When a crime witness names the wrong person, it might result in the death of an innocent person while allowing the true culprit to go free.
Eyewitnesses are frequently asked to choose the face they recall from a list of probable suspects and “fillers.”
This usually entails staring at a line of individuals behind a one-way glass or a collection of pictures.
However, a new study shows that engaging with digital, three-dimensional models—a collection of virtual heads that can be moved about with a computer mouse—could improve the accuracy of eyewitness testimony.
Heather Flowe, a professor of forensic psychology at the University of Birmingham in England and a co-author of the study published in Scientific Reports, says, “We’ve devised a novel interactive lineup technique that allows witnesses to rotate the faces into any position requested.”
“There’s something about handing power over to the witness and allowing them to examine the faces in their manner that aids memory.”
According to John DeCarlo, an associate professor of criminal justice at the University of Current Haven who was not involved in the new study, eyewitness identification is routinely utilized in prosecutions.
Its influence in court is one of the reasons for its popularity.
“Someone takes the witness stand, looks around the courtroom, and declares, ‘That’s the person.’ “To a jury, that’s a pretty compelling indication,” DeCarlo says, adding that this can make “eyewitness identification [seem] more accurate than it is.”
One issue is that people have a difficult time creating accurate memories in stressful settings.
“We generally see the crime happen just once, very quickly, under emotionally stressful and ecologically unfavorable conditions—making eyewitness identification may be the worst kind of identification,” DeCarlo adds.
“There is a reasonably significant mistake rate, or false positives, in every eyewitness identification.”
Eyewitness misidentification has been implicated in 69 percent of cases in which persons were wrongly condemned and later exonerated by DNA evidence, according to the charity Innocence Project.
For decades, researchers have tried to figure out which approaches could enhance eyewitness accuracy, but changes to lineup procedures have had just a little influence.
The team at Flowe believed that digital technology could assist.
“Can we bring in some of the very affordable technical advances—with people’s capacity to shoot high-quality photos on basic camera phones—and even turn them into 3-D things using their phones?” she inquires.
The researchers created their low-cost software and freely distributed it to other researchers.
It converts a video clip or multiple pictures of a face into a 3-D digital model that can be edited with a mouse or (on a tablet) a finger and put in an interactive lineup.
The researchers used internet crowdsourcing to enlist around 1,400 participants to evaluate their digital lineup.
These “witnesses” were given a video clip of a nonviolent crime being committed and then forced to complete diverting tasks for up to a minute to keep their minds off what they had just seen.
Finally, they were given either a collection of photos or a set of digital models to identify the “perpetrator.”
Those who utilized the interactive lineup were considerably more successful in selecting the proper face.
“They’re 18 to 22 percent more accurate,” adds Flowe.
Compared to many previous initiatives to enhance suspect-identification methods, “that is fantastic.”
Some earlier methods improved accuracy while making witnesses less confident in their decisions.
They were less likely to pick anybody at all if they were informed ahead of time that a lineup may be all “filler” and not contain the genuine culprit.
This research, however, did not show a decrease in confidence, according to Flowe.
She points out that the accuracy improved for both the enthusiastic witnesses who were more inclined to guess and the cautious witnesses who only chose when they were certain.
What was it about the movable 3-D digital models that made them so effective? “We believe that when individuals match the position in which they encoded or examined the offender at the moment of the incident, they remember that information and seek it out in the lineup to stimulate their memory for the face,” Flowe adds.
Subjects were shown a lineup of still pictures with their heads either in the same posture as the culprit in the video or a different position in another series of tests detailed in the current study.
When the orientation of the witnesses matched that of the film, they were more accurate.
This, according to Flowe, “increases the likelihood that they will be able to identify the guilty from the innocent.”
“The research they conducted appeared to have a good level of validity and a large [sample size], so it was theoretically generalizable,” DeCarlo adds.
“I believe it will surely… provide people with more topics to discuss and research.”
However, he emphasizes that reality is far different from this type of fictitious circumstance.
Watching a video on the internet is not the same as witnessing a crime in person.
And doing something else for a few minutes later is a poor substitute for sitting in a police lineup for weeks.
When testing multiple identifying situations on a big population, though, DeCarlo believes there is no way around it.
“Most eyewitness study, including [this],” he points out, “does not always reflect the real world.”
“However, it makes every effort to imitate it.”
DeCarlo recommends testing the program in the field as a next step.
The objective for Flowe and her co-author Melissa Colloff, an assistant professor of forensic psychology at the University of Birmingham, is to keep testing and evaluating the data.
They are, however, on the lookout for opportunities to conduct real-world testing.
“Let’s explore what we can accomplish, especially in those early-adopter areas in the United States,” Flowe adds.
Another factor to consider is that some police departments do not have the necessary resources to implement such technology.
Flowe and Colloff work in the United Kingdom, where police departments routinely take video “mug shots” of potential suspects turning their heads from side to side, providing an ideal foundation for a digital model.
However, policy varies more among police departments in the United States, with some having only still images, making it difficult to create a detailed dynamic model.
Colloff believes it would be relatively simple for departments to begin collecting more data.
“Video lineups have been used in the United Kingdom for a long time,” she says. “They do it on a national scale here, so it can be done.”
Given the operational and resource constraints in some smaller US departments, DeCarlo is less confident.
However, as Flowe points out, the cost of such advancements in lineup technology would be a fraction of the cost of erroneous eyewitnesses.
“If it has the benefit of increasing the detection of guilty suspects and decreasing the erroneous identifications that lead to these wrongful convictions,” she says, “we could afford it.” “And that comes at a price on so many levels—societally, economically, and, of course, personally.”
THE AUTHOR IS
Sophie Bushwick is an associate editor covering technology at Scientific American.