AI as a decision support tool in forensic image analysis: A pilot study on integrating large language models into crime scene investigation workflows

Abstract

This study evaluates the effectiveness of artificial intelligence (AI) tools (ChatGPT-4, Claude, and Gemini) in forensic image analysis of crime scenes, marking a significant step toward developing bespoke AI models for forensic applications. The research involved independent analysis of 30 crime scene images by the AI tools, with the resulting reports rigorously assessed by 10 forensic experts. Findings reveal promising potential for AI as a decision support tool in forensic science, serving as a rapid initial screening mechanism to assist human experts in their comprehensive analysis. The results emphasize that current AI tools function optimally as assistive technologies, enhancing rather than replacing expert forensic analysis, particularly in scenarios involving multiple evidence points or high-volume caseloads. The AI tools demonstrated high accuracy in observations but faced challenges in evidence identification, with performance varying across different crime scene types—excelling in homicide scenarios (average score of 7.8) but encountering difficulties in arson scenes (average score of 7.1). This study's findings could significantly impact investigative procedures, forensic training, and the development of AI tools in law enforcement, while emphasizing the importance of establishing robust ethical guidelines for the integration of AI in criminal justice systems.