Diverse types of expertise in facial recognition.

Facial recognition errors can jeopardize national security, criminal justice, public safety and civil rights. Here, we compare the most accurate humans and facial recognition technology in a detailed lab-based evaluation and international proficiency test for forensic scientists involving 27 forensic departments from 14 countries. We find striking cognitive and perceptual diversity between naturally skilled super-recognizers, trained forensic examiners and deep neural networks, despite them achieving equivalent accuracy. Clear differences emerged in super-recognizers' and forensic examiners' perceptual processing, errors, and response patterns: super-recognizers were fast, biased to respond 'same person' and misidentified people with extreme confidence, whereas forensic examiners were slow, unbiased and strategically avoided misidentification errors. Further, these human experts and deep neural networks disagreed on the similarity of faces, pointing to differences in their representations of faces. Our findings therefore reveal multiple types of facial recognition expertise, with each type lending itself to particular facial recognition roles in operational settings. Finally, we show that harnessing the diversity between individual experts provides a robust method of maximizing facial recognition accuracy. This can be achieved either via collaboration between experts in forensic laboratories, or most promisingly, by statistical fusion of match scores provided by different types of expert.

Corrigendum to "Self-reported face recognition abilities moderately predict face-learning skills: Evidence from Italian samples" [Heliyon 9 (3) (March 2023) Article e14125].

[This corrects the article DOI: 10.1016/j.heliyon.2023.e14125.].

Selecting police super-recognisers.

People vary in their ability to recognise faces. These individual differences are consistent over time, heritable and associated with brain anatomy. This implies that face identity processing can be improved in applied settings by selecting high performers-'super-recognisers' (SRs)-but these selection processes are rarely available for scientific scrutiny. Here we report an 'end-to-end' selection process used to establish an SR 'unit' in a large police force. Australian police officers (n = 1600) completed 3 standardised face identification tests and we recruited 38 SRs from this cohort to complete 10 follow-up tests. As a group, SRs were 20% better than controls in lab-based tests of face memory and matching, and equalled or surpassed accuracy of forensic specialists that currently perform face identification tasks for police. Individually, SR accuracy was variable but this problem was mitigated by adopting strict selection criteria. SRs' superior abilities transferred only partially to body identity decisions where the face was not visible, and they were no better than controls at deciding which visual scene that faces had initially been encountered in. Notwithstanding these important qualifications, we conclude that super-recognisers are an effective solution to improving face identity processing in applied settings.

Self-reported face recognition abilities moderately predict face-learning skills: Evidence from Italian samples.

Face Recognition Ability (FRA) varies widely throughout the population. Previous research highlights a positive relationship between self-perceived and objectively measured FRA in the healthy population, suggesting that people do have insight into their FRA. Given that this relationship has not been investigated in Italian samples yet, the main aim of the present work was to develop an Italian translation of the Prosopagnosia Index-20 (PI-20), a self-report measure of FRA, to investigate the relationship between PI-20 performances and an objective assessment given by the Cambridge Face Memory Test Long Form (CFMT+) in the Italian population. A sample of 553 participants filled in the PI-20 Italian version 1 or 2 (PI-20_GE or PI-20_BA) and completed the CFMT+. Results showed a negative correlation between both versions of the Italian PI-20 and CFMT+ scores, meaning that the more self-evaluations were negative, the worse they objectively performed. The same results applied to the extreme limits of the distribution (i.e., 10% of the highest and lowest PI-20 scores). Furthermore, both age and administration order of the tests were predictor variables of CFMT+ scores. Overall, our results suggest that people possess insight, although relatively limited, into their FRA.

Verifying unfamiliar identities: Effects of processing name and face information in the same identity-matching task.

Matching the identity of unfamiliar faces is important in applied identity verification tasks, for example when verifying photo ID at border crossings, in secure access areas, or when issuing identity credentials. In these settings, other biographical details-such as name or date of birth on an identity document-are also often compared to existing records, but the impact of these concurrent checks on decisions has not been examined. Here, we asked participants to sequentially compare name, then face information between an ID card and digital records to detect errors. Across four experiments (combined n = 274), despite being told that mismatches between written name pairs and face image pairs were independent, participants were more likely to say that face images matched when names also matched. Across all experiments, we found that this bias was unaffected by the image quality, suggesting that the source of the bias is somewhat independent of perceptual processes. In a final experiment, we show that this decisional bias was found only for name checks, but not when participants were asked to check ID card expiration dates or unrelated object names. We conclude that the bias arises from processing identity information and propose that it operates at the level of unfamiliar person identity representations. Results are interpreted in the context of theoretical models of face processing, and we discuss applied implications.

Face-Information Sampling in Super-Recognizers.

Perceptual processes underlying individual differences in face-recognition ability remain poorly understood. We compared visual sampling of 37 adult super-recognizers-individuals with superior face-recognition ability-with that of 68 typical adult viewers by measuring gaze position as they learned and recognized unfamiliar faces. In both phases, participants viewed faces through "spotlight" apertures that varied in size, with face information restricted in real time around their point of fixation. We found higher accuracy in super-recognizers at all aperture sizes-showing that their superiority does not rely on global sampling of face information but is also evident when they are forced to adopt piecemeal sampling. Additionally, super-recognizers made more fixations, focused less on eye region, and distributed their gaze more than typical viewers. These differences were most apparent when learning faces and were consistent with trends we observed across the broader ability spectrum, suggesting that they are reflective of factors that vary dimensionally in the broader population.

The low prevalence effect in fingerprint comparison amongst forensic science trainees and novices.

The low prevalence effect is a phenomenon whereby target prevalence affects performance in visual search (e.g., baggage screening) and comparison (e.g., fingerprint examination) tasks, such that people more often fail to detect infrequent target stimuli. For example, when exposed to higher base-rates of 'matching' (i.e., from the same person) than 'non-matching' (i.e., from different people) fingerprint pairs, people more often misjudge 'non-matching' pairs as 'matches'-an error that can falsely implicate an innocent person for a crime they did not commit. In this paper, we investigated whether forensic science training may mitigate the low prevalence effect in fingerprint comparison. Forensic science trainees (n = 111) and untrained novices (n = 114) judged 100 fingerprint pairs as 'matches' or 'non-matches' where the matching pair occurrence was either high (90%) or equal (50%). Some participants were also asked to use a novel feature-comparison strategy as a potential attenuation technique for the low prevalence effect. Regardless of strategy, both trainees and novices were susceptible to the effect, such that they more often misjudged non-matching pairs as matches when non-matches were rare. These results support the robust nature of the low prevalence effect in visual comparison and have important applied implications for forensic decision-making in the criminal justice system.

Statistical feature training improves fingerprint-matching accuracy in novices and professional fingerprint examiners.

Forensic science practitioners compare visual evidence samples (e.g. fingerprints) and decide if they originate from the same person or different people (i.e. fingerprint 'matching'). These tasks are perceptually and cognitively complex-even practising professionals can make errors-and what limited research exists suggests that existing professional training is ineffective. This paper presents three experiments that demonstrate the benefit of perceptual training derived from mathematical theories that suggest statistically rare features have diagnostic utility in visual comparison tasks. Across three studies (N = 551), we demonstrate that a brief module training participants to focus on statistically rare fingerprint features improves fingerprint-matching performance in both novices and experienced fingerprint examiners. These results have applied importance for improving the professional performance of practising fingerprint examiners, and even other domains where this technique may also be helpful (e.g. radiology or banknote security).

Match me if you can: Evidence for a domain-general visual comparison ability.

Visual comparison-comparing visual stimuli (e.g., fingerprints) side by side and determining whether they originate from the same or different source (i.e., "match")-is a complex discrimination task involving many cognitive and perceptual processes. Despite the real-world consequences of this task, which is often conducted by forensic scientists, little is understood about the psychological processes underpinning this ability. There are substantial individual differences in visual comparison accuracy amongst both professionals and novices. The source of this variation is unknown, but may reflect a domain-general and naturally varying perceptual ability. Here, we investigate this by comparing individual differences (N = 248 across two studies) in four visual comparison domains: faces, fingerprints, firearms, and artificial prints. Accuracy on all comparison tasks was significantly correlated and accounted for a substantial portion of variance (e.g., 42% in Exp. 1) in performance across all tasks. Importantly, this relationship cannot be attributed to participants' intrinsic motivation or skill in other visual-perceptual tasks (visual search and visual statistical learning). This paper provides novel evidence of a reliable, domain-general visual comparison ability.

UNSW Face Test: A screening tool for super-recognizers.

We present a new test-the UNSW Face Test (www.unswfacetest.com)-that has been specifically designed to screen for super-recognizers in large online cohorts and is available free for scientific use. Super-recognizers are people that demonstrate sustained performance in the very top percentiles in tests of face identification ability. Because they represent a small proportion of the population, screening large online cohorts is an important step in their initial recruitment, before confirmatory testing via standardized measures and more detailed cognitive testing. We provide normative data on the UNSW Face Test from 3 cohorts tested via the internet (combined n = 23,902) and 2 cohorts tested in our lab (combined n = 182). The UNSW Face Test: (i) captures both identification memory and perceptual matching, as confirmed by correlations with existing tests of these abilities; (ii) captures face-specific perceptual and memorial abilities, as confirmed by non-significant correlations with non-face object processing tasks; (iii) enables researchers to apply stricter selection criteria than other available tests, which boosts the average accuracy of the individuals selected in subsequent testing. Together, these properties make the test uniquely suited to screening for super-recognizers in large online cohorts.

Familiarity does not inhibit image-specific encoding of faces.

When matching and recognizing familiar faces, performance is unaffected by changes to image-specific details such as lighting, head angle, and expression. In contrast, these changes have a substantial impact on performance when faces are unfamiliar. What process can account for this difference? Recent evidence shows a memory disadvantage for remembering specific images of familiar people compared to unfamiliar people, suggesting that image invariance in familiar face processing may be supported by loss of image-specific details in memory. Here, we examine whether this cost results from loss of image specific details during encoding of familiar faces. Participants completed four tasks that required participants to retain image-specific information in working memory: duplicate detection (Experiment 1), change detection (Experiment 2), short-term recognition memory (Experiment 3 and 5), and visual search (Experiment 4). Across all experiments (combined n = 270), our results consistently show equivalent memory performance for specific images of familiar and unfamiliar faces. We conclude that familiarity does not influence encoding of pictorial details, suggesting that loss of image-specificity reported in previous work is a result of longer-term storage mechanisms. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Do professional facial image comparison training courses work?

Facial image comparison practitioners compare images of unfamiliar faces and decide whether or not they show the same person. Given the importance of these decisions for national security and criminal investigations, practitioners attend training courses to improve their face identification ability. However, these courses have not been empirically validated so it is unknown if they improve accuracy. Here, we review the content of eleven professional training courses offered to staff at national security, police, intelligence, passport issuance, immigration and border control agencies around the world. All reviewed courses include basic training in facial anatomy and prescribe facial feature (or 'morphological') comparison. Next, we evaluate the effectiveness of four representative courses by comparing face identification accuracy before and after training in novices (n = 152) and practitioners (n = 236). We find very strong evidence that short (1-hour and half-day) professional training courses do not improve identification accuracy, despite 93% of trainees believing their performance had improved. We find some evidence of improvement in a 3-day training course designed to introduce trainees to the unique feature-by-feature comparison strategy used by facial examiners in forensic settings. However, observed improvements are small, inconsistent across tests, and training did not produce the qualitative changes associated with examiners' expertise. Future research should test the benefits of longer examination-focussed training courses and incorporate longitudinal approaches to track improvements caused by mentoring and deliberate practice. In the absence of evidence that training is effective, we advise agencies to explore alternative evidence-based strategies for improving the accuracy of face identification decisions.

Error Rates in Users of Automatic Face Recognition Software.

In recent years, wide deployment of automatic face recognition systems has been accompanied by substantial gains in algorithm performance. However, benchmarking tests designed to evaluate these systems do not account for the errors of human operators, who are often an integral part of face recognition solutions in forensic and security settings. This causes a mismatch between evaluation tests and operational accuracy. We address this by measuring user performance in a face recognition system used to screen passport applications for identity fraud. Experiment 1 measured target detection accuracy in algorithm-generated 'candidate lists' selected from a large database of passport images. Accuracy was notably poorer than in previous studies of unfamiliar face matching: participants made over 50% errors for adult target faces, and over 60% when matching images of children. Experiment 2 then compared performance of student participants to trained passport officers-who use the system in their daily work-and found equivalent performance in these groups. Encouragingly, a group of highly trained and experienced "facial examiners" outperformed these groups by 20 percentage points. We conclude that human performance curtails accuracy of face recognition systems-potentially reducing benchmark estimates by 50% in operational settings. Mere practise does not attenuate these limits, but superior performance of trained examiners suggests that recruitment and selection of human operators, in combination with effective training and mentorship, can improve the operational accuracy of face recognition systems.