Forensic Psychology Lab
People are very good at recognising the faces of people they know, even in very poor quality photographs. However, years of research now show that this ability does not extend to identifying unfamiliar faces – for example by matching a traveller to an image contained in their passport. We label this phenomenon as the 'passport problem', and have recently shown that passport issuance staff also succumb to errors on this task. When matching images of unfamiliar faces, they make the same high error rates to a group of untrained students.
Despite these poor levels of performance, there is continued reliance on photo-ID to identify unfamiliar people at national borders. Moreover, recent advances in automatic face recognition software – combined with the increased use of face images captured by CCTV and mobile devices – mean that facial image comparison plays an increasingly important role in modern forensic investigation.
Given the importance of these decisions, the focus of our work in this area is to improve human accuracy in unfamiliar face matching. So far we have made some encouraging discoveries. We observe benefits of certain forms of training, response aggregation (wisdom of the crowds), and – most promisingly – selection of experts through proficiency testing. In current work, we are evaluating performance of forensic experts in face matching and looking to optimise the benefits of training, response aggregation and personnel selection. This will provide realistic estimates of error rates that can be factored into legal decision making and identity verification processes. We also hope that it will provide important theoretical insights into cognitive processes driving visual expertise.