Forensic confirmation bias: the case of facial image comparison

Abstract

Facial image comparison is the process of comparing a face with another (referred to as one-to-one facial image comparison or facial examination) or many others (one-to-many facial image comparison or facial review). As such it encompasses much of the basic visual cognition present in other forensic disciplines. The discipline has grown considerably since the introduction of facial recognition technology in the last decade (Spaun, 2007). Such technology is currently being used widely in the government and private sectors, including law enforcement (Phillips, 2011). Much like automated fingerprint identification systems (AFIS), the current generation of identification facial recognition systems require a human in the loop (Spaun, 2009). Research conducted over the past decade has consistently found that, although significant, the role of the human is often underestimated by organisations implementing facial recognition technology (O’Toole, Jiang, Roark, & Abdi, 2006; Spaun, 2009). This is not surprising, given that the importance of understanding the human role has often been overshadowed by the promises of the technology. As a consequence, the requirements for human-centered system design and training have often been neglected.