Forensic facial examiners receive extensive training at comparing images of two faces

This Cooperative Agreement award from the National Institute of Standards and Technology (NIST), under the Federal Grant Program for Measurement Science and Engineering (CFDA 11.609), provides $199,989 to the University of Texas at Dallas (UTD) to conduct research on forensic facial examiner performance and methods for effectively combining human and machine face recognition systems.

The key products and services to be delivered include:

• Administering a test to forensic examiners to assess their accuracy in facial comparisons when the race of the examiner and the faces are different
• Investigating methods for creating optimal "smart teams" of humans and machines to increase accuracy and oversight of face recognition systems

The award aims to provide insights that will benefit forensic examiners, law enforcement, the legal system, and researchers in related fields. The research findings are expected to be presented in open scientific forums and published in the scientific literature.

Activities

This project will coordinate with NIST efforts to study the effect of cross race in facial comparisons and combining humans and machines for face recognition. Facial forensic examiners perform comparisons in forensic laboratories, where they have access to special tools and methods. Previously, the team designed a test to measure accuracy when the race of the examiners and the faces are different. The team will administer this test to examiners so they have access to their tools and methods. The examiners will return their answers for analysis. Recent studies found that combining humans and machines are more accurate than either the human or machine alone. Based on these results and the decision-making theory, the project will investigate methods for creating smarts teams of humans and machines optimized to increase accuracy and oversight.

Expected outcomes

For forensic science, this research will produce results from administering a facial comparisons test when the race of the examiners and the faces are different. The studies on combining humans and algorithms will produce insight into creating effective oversight of systems. The research supported by this grant will be presented in open scientific forums and published in the open scientific literature.

Intended beneficiaries

Forensic examiners, law enforcement and the legal system will benefit from knowing the accuracy of comparisons when the race of the examiners and the faces are different. This knowledge will facilitate investigators, lawyers, and judges interpreting the results of these facial comparisons. Forensic scientists, psychologist, and neuroscientist will have a better understanding of methods and techniques for measuring the performance of forensic examiners. Engineers, forensic scientists, and psychologists will benefit from the additional insight into designing face recognition systems that require human oversight.

Subrecipient activities

There are no planned subawards.