Facial Recognition Technologies in the Wild: A Primer

Buolamwini, Joy; Ordóñez, Vicente; Morgenstern, Jamie; Learned-Miller, Erik

In recent years, facial recognition technologies (FRTs) have experienced enormous growth and
rapid deployment. The potential benefits of FRTs such as increased efficiency, diagnosis of 1
medical conditions, and the ability to find persons of interest are tempered with risks of mass
surveillance, disparate impact on vulnerable groups, algorithmic bias, and lack of affirmative
The passage of city and statewide restrictions [9, 8, 10, 7] and proposed federal legislation [58,
22, 23, 21] show growing public concern. They also demonstrate the need for comprehensive
policies to address the wide range of uses across private and public sectors. Current legislative
efforts address a patchwork of different applications, jurisdictions, and time periods. They do not cover the full scope and spread of FRTs.
The ubiquitous scenarios that lawmakers have not yet addressed require oversight and guidance
for industry practice, research norms, procurement procedures, and categorical bans where
deemed appropriate. Depending upon the application, societal, legal, ethical, financial and even physical risks demand a thorough understanding of real-world impacts. How can we manage such a complex set of technologies with such enormous societal implications?
Many other authors have addressed the ethical and societal implications of FRTs [65, 11, 47, 69]
and of artificial intelligence more broadly [24, 40, 4]. Several groups have argued for new laws
and regulation of face recognition technologies (see, for example, [67, 20]). These earlier works have illuminated the widespread problems that emerge with the deployment of FRTs and related technologies. Yet in isolation they are not enough. It is time to take the next step and make a specific proposal about how to move forward. This white paper makes the following central claim.