Super-recognizers: From the lab to the world and back again

Abstract

The recent discovery of individuals with superior face processing ability has sparked considerable interest amongst cognitive scientists and practitioners alike. These 'Super-recognizers' (SRs) offer clues to the underlying processes responsible for high levels of face processing ability. It has been claimed that they can help make societies safer and fairer by improving accuracy of facial identity processing in real-world tasks, for example when identifying suspects from Closed Circuit Television or performing security-critical identity verification tasks. Here, we argue that the current understanding of superior face processing does not justify widespread interest in SR deployment: There are relatively few studies of SRs and no evidence that high accuracy on laboratory-based tests translates directly to operational deployment. Using simulated data, we show that modest accuracy benefits can be expected from deploying SRs on the basis of ideally calibrated laboratory tests. Attaining more substantial benefits will require greater levels of communication and collaboration between psychologists and practitioners. We propose that translational and reverse-translational approaches to knowledge development are critical to advance current understanding and to enable optimal deployment of SRs in society. Finally, we outline knowledge gaps that this approach can help address.

Keywords: face identification; face matching; face processing; face recognition; super-recognizers.

Figure 1

Super‐Recognizer identification in the lab, and potential for deployment in the real world. In laboratory settings (left box), superior face processing abilities are commonly assessed with experimental paradigms involving (top to bottom) simultaneous discrimination of pairs of stimuli (Robertson et al., 2016; Phillips et al., 2018), simultaneous one‐to‐many matching (Bruce et al., 1999), and memory paradigms designed to assess learning of facial identity using videos (left: Bobak et al., 2016, pics.stir.ac.uk) and static images (right: Russell et al., 2009). In the real‐world, SRs are selected using lab‐based tests and “on the job performance” (e.g. Davis et al., 2016), and have supported criminal investigations (Ramon, 2018a). They could be deployed in a diverse range of operational law enforcement and security settings (right box), including (top to bottom) e.g. passport control, investigative purposes (left image: West Midlands Police, https://www.flickr.com/photos/westmidlandspolice/39164763734/; right: Landespolizei Schleswig‐Holstein Filmgruppe), or crowd surveillance (left: Community Safety Glasgow; right: Landespolizei Schleswig‐Holstein Filmgruppe). [Colour figure can be viewed at wileyonlinelibrary.com]

Figure 2

Monte Carlo simulation to estimate the benefit of recruiting SRs. (1) We simulated 100 normal bivariate distributions representing the correlation between a recruitment test and a real‐world task for 1,000 ‘candidates’. The level of correlation between accuracy on the recruitment test and the real‐world task was set randomly for each simulation (.5 in this example). (2) For each of these 100 simulations, three criteria were applied to recruitment test scores in order to select face processing specialists (no selection, greater than one standard deviation above the mean, greater than two standard deviations above the mean). We then calculated the mean accuracy of these groups on the real‐world task. (3) Simulation data showing the mean real‐world accuracy of selected groups for all 100 simulations, as a function of the level of correlation between recruitment test and real‐world task. Estimated benefits of selection are signified by the difference between regression lines for selected groups (blue, red) and the non‐selected group (grey). The orange shaded area represents the ‘best‐case’ correlation between laboratory‐based tests and real‐world tasks based on existing estimates (r = .5, see text for details). At this level of correlation, benefits of selection are approximately 8% for >1SD and 12% for >2SD selection criteria. [Colour figure can be viewed at wileyonlinelibrary.com]

Figure 3

Relationship between cognition and experimental assessment of overt behaviour. (a) A cognitive process of interest, such as face cognition, can involve different subprocesses, which are ideally measured in isolation through dedicated experiments designed to this end. (b) More commonly, experiments designed to measure predominantly one subprocess through observers’ registered responses (filled box) also rely upon additional subprocesses (not filled, thick‐lined boxes), but not others (unconnected box). [Colour figure can be viewed at wileyonlinelibrary.com]

Figure 4

A framework for practice‐oriented development of performance measures. An initial analysis of the real‐world task serves to identify task constraints, practitioners’ goals, and cognitive processes (c.f. Schraagen, 2006). Researchers can then use this information to derive hypotheses about the cognitive subprocesses underlying performance and design experiments to test these hypotheses. This leads to the development of measures, which can be optimized to capture the real‐world task through additional task analyses, and the observed correspondence between accuracy in the measures and performance in on the real‐world task. This process serves to increase the predictive power of tests in terms of predicting performance in real‐world settings. [Colour figure can be viewed at wileyonlinelibrary.com]

Figure 5

Continued exchange between scientists and real‐world practitioners. This continued development cycle can serve to improve theoretical knowledge of superior face processing, which can in turn help to generate improved processes deployed in professional settings. [Colour figure can be viewed at wileyonlinelibrary.com]