|
Getting your Trinity Audio player ready...
|
Columbia Engineering senior, Gabe Guo and his team have embarked on a study utilising an artificial intelligence (AI) system that has cast doubt on the longstanding belief in forensics that intra-person fingerprints (fingerprints from different fingers of the same person) are inherently diverse and unmatchable. Guo’s approach involved feeding pairs of fingerprints, some from the same individual (but different fingers) and others from different people, into a modified deep contrastive network—an AI-based system.
Despite facing scepticism within the forensics community and initial rejection from forensics journals, the team persisted in his pursuit. They recognised their work’s importance and its potential impact on solving these intricate cases.
In light of this, Guo’s team are starting their experiment. Guo utilised a public U.S. government database containing around 60,000 fingerprints. The AI system, designed by the team, demonstrated results over time.
The AI system demonstrated a capacity to identify patterns, nuances, and associations within the fingerprint data that human analysts might overlook or struggle to process efficiently. Its ability to handle multiple pairs simultaneously showcased the potential for rapid and accurate processing of large datasets.
The accuracy for a single pair reached 77%, and when multiple pairs were presented, the accuracy soared significantly, potentially revolutionising forensic efficiency by more than tenfold.
The central focus of this study revolved around uncovering the alternative information that the AI utilised, evading forensic analysis for decades. The AI system’s decision-making process visualisations exposed a novel forensic marker. Unlike depending on minutiae, the conventional branchings and endpoints in fingerprint ridges, the AI concentrated on the angles and curvatures of the swirls and loops in the centre of the fingerprint.
While AI’s accuracy isn’t sufficient to officially decide a case, it can be valuable in prioritising leads in ambiguous situations. The team was acutely aware of potential biases in the data used to train the AI system. Understanding the importance of fair and unbiased algorithms, they actively sought to address these concerns.
To mitigate biases, the team continuously diversified their datasets, incorporating fingerprints from individuals across various demographics, ethnicities, and backgrounds. This comprehensive approach ensured that the AI system would not inadvertently favour certain groups or demographics over others.
This discovery is indicative of the surprising capabilities of AI. Professor Lipson noted that many people perceive AI as a tool that merely regurgitates existing knowledge. Still, this research showcases how even a relatively simple AI, working with a dataset available for years, can provide insights that have evaded experts for decades.
Further, the study highlighted the democratising potential of AI-led scientific discovery. Gabe Guo, an undergraduate student with no background in forensics, successfully challenged a widely held belief in the field using AI. This foreshadows an imminent era of AI-led scientific exploration by non-experts, prompting the expert community, including academia, to prepare for the upcoming surge in AI-led discoveries.
Transparency became a guiding principle for Guo and his team. They committed to sharing their methodologies, validation processes, and any identified limitations openly with the forensic community. This transparency built trust among their peers and encouraged a collaborative effort to continually refine and improve the technology.
As AI applications in forensics continued to evolve, so did the ethical standards and guidelines governing their use. The scientific community, legal professionals, and regulatory bodies collaborated to establish best practices and ethical frameworks for integrating AI into the criminal justice system. This ongoing dialogue ensured that the benefits of AI in forensic science were realised without compromising the principles of fairness, accountability, and justice.
Guo’s team’s commitment to addressing biases and advocating for transparency and validation set a precedent for responsible AI implementation in forensics, fostering a culture of ethical innovation. Their work not only pushed the boundaries of technology but also paved the way for a future where AI and human expertise collaborated seamlessly to enhance the accuracy and efficiency of criminal investigations.
