Will AI Help Predict Future Killers?
Machine learning shows promise for pre-homicide intervention.
Posted January 16, 2026 | Reviewed by Tyler Woods
A twelve-year-old girl was arrested in Florida last October after her notebook was found in the girls’ bathroom at her junior high school. She’d listed the names of other students she wanted to kill.
In Ohio the month before, two juvenile boys, ages 9 and 10, sexually assaulted and attempted to kill a five-year-old autistic girl. They lured her to a field and left her lying naked and unconscious .
What influenced such malign intent in kids so young? Can we spot them before they carry out a murder plan? Recent research offers hope.
Burtăverde, et al. (2026) at the University of Bucharest investigated links between childhood trauma , socioeconomic status, and emerging traits of the Dark Triad ( narcissism , Machiavellianism , and psychopathy). Two hundred seventy students recruited via online social media ads completed assessments like the Short Dark Triad Measure, the Childhood Trauma Questionnaire, and queries about living conditions during their childhood. Those who reported significant levels of abuse, neglect, or trauma also reported low socioeconomic status and subpar parental care. The researchers concluded that those who experience early trauma are more likely to display psychopathic traits. This is consistent with other studies, as well as with the development of secondary psychopathy as a reaction to abuse (Sethi, et al, 2018).
However, clinical data like this appears to offer only moderate levels of predictive accuracy for adolescents who might eventually commit homicide. Rodriguez, et al (2025) improved the forecast when they added neural data and machine learning analytics. “This study,” they wrote, “is the first to investigate the efficacy of both clinical and neuroimaging data collected from juvenile boys who are already deemed to be at a high risk for future antisocial behavior in predicting who will commit a future homicide during adulthood.”
They tested 202 formerly incarcerated youths for sixteen years, post-release. All had served a sentence in a New Mexico-based maximum-security juvenile detention facility between 2007 and 2011. Excluded were those with a low IQ , a TBI , or evidence of psychosis . Before the study commenced, the subjects had undergone baseline assessments and MRI scans. Records or self-reports showed that 35 (17%) had committed a homicide, post-release. They became the H group. Those who had not committed a homicide were the No-H group.
The researchers used Weighted Linear Support Vector Machines to develop models from the combinations of variables that would assist with the prediction of future homicidal acts. The model that included clinical plus neural variables showed 76% accuracy vs. clinical data alone at 65% accuracy.
Results and Implications
Implicated in the risk for future homicidal acts were a specific neural profile, an earlier age of first arrest, and elevated psychopathic traits, according to scores on the Psychopathy Checklist: Youth Version (PCL:YV). This assessment, based on the well-established PCL-R adult psychopathy assessment, identifies potential patterns of cheating, deception , fighting, bullying , callousness, and similar antisocial acts. Subjects in the H group scored higher on baseline measures of these traits and also showed antisocial behavior at an earlier age than the No-H group. In addition, their MRI scans indicated reduced gray matter in the bilateral amygdalae, middle temporal pole, and right superior temporal pole.
The study concludes that “there is improved utility when incorporating both clinical measures and neuroanatomical metrics in predicting future homicidal behavior among high-risk youth.”
The treatment implications are clear: Evidence of psychopathic traits and early antisocial behavior, coupled with abnormalities in those brain regions associated with emotional and social learning , could provide a foundation for preventative treatment in those deemed likely to commit future fatal violence.
It’s a promising start, but this same research must first be expanded to larger, more diverse, and more representative groups. Still, gaining neural profiles of kids could pose a challenge. In certain cases, within clearly defined parameters, this research could justify it.
Burtăverde, V., Jonason, P. K., Minulescu, A., et al. (2026). Childhood trauma and life history strategies—the moderating role of childhood socio-economic status and the dark triad traits. Personality and Individual Differences. 248 . 113467. 10.1016/j.paid.2025.113467.
Forth, A. E. (2005). Hare Psychopathy Checklist: Youth Version. In T. Grisso, G. Vincent, & D. Seagrave (Eds.), Mental Health Screening and Assessment in Juvenile Justice (pp. 324–338). The Guilford Press.
Rodriguez, S.N., Gullapalli A.R., Stephenson, D.D. et al. (2025). Machine learning of clinical and neural data predicts future homicide in high-risk youth. Scientific Reports. https://doi . org/10.1038/s41598-025-32782-5
Sethi, A., McCrory, E., Puetz, V., et al. (2018). Primary and secondary variants of psychopathy in a volunteer sample are associated with different neurocognitive mechanisms. Biological Psychiatry. Cognitive Neuroscience and Neuroimaging , 3 (12), 1013–1021. https://doi.org/10.1016/j.bpsc.2018.04.002
Share this post Facebook Bluesky Linkedin Email
There was a problem adding your email address. Please try again.
By submitting your information you agree to the Psychology Today Terms & Conditions and Privacy Policy
Katherine Ramsland, Ph.D., is a professor of forensic psychology at DeSales University and the author of 69 books.
Get the help you need from a therapist near you–a FREE service from Psychology Today.
This article is part of the Bringwise Psychology Journal — daily insights on human behavior, mental health, and personal growth.