Controversy surrounds a novel AI-driven fingerprint-matching tool introduced to forensic science, capable of connecting prints from different digits belonging to the same individual. The groundbreaking technique, utilizing machine learning, raises debates among forensic experts regarding its practicality. While some foresee its potential to solve cold cases, others downplay its significance. Traditionally, forensic methods only accurately link fingerprints from the same digit, leaving law enforcement unable to match prints across different digits.
Published in Scientific Advances on January 12, the study claims that a machine learning system can link fingerprints from distinct digits of the same person with a 77% accuracy rate, surpassing the 50% expected by chance. Lead author Gabe Guo, a Columbia University computer science undergraduate, emphasizes the newfound ability to link fingerprints from different digits efficiently, aiding in criminal apprehension.
The researchers trained the AI tool with 53,315 fingerprints from 917 deceased individuals, then tested it on 133 individuals from a separate database. Unlike conventional forensic systems that focus on arches, whorls, and loops, the AI identified a new forensic marker—similarities in the angles of patterns across each finger.
Guo and his team faced challenges in getting their findings accepted by academic journals due to skepticism about the uniqueness of fingerprints. Simon Cole, a criminology professor at the University of California, contends that the similarities in fingerprints from different digits have long been recognized, though matching them precisely has been elusive for forensic experts.
Despite skepticism, forensic researchers acknowledge potential utility in the new findings. Ralph Ristenbatt, a criminalist and assistant teaching professor at Pennsylvania State University, notes that if the study’s claims hold true, even two prints from different locations could establish a connection. However, he emphasizes the AI’s current lack of courtroom accuracy.
While the researchers acknowledge the tool’s current limitations, they express intentions to collaborate with law enforcement to gather more fingerprints and enhance the AI’s accuracy for potential future use in solving crimes.