There's a problem with photographic police lineups – they typically only include front-on photos of suspects, which may not match the angle at which a witness saw the criminal. A new system, in which witnesses can rotate the images, has been shown to be more reliable.
We've all seen TV shows in which the eyewitness to a crime views a lineup of people standing in another room, through a one-way glass window. One of those people is suspected of committing the crime, while the others are uninvolved people who simply look somewhat like that person. If the witness identifies the suspect as being the person they saw, a conviction is much more likely.
In real life, many police forces utilize facial photos instead of a live lineup. The basic idea is still the same, though, in that it is hoped the witness will select the photo of the suspect, as opposed to a photo of someone who is merely similar in appearance.
That said, an accurate ID is less likely if the angle at which the photo was taken isn't the same as the angle at which the criminal's face was seen. As a result, the witness may not recognize the person in the photo. Worse yet, they may think and say that they do, when it fact the suspect is not the person they actually saw – this means an innocent person could go to prison.
In an attempt to address this problem, scientists at Britain's University of Birmingham developed an experimental computer system in which witnesses can rotate 3D photographic models of people's heads up to 90 degrees to the left or right, stopping at any point in between.
The setup was tested on approximately 3,000 volunteer witnesses, who were initially shown a video in which a man in his thirties stole a woman's handbag. Some of the test subjects saw a version of the video in which the man's face was only seen from the left, while others saw a version in which it was only seen from the right.
The volunteers were then presented with a set of onscreen facial photos, one of which was of the man in the video. Some of the participants could only view the faces front-on, some could see oblique left- or right-side angles, while others could manually rotate the faces as they saw fit.
It was found that when it came to selecting the photo that matched the man in the video, the rate of accuracy was lowest with the front-only photos, it improved when the supplied angle of the photo approximately matched that of the viewed video, and it was highest when users could adjust the angle in order to more precisely match that of the video.
"There are good reasons to predict that accuracy will be higher if witnesses are able to manipulate images and align them to be similar to the angle of the face in their memory," says Dr. Melissa Colloff, co-author of a paper on the study. "We know that retrieving memories accurately relies on the context in which they are retrieved being similar to when the memory was formed. The experiments we carried out show this theory in practice."
The paper was recently published in the journal Scientific Reports. You can see the system in use, in the video below.
Source: University of Birmingham
