Purchase this article with an account.
P Jonathon Phillips, Rebecca Heyer, Dana Michalski; Effectiveness of a Facial Forensic Training Course. Journal of Vision 2018;18(10):560. doi: 10.1167/18.10.560.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Forensic facial examiners are professionals trained to identify faces in images. They have extensive training, and their identity comparisons involve a rigorous process. Part of their training includes completion of courses that teach them methods for identifying faces. The goal of this study was to determine if a course improved the face matching accuracy of the students. The course was 10 days in length and took place over two consecutive business weeks. Ten students in the class volunteered to participate in the study. To be able to compare face matching ability with published results in the literature, the Glasgow Face Matching Test was administered on the morning of the first day of class. To measure the change in face matching ability over the course, two tests of equal difficulty were created. The image-pairs in the two tests (n1=30 and n2=29) were from White et al 2015 Proceedings of the Royal Society. Accuracy for each subject was measured by area under the ROC (AUC). Results: Half of the students' accuracies increased between the first and second tests; the other decreased. The accuracies of all the students were comparable to the super-recognizers in Robertson et al, 2016, PLoS One; the students were comparable to the forensic facial examiners in White et al 2015. Conclusions: The test results did not show an increase in performance over the course. The students already had superior face matching ability. The distributions of students' AUCs on the before and after tests were comparable; however, there was a high within subject variability. Recommendations: Future tests are designed for students with superior face matching ability. The tests allow students sufficient time to use the tools and methods taught in the class. The within subject variability on face matching and recognition tasks needs to be quantified.
Meeting abstract presented at VSS 2018
This PDF is available to Subscribers Only