Biometrics Institute identifies dire need for clear language in biometrics and AI

These days, biometrics are most often deployed in tandem with AI. A new paper from the Biometrics Institute, “Members’ Viewpoints: The Relationship Between Biometrics and Artificial Intelligence (AI),” collects expert opinions that show “significant disparities,” suggesting that not everyone is comfortable with the pairing.

“Rarely has the biometrics community disagreed on an issue at this level before,” says Isabelle Moeller, CEO of the Biometrics Institute, in a post. “This paper reflects the conflicting perspectives of our global community on an evolving topic that is critical technology for biometric success. Understanding the relationship between biometrics and AI is essential for responsible innovation and the development of ethical guidelines for their use.”

The problems begin at the baseline, in attempting to define the terms. “Some say that biometrics are an adjunct to AI technology and, as a consequence, are always an integral part of it,” says the report. “Others point out that while biometrics and AI can be used together in a variety of applications it is also the case that some biometric applications exist quite separately from AI.”

“The interpretation of this relationship relies heavily on the definition being used for both AI and biometrics.”

In short, most biometrics experts agree that no one is exactly sure what anyone is talking about. The Biometrics Institute is trying to help, via its Explanatory Dictionary, a resource that aims to capture the nuances in biometric terminology, “considering both formal definitions and how they are perceived by the public – for example, how someone might explain biometrics or AI to a friend.”

Because, as of now, there isn’t a standard that is universally agreed-on, nor is there really a clear way to explain biometrics and AI to your neighbour Ted who works in marketing.

“There are no universal definitions of biometrics or AI and those put forward by ISO and some governments are either too technical, obtuse or are not fully aligned with one another or are hidden behind paywalls and not accessible to the majority of the general public.”

The paper drills down on the semantics of biometric grammar. What does it mean for a biometric application to “have AI”? Conflation of certain terms in both regulatory and public contexts exacerbates the problem. Media struggles to pick apart the web of language, and contributes its own strands in the process.

Is a tool “AI-driven,” or “AI-equipped”? Where do algorithms fit in? What’s the difference between facial recognition and face matching, and how does the distinction affect perceptions of biometrics from a civil liberties perspective? Per the report, “AI and biometrics appear to have become conflated in some regulations e.g. EU AI Act and especially the use of live and remote biometric surveillance.”

The public can be forgiven for asking, what is all this stuff and what’s it for?

The foremost international biometrics experts don’t have a satisfactory answer. They appear to agree that “there is no agreed definition of AI or biometrics.” Their befuddlement points to the elephant in the roomful of biometrics providers: no matter how accurate your algorithm is – no matter how much friction it reduces or efficiencies it gains – the world won’t care unless they know what the heck you’re talking about.

Article Topics

AI  |  biometrics  |  Biometrics Institute  |  ethics  |  explainability  |  responsible AI  |  responsible biometrics