From a linguistic perspective, it’s not easy to put the new Siri voices in a clear racial category, Holliday says. Many tiny cues in a person’s speech offer hints at their background. The new Siri voices have “some voice quality and intonational elements that are more likely to occur in the speech of African Americans,” she says, but they’re shared by some speakers of other groups.
For example, she says, the voices display more vocal fry—a creaky quality often heard at the ends of words—and a greater range of pitch than the old default voice. According to Holliday, those features are sometimes associated with African Americans, but they’re also typical of younger speakers from varied backgrounds.
What’s more, it’s particularly difficult to assign a race or an accent to a synthetic voice. A person’s speech patterns are the product of all sorts of environmental factors that shape who they are as an individual, including their age and ethnicity, and where they grew up. Siri is ageless and has no parents or hometown. “We’re trying to impute an entire identity onto this voice that’s not a person,” Holliday says.
The new Siri voices are drawing some negative reactions in addition to the positive ones. A person said on Twitter that the new female-sounding voice “seems like Siri is trying to be cool.” Another said that both new voices “sound . . . um . . . a bit airheaded.” And someone asked: “Where are the older, more mature voices that actually sound like someone that has been formally educated? If I wanted to listen to unintelligible teenagers, I’d be on YouTube.”
Holliday says those commenters may be reflecting widespread stereotypes about younger, nonwhite speakers. But those attitudes could soften if technologies like Siri gradually “normalize” a variety of voices, she says, expanding people’s conception of who sounds competent. “If people hear me and say, ‘Oh, you sound like the new Siri,’ maybe that’s not a bad thing,” says Holliday, who is Black. “Our voice assistants should be able to represent the diversity of who we are.”
While Apple will now offer more diverse voices for Siri, the devices still have a way to go in understanding the voices of Black people in everyday life, says Halcyon Lawrence, a Towson University professor who studies speech interfaces. Like other experts I spoke with, Lawrence pointed to a 2020 Stanford study that found that major speech recognition products, including Apple’s, misunderstood Black users at nearly twice the rate that they misunderstood white users.
“I am very concerned about representation (who we hear on these devices),” Lawrence wrote in an email. “I am equally concerned about perception (who these devices hear and who they discipline to speak in a particular way).”
Users who update to iOS 14.5 will be asked to choose which Siri they’d like to hear on their iPhone. It will be the first time since the assistant was introduced nearly a decade ago that Siri won’t default to its famous female-sounding voice.
Allen, the PR manager, says he’ll pick the new male-sounding voice. Paul Anthony Webb, a web designer in Silicon Valley, prefers the new female voice.
“Black people, brown people, and everybody else have always had to use these white-sounding voices, because they were the only option,” Webb says. “This is a welcome change. It’s not something I would expect any company to actually care about. Black people tend to be an afterthought.”