By Marcia-Ruth Ndege ‘21

In recent years, school violence has increased at an alarming rate. In a recent study, Dr. Drew Barzman, a child forensic psychiatrist at Cincinnati Children’s Hospital Medical Center (CCHMC), proposes the utilization of artificial intelligence (AI) as a predictor of how likely students are to perform violent acts while at school.

To perform this study, Barzman and his team recruited 103 American teenage students between the ages of 12 and 18. Many participants were patients in emergency rooms, inpatient units, and psychiatry outpatient clinics who had previously demonstrated aggressive behavior towards themselves or others. First, researchers performed audio-recorded risk evaluations on the participants. Factors taken into account during these evaluations included impulsivity, negative attitudes, and previous violent behavior. After performing the evaluations, researchers transcribed the recordings, then manually annotated the transcriptions. Results showed that the students were almost equally divided between moderately-high and low risk. Finally, researchers used artificial intelligence to predict violent behavior. Upon comparing the evaluation and the AI results, the researchers found that the machine learning algorithm had an accuracy rate of 91.02 percent when they used only interview content to generate the prediction and a rate of 91.45 percent when additional data such as socioeconomic and demographic data was accounted for in the prediction.  

Although this technique is still in the early stages of development, the results are extremely promising: the results obtained by using artificial intelligence were nearly as accurate as the tests performed by large teams of professionals trained to identify violent behavior. Ultimately, the researchers hope to further develop this technology to help prevent school violence.

References

1. D. Barzman, et. al., Pilot study validates artificial intelligence to help predict school violence. Psychiatric Quarterly (2018). doi: 10.1007/s11126-018-9581-8.
2. Image retrieved from: https://www.pexels.com/photo/macro-shot-of-water-drops-on-leaf-326461/