Facial recognition and image comparison evidence: Identification by investigators, familiars, experts, super-recognisers and algorithms

Drawing upon decades of scientific research on face perception, recognition and comparison, this article explains why conventional legal approaches to the interpretation of images (eg from CCTV) to assist with identification are misguided. The article reviews Australian rules and jurisprudence on expert and lay opinion evidence. It also summarises relevant scientific research, including emerging research on face matching by humans (including super-recognisers) and algorithms. We then explain how legal traditions, and the interpretation of rules and procedures, have developed with limited attention to what is known about the abilities and vulnerabilities of humans, algorithms and new types of hybrid systems. Drawing upon scientific research, the article explains the need for courts to develop rules and procedures that attend to evidence of validity, reliability and performance - ie proof of actual proficiency and levels of accuracy. It also explains why we should resist the temptation to admit investigators' opinions about the identity of offenders, and why leaving images to the jury introduces unrecognised risks by virtue of the surprisingly error-prone performance of ordinary persons and the highly suggestive (or biasing) way in which comparisons are made in criminal proceedings. The article recommends using images in ways that incorporate scientific knowledge and advance fundamental criminal justice values.