beSure®: The Only Existing Scientific Assessment Tool for Super-Recognizer Identification Using Authentic Police Material

<p>Super-recognizers (SRs) are individuals with exceptional facial identity processing (FIP) abilities. Law enforcement agencies are increasingly interested in identifying SRs, but existing lab-based selection procedures do not reflect practitioners' challenges. Overcoming this limitation via a unique participatory action research collaboration between law enforcement and academia, we developed the Berlin Test for Super-Recognizer Identification (beSure®). This bespoke tool enables assessment of individuals' FIP ability across challenging and professionally pertinent scenarios and thus the identification of Police Super-Recognizers (Police SRs). beSure® features two critical components that all previous SR diagnostic approaches have lacked and ensure its operational relevance for law enforcement: First, authentic police material is used as stimuli. These are embedded across a number of subtests, which were designed based on the outcome of extensive expert interviews conducted throughout the entire Berlin Police, and involve increasing task difficulty (from 1:1 mugshot image matching to 1:n crowd search in video sequences). Second, across its image matching subtests, distractor stimuli were selected based on the false positives of an open source automatic face recognition algorithm. Thus, beSure® is complementary by design: it identifies Police SRs as those individuals who do not exhibit the same mistakes as an algorithm would have. Here, we describe beSure®'s development, the data obtained from among the entire cohort of the Berlin Police's 18K officers, and — for the first time — the details of the Police SR selection process developed thereupon. This yielded 22 Police SRs in the initial rollout and a further 26 in a follow-up assessment: officers whose accuracy remained high precisely where the material was least constrained. By establishing a scientifically grounded framework, and transparently documenting its methods, this work represents a benchmark for future research and practice in the identification of SRs within law enforcement contexts.</p>