Ensuring equitable representation in the development of voice authentication systems

With the rise of voice biometrics, questions are being raised about the use of this technology among groups of people who change their voices, including the transgender community.

In a recent conversation with The New York Academy of Sciences, Juana Catalina Becerra Sandoval, a Harvard University PhD, candidate examined the ethics of developing voice biometrics.

While some debates center on over-policing and over-surveilling, others look at how voice biometrics systems – whether used by the financial sector or the security and surveillance sector – leave out certain people.

“In the security sector, I think my main concern is that there’s a presumption that the relationship between voice and identity is fixed and immutable, which can create problems for people who want to change their voice and or for people whose voice changes in ways outside of their control, like from an injury or illness,” says Sandoval. “There are numerous reasons why people might be left out of these systems, which is why we want to make sure we are creating infrastructures that are equitable.”

Sandoval is also a research scientist in the Responsible and Inclusive Technologies initiative at IBM Research.

Whether voice biometrics are posing challenges for groups like transgender people is also becoming a topic among technology companies, including those that provide biometric-based digital identity and fraud prevention.

Transgender and gender-diverse people often go through medical procedures that help them align with their gender identity, including those that change their voice. Voice biometric systems, however, may still use their pre-transition voice patterns leading to mismatches and exclusion from essential services and potential suspicion, Mitek’s Chief Marketing Officer Cindy White writes in a recent blog post.

“Voice biometric systems rely on extensive datasets to establish a baseline for comparison and identification,” says White. “Unfortunately, transgender voices are often underrepresented or not included at all in these datasets. This exclusion can lead to biased algorithms and inaccurate results when authenticating transgender individuals, further perpetuating discrimination.”

Mitek suggests collaboration with transgender advocacy organizations to ensure representation and reduce bias in voice biometric systems. Developers should also work on adaptive algorithms that can accommodate variations in voice patterns due to factors such as gender transition or hormone therapy while companies should ensure they are more inclusive during voice biometric system implementation, says White.

“These algorithms should be designed to adapt to an individual’s unique voice characteristics over time, allowing for accurate and inclusive authentication,” she adds.

Defining the relationship between AI and biometrics

What is the relationship between AI and biometrics? How does AI interact with biometrics and what are the corresponding issues?

These and other questions were asked by the Biometrics Institute during several events held in the past months in order to formulate a final statement on the relationship between AI and biometrics. The statement will be presented to members for review at the upcoming Biometrics Institute Congress in London in October and Showcase Australia in Canberra in December.

The independent and international member organization for biometrics has invited members and other experts to provide their perspectives. The goal of this statement is to provide greater clarity to members, policymakers and regulators, while the output will be entered into the Biometrics Institute Explanatory Dictionary.

The institute launched a discussion paper on the topic in April this year, inviting responses from the public. Since then, the organization has held several Biometrics Institute conferences discussing the issue.

In October last year, the organization also published its fifth State of Biometrics Report.

Article Topics

biometric authentication  |  biometrics  |  Biometrics Institute  |  ethics  |  Mitek  |  research and development  |  voice authentication  |  voice biometrics