Design and Implementation of Forensic Facial Identification Experts Test

As submitted by the applicant: In 2009, the National Academy of Science (NAS) issued a report on the state of forensic science in the US. This followed a spate of DNA-based exonerations of people falsely convicted of crimes. Inaccurate identifications by forensic examiners formed a common thread in these cases. The report specifically recommended research “on sources of human error in forensic examination”. We address the need for research pertaining to one important source of human error in forensic examination: the error introduced by practicing facial identification experts. The purpose of this project is to develop a flexible and challenging test of the face identification skills of practicing forensic examiners. This test will consist of a series of small-scale face identification experiments that can be distributed to laboratories responsible for forensic facial examination in criminal justice applications. The test will target situations that are known to produce identification errors in the general population and are known to be challenging for computer-based face recognition systems. These situations include identification under poor illumination, at a distance, with low-resolution images, and when the pose of the face is non-frontal, or varies between comparison images. We will also include identification across faces with diverse ethnic/racial backgrounds. In the first part of the work, we will select the most challenging images and videos by extensively pre-testing untrained human participants with hundreds of images and videos and by incorporating baseline data on these items from computer-based face recognition software. In the second part of the work, we will analyze existing data from a recent test of forensic facial examiners to select images on which experts’ identification judgments were incorrect or were divergent across examiners. The results of the pre-test experiments, in combination with the analysis of experts’ data, will be used to compile a comprehensive face identification test for forensic facial examiners. In the third part of the work, we will distribute this test for comments and feedback by experts in forensic facial identification. We will also collect preliminary data from experts to refine the tasks to optimally tap skills that are most relevant for the face identification tasks they carry out professionally. The resulting tests can be used to troubleshoot weaknesses in examiner performance and ultimately to guide training programs aimed at mitigating performance vulnerabilities both within and across forensic labs.

This project contains a research and/or development component, as defined in applicable law.

ca/ncf