Accuracy and reproducibility of latent print decisions on comparisons from searches of an automated fingerprint identification system

Abstract

This study was conducted to evaluate the accuracy and reproducibility of decisions made by practicing latent print examiners (LPEs) when comparing latent fingerprints to exemplars that were acquired by searches of the FBI Next Generation Identification system (NGI). This study builds on the 2009 FBI-Noblis latent print examiner black box study (1, 2), incorporating lessons learned from conducting multiple additional forensic examiner studies. Analyses are based on 14,224 responses from 156 LPEs. Each participant was assigned 100 latent-exemplar image pairs (IPs, 80 nonmated and 20 mated) out of a total of 300 IPs. On mated comparisons, 62.6 % of responses were IDs (true positives), 4.2 % were erroneous exclusions (false negatives), 17.5 % were inconclusive, and 15.8 % were no value. On nonmated comparisons, 0.2 % of responses were erroneous IDs (false positives), 69.8 % were exclusions (true negatives), 12.9 % were inconclusive, and 17.2 % were no value. One participant made the majority of the erroneous IDs in the study, which underscores how decision rates can be highly sensitive to the individual participants in a study. No erroneous IDs were reproduced by different LPEs, but 15 % of erroneous exclusions were reproduced. There is a potential concern that NGI’s size and ability to yield much more similar nonmates (as compared to IAFIS used in (1, 2)) poses an increased risk of false IDs; however, we do not find evidence of an increase in the observed false ID rate. These results suggest that risk mitigation strategies for NGI searches may be working for those agencies that have implemented such strategies.