(CN) — Researchers from Columbia University revealed a new artificial intelligence system this week that disproves the long-held idea that each human finger has a unique fingerprint.
The discovery — published in Science Advances on Friday — stems from a research collaboration between the Creative Machines lab at Columbia and the Embedded Sensors and Computing lab at the University of Buffalo, SUNY. Its lead researcher, Gabe Guo, is an undergraduate computer science student at Columbia’s School of Engineering and Applied Science.
Guo — along with three other student researchers and engineering professors Hod Lipson and Wenyao Xu, the directors of the Creative Machines lab and Embedded Sensors lab, respectively — developed an AI system that challenges a prevailing, yet unproven conviction in forensics.
Since the early 20th century, criminal investigators have zealously used modern concepts of fingerprinting as tangible evidence to indict and charge criminals. One of the earliest examples in the U.S. occurred in 1910 when a jury convicted Thomas Jennings of murder after prosecutors distributed enlarged copies of his fingerprints.
However, the practice has since led forensics experts to believe that every fingertip on a person’s hand possesses a unique print, making them unmatchable. The occurrence — known as intra-person fingertips — has made it difficult for investigators to link individuals across crime scenes. It also prevented the researchers from publishing their results for months.
“Unveiling the similarity”
Inspired by previous studies indicating a potential link between genetics and fingerprint patterns, the researchers set out to find similarities between fingerprints of the same person. To do so, they sourced roughly 60,000 fingerprints from a public U.S. government database and fed the data in pairs into an AI-based system called a “deep contrastive network.”
The system — which essentially compares and contrasts images — eventually reached a point where it could detect when seemingly unique fingerprints belonged to the same person and when they didn’t, reaching an accuracy rate for a single pair at 77%.
“Contrary to this prevailing assumption, we show with above 99.99% confidence that fingerprints from different fingers of the same person share very strong similarities,” the authors wrote in their study. “Using deep twin neural networks to extract fingerprint representation vectors, we find that these similarities hold across all pairs of fingers within the same person, even when controlling for spurious factors like sensor modality.”
The finding led the researchers to conclude that the AI utilizes a new type of forensic marker. Rather than comparing the branching and endpoints of fingerprint ridges — the traditional way of analyzing prints — Guo said the AI’s technique involves the angles and curvatures of the swirls and loops from the center of the fingerprint.
“Our experiments suggest that, in some situations, this relationship can increase forensic investigation efficiency by almost two orders of magnitude,” the authors wrote.
Eager to share their results, the team submitted an initial manuscript to an esteemed forensics journal. The expert reviewer from the undisclosed journal rejected the study for publication, telling them, “It is well known that every fingerprint is unique.”
The researchers — now aware of the forensic community’s skepticism — continued to improve the AI system with more data before submitting their next manuscript to a more general audience. The paper met rejection again.
Lipson appealed to the second publication, also undisclosed, reasoning that the finding was too important to ignore.
“If this information tips the balance, then I imagine that cold cases could be revived, and even that innocent people could be acquitted,” Lipson said in a statement.
After nine months, Guo said, Science Advances accepted the paper for publication. And as indicated by Lipson, the new system can potentially improve the criminal justice system — once its accuracy is good enough, that is.
Before using the AI to bust crime, the researchers say broader datasets are necessary to overcome potential data biases. So far, their research indicates that the AI’s outputs have held regardless of race or gender, where samples are available.
“Many people think that AI cannot really make new discoveries, that it just regurgitates knowledge,” Lipson said. “But this research is an example of how even a fairly simple AI, given a fairly plain dataset that the research community has had lying around for years, can provide insights that have eluded experts for decades.”
Lipson also highlighted his excitement around how Guo — an undergraduate student with no prior background in forensics — developed an AI to successfully challenge a widely held belief of the forensics field.
“We are about to experience an explosion of AI-led scientific discovery by non-experts, and the expert community, including academia, needs to get ready,” Lipson said.
Follow @alannamayhampdx
Subscribe to Closing Arguments
Sign up for new weekly newsletter Closing Arguments to get the latest about ongoing trials, major litigation and hot cases and rulings in courthouses around the U.S. and the world.
