Texas Researchers Use Eye-Tracking to Make Lineups Fairer

Wrong eyewitness identifications send innocent people to prison, but researchers in Texas are developing tools that could help police build fairer lineups and reduce false convictions.

Dr. Curt Carlson at East Texas A&M University is leading a groundbreaking study on how artificial intelligence might revolutionize police lineups. His team received a $100,000 grant to examine whether AI-generated faces could help investigators create identification procedures that don't unfairly point witnesses toward suspects.

The challenge with traditional lineups is striking the right balance. Police need to include a suspect alongside "fillers," innocent people who match the general description of the perpetrator. If the suspect stands out too much, the lineup becomes unfair. If the fillers look too similar, witnesses can't make accurate identifications.

AI promises to solve part of this problem by generating realistic faces in seconds based on witness descriptions. Instead of searching through thousands of mug shots, investigators could create custom fillers instantly. But Carlson warns this convenience might introduce new problems.

"Having a real suspect's photo in there with AI fillers could be problematic," Carlson explains. Even highly realistic artificial faces might contain subtle characteristics that make the real photograph stand out, potentially biasing witnesses.

The research goes beyond simply asking which face participants select. The team collaborates with Dr. Dawn Weatherford at Texas A&M University-San Antonio, using Tobii eye-tracking glasses to record exactly where witnesses look while examining faces. The technology reveals which faces attract attention, how long people study them, and whether real and AI-generated faces produce different viewing patterns.

This detailed information helps researchers understand the cognitive processes behind identification decisions. Decades of face perception research show people often remember external features like hairstyles, even though internal features like eyes and noses are more reliable for identification.

Why This Inspires

This research represents exactly the kind of proactive work that can prevent injustice before it happens. Rather than waiting for AI to create problems in the criminal justice system, these scientists are getting ahead of the technology to ensure it serves fairness.

The project also creates opportunities for students to work on research with real consequences. Seven doctoral students and multiple undergraduates will contribute to the work, learning how laboratory science can directly improve people's lives.

The team plans to use their findings to apply for larger federal grants and ultimately provide practical recommendations to law enforcement agencies. Their goal isn't to prove AI is good or bad for lineups, but to discover exactly how and when it should be used to protect both innocent suspects and witnesses' ability to identify actual perpetrators.

The implications extend far beyond Texas courtrooms, potentially influencing how police departments nationwide conduct identification procedures in an age of increasingly sophisticated artificial intelligence.