Industry and Civil Society Organizations Demystify Facial Recognition Systems
Partnership on AI’s “Understanding Facial Recognition Systems” establishes a common language to inform discussions on the role of facial recognition systems
SAN FRANCISCO, Calif. Feb. 19, 2020 – The Partnership on AI (PAI), a nonprofit coalition committed to the responsible development of artificial intelligence, today published the paper, “Understanding Facial Recognition Systems,” to equip policymakers, journalists, and the general public with a shared language to inform discussions on the role of facial recognition systems in society. The paper is accompanied by an interactive graphic to illustrate what facial recognition systems are and how they work.
Concerns about facial recognition systems used to collect, track, or surveil a unique and exposed part of the human body-one that is, for many, directly associated with identity, privacy, safety, democracy, and security-raise important questions about the appropriate role of this technology in society. These considerations have prompted calls for policymakers around the world to take immediate steps to determine whether and how facial recognition systems can be used to benefit society while preserving human rights and civil liberties.
The development and deployment of facial recognition technology increases the capabilities of governments and private actors to collect, control and analyse our most sensitive and unalterable biometric information, namely, our face. Access Now calls for a debate on facial recognition that is democratic, evidence-based and includes those stakeholders who will be most disproportionately impacted by this technology. We further call for stronger enforcement of human rights to protect individuals and communities, and to put new safeguards in place including red lines for areas where the technology must not be developed and deployed at all. Access Now welcomes the Partnership on AI’s efforts to provide clarity on definitions and to dispel misconceptions around facial recognition as an important contribution to the debate on this technology.
“Understanding Facial Recognition Systems,” aims to ground discussions and policy debates surrounding facial recognition systems in a shared understanding of their technical capabilities. It describes how a facial recognition system works, clarifying the methods and goals of facial detection, facial verification, and facial identification. The paper also explains that systems that analyze and categorize facial characteristics are not a part of facial recognition systems, because they do not verify or predict someone’s identity. The paper additionally includes a list of questions that policymakers and other stakeholders can use to elicit additional technical and related information about facial recognition systems.
A productive discussion about the role of these systems in our society starts when we speak the same language. Demystifying these systems can bridge conversations between those developing and using the technology, policymakers, and those whose faces and names are wittingly or unwittingly included in these systems. While it is important to understand how these systems work, PAI also recognizes that facial recognition systems are developed by humans, and their use cannot be separated from existing cultural, social, and economic power dynamics.
Two key findings in the paper reveal that each facial recognition system is unique and that design matters. There is no one standard system design for facial recognition systems, and not only do organizations build their systems differently, but they also use different terms to describe how their systems work. The results that facial recognition systems present to users are also dependent on how the systems were designed, developed, tested, deployed, and operated.
The paper is the result of a series of workshops on facial recognition systems held with PAI Partner organizations, multidisciplinary and experiential experts, between September 2019 and January 2020. The explanations in this paper, informed by these briefings, aim to provide a consistent set of descriptions to ground future discussions.
Facial recognition systems are one of the key human rights threats of our technological age. Face identification systems pose a significant risk to human rights, including the rights to privacy, non-discrimination, freedom of expression, association and peaceful assembly. Amnesty International is calling on governments and companies to address these risks, and agreeing on terminology and using a shared language is a vital first step towards concrete action – including regulation, standards and safeguards – that protect human rights. The Partnership on AI meetings have been critical in furthering Amnesty’s analysis of this technology.