Fingerprinting Not That Unique, Study Suggests
AI Reveals You May Share Fingerprints With 204 Other People
Forensic science has long treated fingerprints as the ultimate personal identifier.
For over 100 years, we've trusted this method as a close-to-foolproof way of solving crimes. You’ve probably heard it said:
No two people have the same fingerprints.
New research, however, challenges this belief. It also raises questions about how fingerprints are used in court.
Random Overlap Probability
An AI-led method called Random Overlap Probability (ROP) has enabled researchers to analyze massive fingerprint databases and uncover patterns that challenge the traditional assumption of uniqueness.
The advanced computational power of AI allows for a more detailed examination of whorls, bifurcations, and other factors. Turns out, these minutiae can repeat across large populations quite frequently.
In a population of 14 million people, there's a 50% chance of two fingerprints matching by coincidence.
At 40 million people, overlaps are almost guaranteed.
Extrapolating those findings to a global population of 8.2 billion people results in 204 people sharing your fingerprints.
The chance that all of you are located in the same geographic location at the moment a crime is committed?
Highly improbable.
Cuellar is an Assistant Professor in the Department of Criminology, the Department of Statistics and Data Science, and the Department of Biostatistics, Epidemiology, and Informatics at the University of Pennsylvania.
Gold writes on his LinkedIn that he is “an undergraduate pursuing a Bachelor of Arts in Justice, Law, and Criminology with a minor in International Relations in addition to a Master of Science in Criminology as part of the School of Public Affairs Combined Program.”
Here’s the link to the full text from Maria Cuellar and Jackson Gold.
Key Takeaways
Note: I do not have a sturdy opinion on the research, but the following are its key findings (aka “the TLDR” if you don’t have time to dig into the full 13-page PDF).
This is what the authors are getting at:
Assumption of Uniqueness Challenged:
Fingerprint uniqueness, long considered a forensic cornerstone, lacks empirical evidence and is increasingly questioned by AI-driven research.
AI studies demonstrate that even fingerprints from different fingers of the same person show strong similarities.
Application of the Birthday Paradox:
By adapting the birthday paradox, researchers calculate that in populations of 14 million, there’s a 50% chance of fingerprint overlap, rising to near certainty in populations of 40 million.
Random Overlap Probability (ROP):
Introduced as a measure to quantify the likelihood of identical fingerprints in a population, highlighting significant overlap in cities with large populations (e.g., 100% ROP for New York City).
Historical and Modern Context:
Francis Galton's 1892 experiment laid early groundwork but overestimated fingerprint uniqueness. Modern AI methods show his assumptions are outdated.
Probabilistic Framework Proposal:
Advocates a shift from categorical fingerprint matching to probabilistic models to improve reliability and fairness in forensic science.
Practical Implications:
Calls for legal systems to recognize the probabilistic nature of fingerprint evidence to avoid overstating its reliability, reducing instances of the “prosecutor’s fallacy.”
Limitations:
Assumptions about fingerprint segment independence and the total number of unique fingerprints may introduce errors, suggesting a need for more nuanced models.
Recommendations:
Educate jurors and legal practitioners about the limitations of fingerprint uniqueness.
Adopt validated probabilistic methods for fingerprint analysis to provide clearer measures of uncertainty in legal cases.
History of Fingerprinting
The belief in fingerprint uniqueness goes back to 1892 when Sir Francis Galton created the first fingerprint classification system. Galton argued that fingerprints are both unique and unchanging.
This idea quickly became central to crime-solving. By the early 1900s, police and courts relied heavily on fingerprints to identify suspects. Training materials confidently declared, "No two prints will ever match."
Experts compared minutiae to decide when two fingerprints were a match. For decades, courts looked at these matches as compelling evidence.
But that didn’t mean they always got it right.
Moving Forward
On the surface, the new research seems like it might weaken the validity of fingerprinting. I see it as just the opposite.
Using AI, prosecutors and defense attorneys can now solidify their cases.
In other words, it’s not weakening the science of fingerprinting. It’s improving the certainty.
Ultimately, this will result in making evidence more reliable and reducing the chances of wrongful convictions.
Full-on AI integration is here, and it’s not limited to forensic science.
Subscribe to my Substack to stay up-to-date on all the ways this groundbreaking technology will continue to show up in our world.