Short description

This literature-master’s thesis invites a comprehensive review of how modelling choices—including feature selection, uncertainty quantification, domain-knowledge integration, and weighting of evidence—shape AI outputs in the forensic and criminal justice context.

This thesis should also draw parallels with high-stakes fields—pre-trial risk algorithms, emergency response triage, aviation autopilot assist systems—to critically assess how modelling standards, fairness, calibration, and transparency are set and whether forensic AI adheres to these best practices. The outcome should be a thematic literature synthesis that identifies key modelling pitfalls, best practices, and avenues for improving reliability in the criminal justice application of AI.

References

1) https://uitspraken.rechtspraak.nl/details?id=ECLI:NL:RBGEL:2024:3636
2) https://washingtondc.jhu.edu/news/ai-in-forensics/
3) https://journals.upress.ufl.edu/fa/article/view/2827
4) https://www.tandfonline.com/doi/full/10.1080/29974100.2025.2503343

Required/recommended skills and expertise

Understanding of AI

Information

Institute/ Company: Amsterdam University of Applied Sciences
Department: Research Group Forensic Science
Country: The Netherlands
Supervisor: Perle Russel
Uva Examiner: t.b.d.
Uva Coordinator: Arian van Asten/ Yorike Hartman

Date of publication: September 9, 2025