FB pixel

Biometric face recognition literacy gap in U.S. causing problems, Mitre report says

Biometric face recognition literacy gap in U.S. causing problems, Mitre report says

Mitre Corporation has published a technical paper offering practical recommendations for implementing biometric facial recognition based on real-world experience, the company has announced. The organization argues that best practices, policies and oversight will be necessary to enable the appropriate use of facial recognition applications while protecting civil rights and liberties.

The four-page document provides definitions, including for “biometric” as both a characteristic and a process, a distinction infrequently seen in documentation, and begins by outlining the tension between the technology’s benefits and problems in the U.S., which Mitre lays at the feet of “untrained and inexperienced entities.” These entities have in some cases attempted to use the technology for legitimate purposes, but confused users have produced inconsistent results, or promote policies and legislation that block or limit the appropriate use of facial biometrics, intentionally or otherwise.

The company notes its experience in helping federal agencies install and operate systems with facial recognition, as well as privacy teams and a former Assistant Director of the White House Office of Science and Technology Policy as providing the basis for the paper’s analysis and recommendations. The latter expert served as a leader for government-wide policy in identification technologies during both the Bush and Obama administrations.

Mitre identifies what it calls “the face recognition literacy gap” as an ongoing issue, pointing to the consistent misrepresentation of NIST demographic report results, and the conflation of the accuracy of software for non-identification purposes such as age estimation and medical condition detection with biometric face recognition. The poor conception and management of implementations by novice operators with limited training on the technology’s use, or the appropriate safeguards that should be applied, or disregard of policies by those operators, add fuel to the fire, and are often held up as representative of all uses.

Four insights are offered in the paper. People evaluating facial recognition systems should consider the entire system, rather than just the algorithm, as other factors can have an even greater impact on the system’s usability and accuracy. They should also understand the differences and similarities between facial biometrics and facial analysis algorithms, and that the technology is probabilistic, making it more appropriate to think in terms of error rates than “accuracy.” Uses cases also vary too widely to support many valid generalizations, with general investigations or considerations of multiple use cases leading to incorrect assessments, Mitre writes.

A bipartisan group that would be created by recently-proposed legislation to formulate policy for facial recognition would therefore only be valuable if those participating are sufficiently informed and experienced in the issues.

Mitre provides a recommended framework for policy deliberations, with four areas of inquiry each considered according to six different criteria. Finally, Mitre recommends Congress issue a resolution on its concerns with and expectations for the technology to signal its intention to update regulations and oversight that have been left behind by the technology’s advances. This step, according to the organization, could provide a deterrent to irresponsible implementations while buying lawmakers the time they need “to study this issue and craft legislation that is evidence- and outcome-based, actionable, equitable, and measurable.”

Related Posts

Article Topics

 |   |   |   |   | 

Latest Biometrics News


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

The ID16.9 Podcast

Most Read This Week

Featured Company

Biometrics Research

Biometrics White Papers

Biometrics Events

Explaining Biometrics