Does filler database size influence identification accuracy?
Amanda N. Bergold,Paul Heaton
DOI: https://doi.org/10.1037/lhb0000289
2018-06-01
Law and Human Behavior
Abstract:Police departments increasingly use large photo databases to select lineup fillers using facial recognition software, but this technological shift's implications have been largely unexplored in eyewitness research. Database use, particularly if coupled with facial matching software, could enable lineup constructors to increase filler-suspect similarity and thus enhance eyewitness accuracy (Fitzgerald, Oriet, Price, & Charman, 2013). However, with a large pool of potential fillers, such technologies might theoretically produce lineup fillers too similar to the suspect (Fitzgerald, Oriet, & Price, 2015; Luus & Wells, 1991; Wells, Rydell, & Seelau, 1993). This research proposes a new factor-filler database size-as a lineup feature affecting eyewitness accuracy. In a facial recognition experiment, we select lineup fillers in a legally realistic manner using facial matching software applied to filler databases of 5,000, 25,000, and 125,000 photos, and find that larger databases are associated with a higher objective similarity rating between suspects and fillers and lower overall identification accuracy. In target present lineups, witnesses viewing lineups created from the larger databases were less likely to make correct identifications and more likely to select known innocent fillers. When the target was absent, database size was associated with a lower rate of correct rejections and a higher rate of filler identifications. Higher algorithmic similarity ratings were also associated with decreases in eyewitness identification accuracy. The results suggest that using facial matching software to select fillers from large photograph databases may reduce identification accuracy, and provides support for filler database size as a meaningful system variable. (PsycINFO Database Record
law,psychology, social