- Josh Davis
- 15 minutes ago
- 10 min read
Professor Josh P Davis
BSc MSc PhD AFBPsS FHEA CPsychol
Face and Voice Recognition Lab
School of Human Sciences
Institute of Lifecourse Development
University of Greenwich
London SE10 9LS
12 September 2025
Hertfordshire Police, Greenwich University, and Super-Recognisers International
You can download the full blog as a PDF using the attachment above, for easy access and future reference.
A brief message from Professor Josh P Davis: For various reasons, I took some time out from work-related social media activities, and I have not posted any blogs on this site for nearly a year. However, I am back, refreshed, and very much ready for action; and once again, me and my team are actively seeking new projects and collaborations. But also, watch this space, we have a large backlog of exciting news to report – all of which will be posted here on the website (www.superrecognisers.com). I apologise to anyone who has tried to contact me. I believe that our fully updated systems and protocols will make it far easier for me to manage the eagerly anticipated increase in the team’s workload.
Because of this lack of activity, this blog is longer than most.
Hertfordshire Police Project (organised by Super-Recognisers International)
I hope that Hertfordshire Police will not mind me re-posting this selfie which was taken and uploaded to LinkedIn by Chief Constable Andy Prophet on 3 September 2025. It depicts Andy and a group of Hertfordshire Police officers and civilian staff who were just about to take the final examination phase (“Phase 3”) of the University of Greenwich Super-Recognition Tests. Because of the exams, it is not surprising perhaps, if some look just a little apprehensive, even if all had already taken the nine Phase 1 and Phase 2 tests (April-May, 2025), to achieve the required standard to be invited to attend the examination session.
Presentation: On the other hand, I (Professor Josh Davis, and also depicted in the selfie) had minutes before delivered a presentation describing research showing that super-recognisers are superior to just about everyone else; at detecting AI-produced artificial faces (Reid et al., currently unpublished – this is a pre-print, so please appraise these results with caution); morphed and other manipulated facial images (i.e. InstaBeauty and face swaps) (Davis et al., 2025); and hyper-realistic silicone masks (Robertson et al., 2024); and they are also significantly better at identifying familiar and unfamiliar faces wearing face masks (Noyes et al, 2021; Ritchie et al., 2024), dark glasses (Noyes et al., 2024), balaclavas and hats (Davis & Tamonytė, 2017); while evidence suggests that White super-recognisers outperform White non-super-recognisers at the identification of infants (Belanova et al., 2018). Black faces (Correll et al., 2022), and those from the middle east (Robertson et al., 2020). The worried faces of the super-recognisers might therefore actually also be due to the heavy expectations of future successes from their other senior management (also present, but not all in the photo).
Hertfordshire Police successes: At the time we did not know this, but I am pleased to report that all the Hertfordshire Police officers and staff who attended the examination session achieved our criteria for either super-recogniser or super-matcher, or both (see glossary at the bottom of this blog). What encouraged me most, was that their chief constable made it very clear that he is determined to capitalise on the skills of these highly talented staff, with the aim of increasing identification rates of suspects caught on CCTV etc.
Super-Recognisers International (SRI): These testing sessions were organised by SRI whose CEO, Mike Neville (ex-Detective Chief Inspector and Instigator of the New Scotland Yard Super-Recogniser Unit for London’s Metropolitan Police Service) as part of an excellent four-day training event, which included the Phase 3 Greenwich super-recogniser examinations; sessions by super-recognisers, sessions on the legal requirements associated with identifying suspects from visual images; and to cap off the week, a session involving the identification of real people in a shopping centre from old photographs.
Could you be a super-recogniser?
· It is possible for anyone to take the Greenwich Super-Recognition Test Battery by contacting SRI (https://superrecognisersinternational.com/) directly. Face-to-face or remotely invigilated online examinations can be arranged for police forces or businesses (remote testing is advised if not in the UK).
· High scorers (the key eligibility criterion) who take the University of Greenwich tests for members of the public; and who sign up to join the 49,000-strong Greenwich Face and Voice Recognition Lab’s Research Participant Volunteer Pool, may also receive an invite to take the tests for Super-Recognisers International. So far, more than 25 journal articles that involved participation from members of the volunteer pool have been published. We recommend anyone wishing to join the volunteer pool to take the pop-up test here (www.superrecognisers.com). Instructions will follow. We recommend this route for individuals who have no knowledge of their level of skill.
Who has taken the battery of 12 University of Greenwich Super-Recognition Tests? More than 50,000 participants have taken the gradually evolving battery of super-recogniser tests since 2013. Most participants have been police officers from more than 30 police forces whose management teams have agreed research consultancy contracts with the University of Greenwich (Davis et al., in preparation).
We have also separately assisted some businesses to recruit super-recognisers. We created special tests for these projects often using images provided to the Greenwich Self-Photo and -Video Donated Stimuli Databases. Quick and easy access to the Greenwich Volunteer Pool, whose members come from 170 countries has also helped. Fewer than 10% are UK-based, meaning we should be able to help organisations recruit suitable staff in most countries.
The Big Picture: AI-Face Recognition and Artificial Face Detection
What about AI-face recognition and fraud detection systems? I often get asked if AI is better than super-recognisers at these tasks. Yes and no is the easy answer, although of course, AI works 24-hours-a-day, 365-days-a-year, so in some contexts, there is no contest. Nevertheless, relevant research demonstrates that when both AI and superior face recognisers work together on difficult decisions of facial identity, accuracy is significantly superior to when super-recognisers or AI is relied upon alone (e.g., Ibsen et al., 2024; Philipps et al., 2018).
Training humans? On the other hand, all research has shown that humans with typical levels of ability are significantly worse at these tasks than AI (e.g., Ibsen et al., 2024; Ritchie et al., 2024), just like they are significantly worse than super-recognisers (e.g., Davis et al., 2025). Training can help, but only slightly (see Alice Towler’s articles for a series of studies examining training of face identification tasks: https://www.alicetowler.com/publications). Super-recognisers will also improve with training and may maintain their level of superiority (Davis et al., in preparation).
The proper question to ask? I would also argue that the question about which system is best (super-recognisers or AI?), is not necessarily the right question to ask. Even if a business or police force thinks they are relying on AI alone; a human being is always required to make the final legal decisions about identity, and indeed, about what steps should next be taken if the system flags up an alert. Years of experience in the facial analyst workplace (i.e. passport officials) and/or specialist training regimes will not maketh a super-recogniser. Full stop. And if staff making these decisions have never been tested, can those organisations be 100% sure that none of their team are prosopagnosic (face blind)?
Forensic scientists: Forensic facial experts may apply their techniques to generate highly reliable judgments of identity from facial images. They are also significantly superior to typical-ability humans (Noyes et al., 2025). However, the application of these methods can sometimes take days. Super-recognisers tend to make significantly quicker and more accurate facial identity decisions than everyone else, albeit there may be some exceptional tasks (Davis et al., 2025), which require an extra second or two of decision-making time. Most importantly, super-recognisers are significantly superior at the recognition of previously seen faces, or the perceptual comparison/matching of faces; but also, at recognising that they have never seen a face before or that two facial images depict two different people (Davis et al., 2016; Davis et al., 2015). This should minimise the number of false leads in the policing and security workplaces. As these decisions tend to be made in seconds, the financial benefits in terms of staffing costs should be very obvious.
Organisations who do not test the face recognition ability of their staff: I am therefore incredulous when “naïve” organisations (my personal opinion only) claim that they do not need the human in the loop and that AI will solve all their problems when it comes to faces. I immediately argue back that the human will already be a vital component of that loop; and if I was acting as their consultant, I would suggest that their finance department starts budgeting for increased spending on public relations, as will be inevitable when people are wrongly identified (and they find out about it); or when systems miss the real fraudsters.
Brief glossary
1. Super-Recogniser: Exceptional skills at short-term face recognition; long-term face recognition; simultaneous face matching; and spotting faces in a crowd; all based on consistently high scores on 12 tests of face recognition.
2. Super-Matcher: Exceptional skills at simultaneous face matching based on the tests in the battery that measure that skill. This perceptual skill is possibly the commonly used component by police, for instance, when verifying the identity of someone from photo-ID. When faces are unfamiliar, most people are far worse at this task than they probably imagine (Davis & Valentine, 2009; 2015).
3. Super-recognition and super-matcher criteria: At least 100 (sometimes more than 1,000) super-recognisers (based on test scores achievable by the top 2% of the population on at least three published tests) pilot tested each test to help us set the appropriate standards. Those achieving this standard on all 12 tests are extremely rare. Indeed, anyone achieving the standard on 50% of the tests is virtually certain to possess a level of face recognition ability that any lab in the world would define as being of super-recogniser standard.
Chloe Dejonge: Thanks most definitely go out to Chloe Dejonge (1st class BSc Psychology graduate, University of Greenwich, 2025), who is “looking for career-enhancing opportunities” while taking some time out before starting an MSc. Chloe (also depicted in the photo) provided great assistance with the invigilation and the statistical analyses of the results of the examinations at Hertfordshire Police Training Centre in Hatfield.
Viktorija Peresta: Thanks also go to Viktorija who has now left the Greenwich Face and Voice Recognition Lab’s Team. Viktorija managed most of the preparation for the Hertfordshire Police, and indeed, most of our other recent projects.
Chloe and Josh outside Hatfield Police Station and Training Centre
List of References
· Articles marked with a star involved contributions from members of the Greenwich Face and Voice Recognition Lab’s Volunteer Research Participant Pool.
· The stimuli used in articles marked with an X were kindly donated by Volunteer Participant Pool members and others uploading images to our Greenwich Self-Photo and -Video Donated Stimuli Database system (more details here: https://www.superrecognisers.com/post/donate-photos-5-for-8-of-you-at-different-ages-easy-upload-system)
· Names in bold were/are staff or students from the University of Greenwich.
Belanova, E., Davis, J. P., & Thompson, T. (2018). Cognitive and neural markers of super-recognisers' face processing superiority and enhanced cross-age effect. Cortex, 98, 91-101. https://doi.org/10.1016/j.cortex.2018.07.008 (Free to download pre-print: https://bit.ly/bdtb2018).
Correll, J., Ma., D. S., & Davis, J. P. (2021). Perceptual tuning through contact? Contact interacts with perceptual (not memory-based) face-processing ability to predict cross-race recognition. Journal of Experimental Social Psychology, 92, 104058.
Davis, J. P., Robertson, D. J., Jenkins, R. E., Ibsen, M., Nichols, R., Babbs, M., Rathgeb, C., Løvåsdal, F., Raja, K., & Busch, C. (2025). The super-recogniser advantage extends to the detection of digitally manipulated faces. Applied Cognitive Psychology, 39(2), e70053. https://doi.org/10.1002/acp.70053 *
Davis, J. P., & Tamonytė, D. (2017). Masters of disguise: Super-recognisers’ superior memory for concealed unfamiliar faces. Proceedings of the 2017 Seventh International Conference on Emerging Security Technologies (EST), 6-8 September 2017, Canterbury, UK. https://doi.org/10.1109/EST.2017.8090397 (Free to download pre-print: https://bit.ly/dtb2017). *
Davis, J. P., & Valentine, T. (2009). CCTV on trial: Matching video images with the defendant in the dock. Applied Cognitive Psychology, 23, 482-505.
https://doi.org/10.1002/acp.1490 (Free to download pre-print:
Davis, J. P., & Valentine, T. (2015). Human verification of identity from photographic images. In T. Valentine and J. P. Davis (Eds.), Forensic Facial Identification: Theory and Practice of Identification from Eyewitnesses, Composites and CCTV (pp. 211-238). Chichester: Wiley-Blackwell. https://doi.org/10.1002/9781118469538.ch9 (Free to download pre-print: https://bit.ly/djpv2015).
Gray, K. L. H., Davis, J. P., Bunce, C., Noyes, E., & Ritchie, K. L. (under second peer review). Training super-recognisers’ detection and discrimination of computer-generated faces. (Free to download pre-print: https://doi.org/10.31234/osf.io/5jqh8_v2). *
Ibsen, M., Nichols, R., Rathgeb, C., Robertson, D. J., Davis, J. P., Løvåsdal, F., Raja, K., Jenkins, R., & Busch, C (2024). Application-oriented face manipulation detection: combining algorithm and human examiner decisions. In Proceedings of the 2024 ACM Workshop on Information Hiding and Multimedia Security (IH&MMSec’24), 41-46. https://doi.org/10.1145/3658664.3659649 *
Noyes, E., Davis, J. P., Petrov, N., Gray, K. L. H., Ritchie, K. (2021). The effect of face masks and sunglasses on identity and expression recognition with super-recognizers and typical observers. Royal Society Open Science, 8, 201169.
Noyes, E. Moreton, R., Hancock, P. J. B., Ritchie, K., Martinez, S. C., Gray, K., & Davis, J. P. (2023). A forensic facial examiner and professional team advantage for masked face identification. Applied Cognitive Psychology, 39(4), e70092. https://doi.org/10.1002/acp.70092 * x
Phillips, P. J., Yates, A. N., Hu, Y., Hahn, C. A., Noyes, E., Jackson, K., ... & O’Toole, A. J. (2018). Face recognition accuracy of forensic examiners, superrecognizers, and face recognition algorithms. Proceedings of the National Academy of Sciences, 115(24), 6171-6176. https://doi.org/10.1073/pnas.1721355115
Ritchie, K., Carragher, D., Davis, J. P., Read, K., Jenkins, R., Noyes, E., Gray, K., Hancock, P. (2024). Face masks and fake masks: The effect of real and superimposed masks on face matching with superrecognisers, typical observers, and algorithms. Cognitive Research: Principles and Implications, 9, 5. https://doi.org/10.1186/s41235-024-00532-2 * x
Robertson, D., Black, J., Chamberlain, B., Megreya, A. M., & Davis, J. P. (2020). Super-recognisers show an advantage for other race face identification. Applied Cognitive Psychology, 34(1), 205-216. https://doi.org/10.1002/acp.3608 *
Robertson, D. J., Davis, J. P., Sanders, J., & Towler, A. (2024). The super-recogniser advantage extends to the detection of hyper-realistic face masks. Applied Cognitive Psychology, 38(4), e4222. https://doi.org/10.1002/acp.4222 *
Alternative sources of information
This site includes a full list of downloadable published articles and blogs is available here: www.superrecognisers.com (private website)
University of Greenwich webpage: https://www.gre.ac.uk/people/rep/faculty-of-education-and-health/josh-davis
Research Gate: https://www.researchgate.net/profile/Josh-Davis-9
You can download the full blog as a PDF using the attachment above, for easy access and future reference.
Comments