Can AI-generated faces serve as fillers in eyewitness lineups?
Journal:
Memory (Hove, England)
PMID:
39977276
Abstract
To create a photo lineup for an eyewitness, police embed the suspect in a group of similar-looking individuals (i.e., fillers). If the witness selects the suspect from these photos of similar-looking people, then this provides evidence they remember the suspect from the crime event. Selecting appropriate fillers is one of the most difficult aspects of lineup construction. In these studies, we explore the feasibility of using artificial intelligence (AI) to generate images of fake people to serve as fillers in a photo lineup. Our results show that people largely fail to detect the differences in a lineup between the real photo of the suspect and the AI-generated fillers. Across different identification outcomes and related decision criteria, we found no evidence of differences between lineups with real and AI-generated fillers. Our study focused on a white, male target and future research is needed to test the generalizability of these findings. We conclude by discussing the feasibility and implications of the use of this technology in police investigations.