Skip to content

Improving Witness Identification

INACCURATE eyewitness identification of suspects has been pinpointed as a major factor behind wrongful convictions. Two factors are believed to contribute to mistaken identification when witnesses are asked to pick a suspect from a police lineup: the first is that the person conducting the process—usually the police officer associated with the investigation—may inadvertently influence the choice; the second is that the other persons selected as “fillers” in the lineup may affect the result.

To address the first issue, many eyewitness identification experts recommend that the person administering an eyewitness lineup be “blind” or unaware of who the suspect is and who the fillers are in the lineup (this process is also known as “double blind”). But guaranteeing that an officer unfamiliar with an investigation carries out the lineup can be difficult and inconvenient.

One new tool that may be able to facilitate blind lineup administration is a virtual police officer known as Officer Garcia, described in a study in Police Quarterly.

Here is how Officer Garcia works: he is a computer animated (virtual) human who administers the lineup photo array to a witness. He has the witness confirm each answer twice. A sequential photo array is displayed for the witness, and although an officer may still be in the room for this administration, he would have no idea which photo was being displayed at which time.

Brent Daugherty designed Officer Garcia as part of his computer science master’s thesis at the University of North Carolina at Charlotte. He has a background in a field of computer study called virtual humans, and he says he embarked on this project in part to aid witnesses who may not be familiar with computers. “A lot of these witnesses either come from a background where they’re not exposed to computers or they don’t use them in their daily lives,” says Daugherty.

One police department in North Carolina has agreed to test the software. Others say it is not ready for prime time.

“I think that it could be very promising, and I like the idea of figuring out some way to have a reliable administration of a procedure,” says Stephen J. Ross, a visiting assistant professor of psychology at Florida International University who has been involved in several well-known eyewitness identification and lineup projects. “And if we can have computers do it, that’s great, but I think [Officer Garcia] is going to need more development and testing,” he says.

Christine Mumma, executive director of the North Carolina Center on Actual Innocence, says that the version of Officer Garcia that she saw “was still too animated. To me, it took away from the seriousness of the procedure.”

Mumma prefers more traditional laptop identification software that does not use virtual humans. Such software is currently employed in many police departments.

One program Mumma has worked with, developed by IT firm SunGard, has an option for voice recognition, which allows the witness to answer verbally (as he or she could do when using Officer Garcia) rather than by using the computer’s controls, which would help people who are unfamiliar with computers. However, she says, most police departments haven’t needed to use the voice recognition feature.

Mumma is participating in studies that are currently being conducted with the laptop software to see how it compares with human administration of lineups. Mumma says the researchers are waiting until they have more data to be able to complete the study.

Ross notes that some in law enforcement say there’s a logical reason for the investigating police officers to stay involved with the witness-identification process: The theory is that “they have generated some sort of rapport with the witness, and that removing that rapport from the procedure can lead to witnesses not being comfortable in the identification procedure.”

If these software programs are refined enough to be used effectively, they may provide a way for the officer who has the rapport to be present but sufficiently removed from the process of how the lineup is conducted, eliminating the possibility that he or she would accidently influence the process.

The double-blind lineup recommendation is not without controversy, and for various reasons (including practicality), it was not explicitly recommended as part of the Department of Justice’s 1999 guide for eyewitness identification. But that guide did state that it was an area for further research and testing.

The experts interviewed for this article find there to be sufficient evidence that double-blind lineups are effective, but they also pointed out that there are other areas for improvement in lineups—for example, the choice of fillers.

Fillers are the individuals placed into the lineup with a suspect. Ross has recently been looking at how computers process facial similarities in comparison with how humans do, as a way to better figure out which filler individuals to use in a lineup.

This information “can possibly lead us to having computer programs that will help us construct better lineups,” he says. “The idea here is that fair lineups are important to have, and that they’re more diagnostic, meaning that identifications from lineups that are fair are more likely to be correct identifications of a guilty individual rather than false identifications.”

Ross is currently comparing various human-directed lineup construction techniques and seeing how they compare with a computer. He hopes to publish his findings next year.