Faces in the real world

Josh P. Davis of the University of Greenwich knows a thing or two about public engagement. At one point during his study into super-recognisers, people who have an excellent ability to recall faces, he had to employ someone to answer his emails. After being reported across a number of media outlets, his study attracted one million people in a year from around the world, including 26 from Antarctica. Now, even his undergraduate students are able to achieve samples of up to 14000 participants. At the start of this symposium about faces in the real word, Davis’s experience of participant recruitment was an outstanding example of why we should all attempt to engage with the media, in the drive to disseminate our findings, as well as to improve the representativeness of our samples. His work has featured on TV programmes and there is even a section of the BBC website dedicated to his projects.

People identified as super-recognisers are sought out in the police service, in order to identify offenders, often using CCTV images. However, as Allan McNeill of Glasgow Caledonian University highlighted in his talk, the criminal justice system is awash with wrongful convictions, and eyewitness misidentification accounts for a high proportion of these cases. McNeill described evidence that suggested people are able to identify faces when images have been enhanced using mapping techniques with around 70% accuracy when targets are present.  However when faces are absent from the line-up there can be up to 30% false positives. He is concerned that the court system does not question the scientific basis of these techniques, and that this will inevitably lead to further miscarriages of justice. Indeed, in another talk, Charlie Frowd from the University of Winchester gave an overview of his meta-analysis of facial composite construction which demonstrated considerable variability in the standard police systems used. Improvements were found when holistic information about faces was added, and when composites were constructed from a shorter retention time. 

While people who are very good at recognising faces might improve rates of accuracy in eyewitness testimony and be better at creating accurate facial composites, at the opposite end of the spectrum lies prosopagnosia, an inability to remember faces. Sarah Bate from Bournemouth University focused her talk on the challenges of diagnosing this condition. Many people say that they are bad at remembering faces, but of course we often have limited insight into our own abilities and so objective tests are needed.  ate argues that prosopagnosia may be over-reported and because a wealth of theoretical work is being conducted using samples who self-report the condition, there are serious implications for the field. Other issues that may affect both the recognition and judgement of faces are race, age, and context. Andrew Dunn from Nottingham Trent University manipulated a face judgement task where participants were required to judge whether a person had committed a crime. He found that participants judged black faces are more culpable of murder, rape, and theft; even though after the task they said they guessed the purpose of the study and were not racist. Participants were also told that the people they were judging either worked in a law firm or a packing company to manipulate socio-economic status (SES). Dunn found that the higher SES faces were judged as more likely to commit forgery and less likely to commit rape, theft and murder. These findings suggest that contextual information has an important influence. In real world face recognition situations there is a wealth of contextual information present, which of course could influence memory for a potential offender.

Mike Burton from the University of York drew these talks together questioning whether there was a clear direction of travel from applied work to theoretical work in this field. Decades of research has shown us the eyewitness memory is fallible, and that we can only make small improvements to memory in experimental studies. However, Burton suggests that researchers continue to prize theory over real world evidence. He suggests that we can use experiences of those working in the field to inform research and theory development rather than solely vice versa. One real world example cited was that of border control officers, who must check that photographs on passports match the person in front of them. It is also important that officers are able to sport fraudulent documents, which often have errors within the spelling of place names or gender relevant names, due to being created in different countries. Research has shown that people were more likely to say the passport was valid when the faces matched, even when the data was fraudulent. When the faces did not match, the fraud was more likely to be spotted. This has huge implications for the real world and Burton ended his talk by suggesting that researchers should ‘get out more’ and understand more of the real world applications of their theories.