Do you never forget a face? Are you one of those people who can spot the same nameless extras across different TV programs and adverts?
Are you the family member always called on to identify or match faces in old photographs?
If so, you may be a “super-recognizer” – the term science uses to describe people with an exceptional ability to recognize faces.
Over the past decade, psychologists have established that our ability to recognize faces varies a lot – much like the ability to sing, for instance. While a small proportion of the population simply can’t hold a note at all, and most are content to confine their very average efforts to the shower, at the top end there are outstanding singers, such as Adele.
Researchers believe the same applies to facial-recognition ability. A small proportion of people struggle to recognize friends and family (a condition known as prosopagnosia), most people are “typical recognizers”, and at the top there is a small number of people who excel at recognizing faces – super-recognizers.
Since 2009, researchers have been assessing super-recognizers and their abilities. These people are usually classed as such if they reach a threshold score on the Cambridge face memory test.
The test asks participants to learn a series of faces and then recognize them from different photographs. More recently, though, the focus has shifted to ensuring that super-recognizers are defined by consistently high scores across a range of face tests, including matching pairs of unfamiliar faces – a task border-control officers perform when matching passport photos to real faces.
Credit: Josh P Davis.
Research shows three main things. One, that super-recognizers outperform most people at learning new faces and then recognizing them from headshots. Two, they are better at deciding whether two photos of unfamiliar people show the same person or two different people. And three, that the ability appears to have a genetic basis and is limited to faces.
The emergence of super-recognizers and the growing body of evidence that has identified their skills has enabled border-control agencies and police to recruit people who excel at facial-identity verification.
While research has shown that in most circumstances training and experience don’t make people better at recognizing faces, recruiting people who excel at this could significantly increase fraud detection rates at border control, where a traveler’s face doesn’t match their passport photo, for example.
It could also decrease false conviction rates from the evidence that is based on a suspect’s face being incorrectly matched to CCTV footage.
While not all super-recognizers perform perfectly all of the time or across all tests, one of the best solutions would be to pair the best among them with our best computer face-recognition algorithms to try and establish a level of best performance.
Several agencies in the UK, including London’s Metropolitan Police, have established dedicated super-recognizer units to assist in facial identification tasks. But, until now, a question remained as to whether super-recognizers would still show enhanced performance for faces that were outside their own ethnic group.
Most super-recognizer tests have used Caucasian (white) faces, and most super-recognizers who have been tested have been white. Within psychological science, it is well established that recognizing faces from another ethnic group is significantly harder than recognizing faces from your own ethnic group.
Given that border control and police officers are likely to encounter people from a large range of ethnic groups, it is important to assess whether super-recognizer ability also applies to faces from other ethnicities.
This question was tested in our recent study, where a group of white super-recognizers outperformed a white control group of typical-recognizers when asked to match pairs of Egyptian faces. The study showed that while super-recognizers showed superior performance to typical-recognizers, there was still a “cost” to their accuracy in relation to the white-face test.
When it comes to faces, most of us are typical-recognisers, with just a small percentage classed as super-recognisers.
This lends support to the idea that super-recognizers are people performing at the top end of the facial recognition scale, rather than doing something completely different to the average person.
This finding shows that deploying super-recognizers at border control or in policing should still provide a benefit, even when they are tasked with reviewing faces outside their ethnic group.
Despite the super-recognizers outperforming the control group, they were not as good at matching Egyptian faces as they were at matching white faces.
Recent research from the University of Bournemouth shows that while white super-recognizers outperform the white control-group members, they do not match native observers’ levels of accuracy.
So white super-recognizers will outperform white control-group members when asked to match Egyptian or Asian faces, for example, but native Egyptian or Asian observers will still outperform the white super-recognizers, meaning agencies seeking to recruit super-recognizers will see a performance boost.
But if they are assessing a particular ethnic group, they will need to seek the help of native observers as well.
Our work builds on this previous research, and the two combined show that the super-recognizer ability does extend to faces of other ethnicities. It provides further evidence to support the selection of super-recognizers for roles in which correctly identifying faces is critical.
Funding: Ahmed Megreya receives previous funding from Qatar National Research Fund (QNRF)
Through the University of Greenwich, Josh Davis consults for Super-Recognisers International and Yoti on the use of super-recognizers in their businesses. He has received university administered research and enterprise funding from the EC, Super-Recognisers International, Yoti, European and UK police forces and government agencies, Singapore government agencies, and Australian police forces. He does not receive any direct funding personally from these organisations.
David James Robertson does not work for, consult, own shares in or receive funding from any company or organization that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.
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.
Super‐recognizers (SRs) are individuals who are extremely proficient at processing facial identity. In the past decade, it has become clear that people vary in their proficiency on laboratory‐based tasks of facial identity processing (see, e.g., Lander, Bruce, & Bindemann, 2018 for a review).
These tests, which typically require participants to discriminate between or recognize previously unfamiliar faces, have demonstrated that face processing ability is characterized by large individual differences with some individuals attaining high performance (e.g., Bobak, Pampoulov, & Bate, 2016; Bowles et al., 2009). Moreover, such inter‐ individual differences have been linked to stable genetic factors (Shakeshaft & Plomin, 2015; Wilmer et al., 2010).
These discoveries followed decades of empirical work, showing that people in general are poor at processing facial identity of unfamiliar, compared to familiar individuals (e.g., Hancock, Bruce, & Burton, 2001).
More recently, studies with professionals trained to perform security‐critical identity verification tasks have shown that they perform no better than students on tasks that are representative of their daily work (Wirth & Carbon, 2017; White, Kemp, Jenkins, Matheson, & Burton, 2014; cf., Figure 1).
SRs have been viewed as a solution to this problem, and there is increasing interest in deploying SRs in real‐world settings that stand to benefit from their superior ability, such as policing, national security, and surveillance.
For instance, individuals selected based on their face processing abilities have been deployed within the London Metropolitan Police (Davis, Forrest, Treml, & Jansari, 2018; Davis, Lander, Evans, & Jansari, 2016; Robertson, Noyes, Dowsett, Jenkins, & Burton, 2016), as well as the Police in Cologne, Germany.1
They have been reported to have assisted investigations of several high‐profile cases, for example, Alice Gross’s murder in the United Kingdom,2 the recent Novichok poisonings in Salisbury (UK),3 and the mass assaults on women in Cologne (Germany) on New Year’s Eve 2015.4
In concert with the widespread media coverage of SRs in such operational deployments, other initiatives have emerged. The resulting rapid translation of limited scientific evidence into applied practice in this area has sometimes led to an overstatement of the benefits of deploying SRs and unsubstantiated claims.
For example, one professional agency recently claimed that ‘Super recognisers can remember 80% of faces they have seen. The average person can only remember about 20% of faces they have seen’5 and assure their staff’s high ability through ‘vigorous and continued training’.6
Another professional association7 offers membership accreditation to practice as a SR. Such claims and offers are not corroborated by the limited number of studies of SRs available to date.
These have thus far documented a 5–17% point advantage depending on the empirical test used (Davis et al., 2016; Robertson, Jenkins, & Burton, 2017). Additionally, several studies report that professionals, whose jobs require frequent image matching, are no better than inexperienced student control participants (Bruce, Bindemann, & Lander, 2018; see also Papesh, 2018; White et al., 2014). Finally, it is unclear what an accreditation to practice as an SR entails and in what capacity the associates are encouraged to operate.
Here, we argue that the current level of scientific understanding of superior face processing abilities does not yet warrant broad placement of SRs in diverse operational settings. We briefly outline the present state of scientific knowledge, before highlighting key shortcomings that limit our understanding of the potential benefit of SR deployment.
These shortcomings can be attributed to the limited number of available studies examining exclusively the SR population (Table 1) and, hence, our insufficient understanding of the functional basis of superior face processing skills.
Additionally, we currently lack a detailed understanding of the real‐world tasks that SRs are expected to perform and whether laboratory‐based tests capture the real‐world abilities of interest (see Figure 1).
We propose that solving these problems requires researchers and practitioners to approach this growing field of research in a fundamentally different way. The emergence of effective strategies for selecting and deploying individuals with superior face processing abilities requires regular communication between scientists and practitioners. Specifically, we suggest that future research in this area should incorporate a feedback loop encompassing translational and reverse‐translational research – from the lab to the world and back again (cf. Ledford, 2008). This is critical for developing robust theory that transfers to an understanding of real‐world tasks and for streamlining recruitment processes and legal guidelines to support the deployment of SRs in society.
Josh Davis, Ahmed Megreya & David James Robertson – The Conversation