AI and Ted Bundy: Exploring Artificial Intelligence usage in criminal profiling

Abstract

We have seen a meteoric rise in Artificial Intelligence (AI) creation and usage; yet little attention has been paid to using AI for societal benefits. To leverage AI in new ways, this study focused on integrating AI in criminal profiling. We assessed the accuracy of AI-generated profiles by modifying well-documented criminal cases with a known, and convicted, perpetrator to compare to published perpetrator data created by (human) experts. Using profiling inputs from the crimes of infamous serial killer Theodore ‘Ted’ Bundy, we prompted ChatGPT 3.5 to create profiles for the first eight attacks Bundy committed and compared the accuracy of the outputs to the known information. Initial results show AI creates vague but detailed, and fairly accurate, profiles compared to known information from cases and can find patterns between crimes. But, these profiles were created with specific prompts and the prompt type impacted accuracy. Our findings suggest that using AI in profiling warrants further research and consideration in ongoing investigations, potentially saving time and lives. Caution is advised given the limitations regarding specificity of details and we do not yet know if human-generated profiles are more accurate as we only compared the AI profiles to known information. Future research should compare AI-generated profiles with human-generated profiles and explore paid versions of AI that might reveal further capabilities which might be useful in law enforcement, where costs of using AI may be nominal, especially in relation to the savings of lives and in manpower hours.