New Pilot Study Shows Machine Learning Can Predict Violent Students

After a violent event or tragedy, you often hear a lot of conjecture about the potential prevention of the occurrence.

With shootings, for instance, there’s generally talk surrounding the signs or lead-up to the attack. Could it have been prevented? What did those close to the attacker see that they should have reported?

Lots of people share their opinions on the matter, but that doesn’t necessarily shed any light on the reality of the situation. Is it even possible to find a way to reliably prevent these tragedies?

A recent study may provide some answers and some cause for hope.

Is It Possible to Predict Violence In Adolescents?

When a child is troubled and shows signs of violence or aggression, it’s apparent that they need support, which those close to them or mental health professionals may provide.

Before anyone can get assistance — and this is true for adults, as well — someone must notice the potential problem. Unfortunately, these issues are not always obvious. It’s an easy thing to miss, even for the parents of a child in need of help.

So, what’s the answer? How can we identify or know when someone needs help? It seems that modern technology may provide a solution.

Researchers from the Cincinnati Children’s Hospital Medical Center have discovered that machine learning can accurately determine risk for school violence.

Their study, which they published in the Psychiatric Quarterly journal, points out that it may be just as accurate — if not more so — than a team of child psychiatrists or even a forensic psychiatrist.

How Can Machine Learning Help?

It’s worth noting up front that the research conducted merely shows that predicting aggression is possible. This does not prove that machine learning technologies can prevent school violence.

Further study will be required to know that for sure, but tackling that question is the research team’s “next goal.”

The researchers concluded, though, that they can provide a more accurate way to problematic behavior which is, in itself, a huge revelation.

“Previous violent behavior, impulsivity, school problems and negative attitudes were correlated with risk to others,” says Drew Barzman, MD. Barzman is the lead author of the study and a child forensic psychiatrist at Cincinnati Children’s Hospital Medical Center.

The study involved 103 teenage students, from 74 U.S. schools, who had exhibited behavioral changes and aggression. Many of the students included were recruited from psychiatry outpatient or inpatient clinics and emergency departments.

The research team divided the participants into two groups based on scales they developed for risk level. They did so by using evaluations, which they transcribed from audio recordings.

After a direct interview, the researchers assessed both groups for the potential to exhibit violent behavior. They analyzed the information they obtained using a machine learning algorithm, which achieved a prediction accuracy rating of 91.02 percent.

When they also included demographic and socioeconomic data about the participants, the accuracy rating increased to 91.45 percent.

Co-author Yizhao Ni, Ph.D., a computational scientist working in Biomedical Informatics at Cincinnati Children’s, confirmed the system’s accuracy.

“The machine learning algorithm, based only on the participant’s interview, was almost as accurate in assessing risk levels as a full assessment by our research team, including gathering information from parents and the school, a review of records when available, and scoring on the two scales we developed,” Ni said.

What Does This Mean In the Real World?

According to this study, machine learning can predict violence. But what does this mean for the future of public safety? Could it be possible to prevent violent attacks altogether?

The truthful answer is we don’t know just yet. The researchers will be conducting further studies to see if their technology could help prevent violence altogether. We’ll need to wait for that before we can draw any conclusions, but this initial study suggests that machine learning could be useful in this area.

As for the machine learning industry, this study provides evidence that this is another area of modern society that this formidable technology could improve.

These findings will further bolster the industry and help introduce a wave of new uses for the technology. For instance, if it turns out you can use machine learning tools to prevent school shootings, what other violent events can you predict and potentially stop?

Time will tell, but it’s likely that machine learning will make all our lives safer and better someday soon.

Image by Pixabay