When a crime witness mistakenly identifies the wrong person, the mistake can destroy innocent lives and leave the real perpetrators with impunity. The task of witnesses is usually to select the faces they remember from a group of potential suspects and “fillers.” Traditionally, this involved viewing a row or series of photos standing behind one-way glass. But a new study shows that interacting with a digital 3D model (a set of virtual heads that can be manipulated with a computer mouse) can make eyewitness evidence more accurate.
“We developed a new Interactive queuing program Witnesses are allowed to rotate their faces to any desired position,” said Heather Flowe, a professor of forensic psychology at the University of Birmingham in the United Kingdom and co-author of the paper, which was published in Scientific Reports. “Yes. Something about transferring control, allowing witnesses to explore faces in their own way, which helps to remember. “
John DeCarlo, associate professor of criminal justice at the University of New Haven who was not involved in the new research, said that witness identification is widely used in prosecutions. One of the reasons for its popularity is its influence in court. “Someone is standing in the stands, looking at someone in the court, pointing to them and saying,’That’s the person.’ This is a very powerful sign for the jury,” DeCarlo explained, adding that Can make “eyewitness identification possible [seem] More accurate than reality. “
One problem is that under worrying circumstances, it is difficult for humans to form accurate memories. “Under emotional stress and environmentally unfriendly conditions, we usually see a crime happening once, very fast-which makes eyewitness identification probably the worst way of identification,” DeCarlo said. “All witnesses are prone to a relatively high rate of error or false positives.” This problem can lead to punishment for the wrong person, and the investigation can be ended without catching the offender: According to a non-profit organization Innocence plan, 69% of cases involved misidentification by witnesses Cases where people were wrongly convicted He was later proved innocent through DNA evidence.
For decades, researchers have been trying to understand which techniques can improve the accuracy of eyewitnesses, but the adjustments to the lineup procedure have only had a slight impact. The Flowe team believes that digital technology can help. “Can we introduce some relatively cheap technological advancements-now people can take high-quality images on a simple camera phone-or even render them into 3D objects with a mobile phone?” she asked. Researchers have developed their own low-cost software and shared it with other researchers for free. It converts a video clip or a few images of a human face into a 3D digital model, which is placed in an interactive lineup and can be manipulated with a mouse or (on a tablet) with a finger.
To test their digital lineup, the researchers recruited approximately 1,400 participants through online crowdsourcing. These “witnesses” watched video clips of ongoing non-violent crimes, and then spent several minutes performing distracting tasks to get their attention away from what they had just seen. Finally, they received a set of photos or a set of digital models and were asked to identify the “perpetrator”. Those who use the interactive lineup do a better job of choosing the right face.
“They have improved accuracy by 18% to 22%,” Flowe said. Compared with many other attempts to improve the suspect identification program, “this is absolutely fantastic.” Some previous methods improved accuracy, but made witnesses less confident in their choices. For example, if they are warned in advance that a lineup may be “filler” without real suspects, they are unlikely to choose anyone. But in this study, the loss of confidence did not occur, Flowe said. She pointed out that accuracy has improved for eager witnesses who are more likely to guess and conservative witnesses who make choices only when they feel certain.
What makes movable 3D digital models so effective? “We think this is by matching people to code or study the posture of the perpetrator while committing a crime-then they remember this information and look for it in the lineup to remind them of their facial memory,” Flowe said. In another set of tests also described in the new study, subjects saw a series of still photos, with the head position either the same as the main position of the offender in the video, or in a different position. When the direction matches the direction shown in the video, witnesses will be more accurate. Flowe said this “makes them more likely to correctly distinguish between guilt and innocence.”
“The research they did looks [like] It is highly effective and has a great [sample size], So it can be promoted in theory,” DeCarlo said. “I think it will definitely…let people have more discussion and research. However, he added that reality is very different from this man-made situation. Watching a video online is very different from witnessing a crime with your own eyes. Soon thereafter, performing a few minutes of distracting task is waiting for weeks of police squad. Pale imitation. But DeCarlo said that there is no way to solve this problem when testing a large group of people in various recognition scenarios. “Most eyewitness studies include [this], Does not necessarily reflect the real world,” he pointed out. “But try your best to model. “
As a next step, DeCarlo recommends to try the software in the field. For Flowe and her co-author Melissa Colloff, assistant professor of forensic psychology at the University of Birmingham, the first task is to continue experimenting and analyzing the data they collect. But they are also paying close attention to real-world testing opportunities. “Let’s see if we can make some changes, especially in the early adopted jurisdictions in the United States,” Flowe said.
Another consideration is that some police departments lack the resources to use this technology. Flowe and Colloff work in the UK, where the department often records video “face photos” of potential suspects turning their heads from side to side-this is the ideal basis for digital models. But in the United States, policy differences between police departments are even greater: some may only have static images, so it is difficult to generate detailed dynamic models. Nevertheless, Kolov suggested that it would be relatively easy for the department to start collecting more information. “Video programming groups have been implemented in the UK for some time,” she said. “so were able Done, because they are doing it here nationwide. “
Considering the operation and resource constraints of some smaller departments in the United States, DeCarlo is not so sure. Nevertheless, as Flowe pointed out, the cost of this lineup technology improvement will be much lower than the cost of false witnesses. “If it helps increase the detection of guilty suspects and reduce the misidentifications that lead to these false convictions, we can afford it,” she said. “And this has costs on many different levels-of course, socially, economically, and personally.”