Face to Face with Bias: The Inevitable Impact of Context on Identity Verification

Verifying identity through face images is vital for a variety of different tasks such as border control and identifying criminals via CCTV footage. In many cases other irrelevant information may also be presented, and this can lead to contextual bias. We present four face matching experiments that investigated the influence of fictional contextual information in the form of a prior decision about the face identities from different sources (automatic face recognition, a forensic examiner, a super recogniser and another participant) and whether it influenced face matching accuracy. Across all studies the contextual information exerted some influence with congruent feedback increasing accuracy and incongruent feedback reducing accuracy. The findings revealed some sources were more influential than others suggesting that perceived expertise of the source of the contextual information can have a greater influence as compared to other less credible sources. We suggest that when making identity verification decisions, faces should be viewed without any other contextual information to reduce contextual bias.