Doctoral research by Alice Caputo  

Model: Statistical risk-scoring AI + Intervention 

Judicial decision support  ➔ Theme: Predicting social dangerousness  

Criminal justice and artificial intelligence make for a complex pairing — especially around one of the most debated concepts in Italian criminal law: "pericolosità sociale," a judge's prediction of whether someone will reoffend. Can machines forecast human conduct that even humans can't predict with certainty? This case designs a statistical scoring model, grounded in Article 133 of the penal code, meant to support — not replace — the judge's assessment. The design challenge is building a transparent tool that assists reasoning about risk without reducing a complex human judgment to a single score. 

"For years, judges assumed anyone with a mental illness was socially dangerous — data later showed that presumption was largely false. A well-designed statistical tool has to actively avoid encoding that same kind of bias. "

Alice Caputo, Uninettuno 

Supervisor: Prof.ssa Silvia Massi