make informed predictions — because that’s really how you prevent violence,” Hanlon says. “I hope to help see Hans’ ideas come to fruition.”
Of course, the system is not without possible risks. While Breiter acknowledges that potential fears of security, privacy and misuse are real concerns, he explains that the software looks for patterns — not specific information. “The social media content is actually irrelevant,” Breiter says. “It’s the pattern of how you say things, and how they might shift from normative patterns, that is important.” To try and counter the potential for its abuse, the anti-violence system would have to be developed with oversight from nonprofit organizations, as well as an ethics board and community engagement.
After seeing positive test results for critical pieces of the program, Breiter is currently exploring funding to build, field-test and optimize such an anti-violence system.
“Violence is a domain that as a society we have to attack head-on,” he says. “Instead of spending millions on someone in the criminal justice system, let’s get ahead of the violence, before something becomes an irreversible event. The goal is targeted intervention. If we can do that, I believe we can positively affect the world around us.”
More Violence Prevention Research at Feinberg
Breiter isn’t the only Northwestern investigator leading violence prevention research. A sampling of scientists throughout the medical school focused on related work:
Linda Teplin, PhD, Owen L. Coon Professor of Psychiatry and Behavioral Sciences
Teplin, who has studied violence for more than 30 years, recently received a grant from the U.S. Department of Justice to study the perpetration and victimization of firearm violence among youth. The grant is an expansion of the Northwestern Juvenile Project, a comprehensive prospective study of health needs and outcomes of delinquent youth. With more than $43 million in funding since 1998, it is one of the largest studies in Feinberg’s history.
Lori Post, PhD, Buehler Professor of Geriatric Medicine
Post, a violence prevention expert who joined Feinberg from the Yale School of Medicine in early 2017, is the inaugural director of the Buehler Center for Health Policy and Economics. She recently received funding to investigate gender-based violence in low-income countries, including widow burning and female genital mutilation.
Hanlon and Brook, who co-direct the forensic psychology research lab at Feinberg, investigate the neuropsychology of violence, including the behavioral personalities and cognitive aspects of psychopaths. To date, the team has published more clinical studies on the neuropsychological profiles of murders than any other lab in the U.S. In two upcoming papers, they identify features of murderers of children, as well as murderers who are female, a rarely studied demographic.