With podcasts and smart speakers on the rise, Northwestern professor Candy Lee wondered how consumer preferences in voices and accents might affect the news business. Do people prefer a man’s voice to a woman’s? Would they rather hear a local accent or an “NPR” accent?
Lee, a professor at Northwestern’s Medill School of Medill School of Journalism, Media, Integrated Marketing Communications, launched a research project to explore the issue.
Among the findings: Female news readers were rated more credible, as were readers speaking with a local accent.
Lee’s survey involved 105 people at one location, so wider research would be required to reach definitive conclusions. But some of the initial findings are supported by other studies. For example, a study last year found that both men and women preferred a female voice to a male voice in their smart speakers.
Yet that raises other issues, with Apple’s Siri and Amazon’s Alexa sometimes accused of promoting gender stereotypes. “Obedient and obliging machines that pretend to be women are entering our homes, cars and offices,” United Nations official Saniye Gulser Corat wrote last year. “The world needs to pay much closer attention to how, when and whether AI technologies are gendered and, crucially, who is gendering them.”
Obedient and obliging machines that pretend to be women are entering our homes, cars and offices.Saniye Gulser Corat, United Nations official
There are, of course, male voice options for both Siri and Alexa. And researchers have developed a gender-less speaker voice called Q. (It also must be noted that real-life men and women relate differently to their smart speakers. For example, women more frequently say “please” to their speakers.)
If local news outlets want to make the most of podcasts and smart speakers, they should try to better understand consumer preferences regarding voices and accents, Lee said. Listeners’ options are particularly important because many of them accept the default voice and stick with it for a long time. “I’m not getting diversity out of my speaker,” Lee said.
“When you look at … the news at night or in the morning, you tend to see a rainbow,” she said. “If you think about the panoply, there’s usually someone who’s white, someone who’s black. There’s usually a male and there’s usually a female. But when you have your speaker on, there’s almost always the same voice.”
Study Done in Nashville
Professor Lee’s study was conducted in October 2019 at a public library in Nashville, Tennessee. Lee wrote two news stories — about a hailstorm and a health fair – and arranged for audio recordings by people with the following demographics:
- A white female with an “NPR” accent2. A white female with a local accent3. A black female with an “NPR” accent4. A black female with a local accent5. A white male with an “NPR” accent6. A white male with a local accent7. A black male with an “NPR” accent8. A black male with a local accent
Each participant in the survey listened to a recording of one of the people representing the various demographics. They rated the news reader on whether he or she was “engaging” and “credible,” among other things. The description of an “NPR” accent is Lee’s, but she didn’t present it that way to the respondents. When she asked them afterward about the accent, they used words like “proper” and “professional.” The local accent was intended to represent what’s spoken by people in the survey city, Nashville.
About 70 percent of respondents said they listened to local news, while 21 percent said they didn’t. Only 32 of the 105 respondents said they used a device with a voice assistant. Most common uses were for music, weather and Google.
When survey participants rated the readers’ credibility, the top average score was a tie between the white woman with the local accent and the white male with an ‘NPR” accent (both at 4.21). Next were the black female with the local accent (4.12), the white male with the local accent (3.88), the black female with the “NPR” accent (3.77) and the white female with an “NPR” accent (3.76). The two lowest ratings were for the black male readers: 3.58 for the “NPR” accent and 3.46 for the local accent.
In another aspect of the survey, respondents were asked whether they preferred a female’s programmed voice exclusively, and 64 percent said yes.
Lee was surprised that local accents rated high in credibility in her study. “They liked … an accent that was similar to themselves,” Lee said. “I didn’t expect that.”
This suggests that local news outlets may want to be more open-minded about what accents they feature in audio reports, rather than requiring the news reader to meet a generic national standard. Or at least it suggests that local news outlets should do more research to see what works with their audiences.
They liked … an accent that was similar to themselves. I didn’t expect that.Candy Lee, professor, Northwestern’s Medill School
When respondents were asked in a separate question about whether they preferred a voice assistant to have an accent similar to them and their family, more participants said yes than no. “Does not matter to me as long as I get the information,” said one person who answered no. A respondent who said yes explained that a local accent would make the voice assistant easier to understand — but added that if it sounded like a family member, “I would tune it out.”
Lee said a recent discussion she had with Amazon made her think a lot more could be learned about smart speaker voices.
“I called and talked to the folks at Amazon and they had done a lot of research into whether Alexa should be a woman and how slow or fast she should speak etc.,” Lee said. “They hadn’t done a lot on the effect of dialect, which was interesting.”
Because one of the survey differentiators was race, Lee hoped for diversity among those participating. The survey group was composed of 49 whites, 45 blacks, 4 native Americans, 2 Hispanics, 1 Hawaiian/Pacific Islander, 8 of “multiple” race/ethnicity and 3 “did not answer/unclear.”
One aspect of the results was eyebrow-raising: Whites in the study found a white voice more credible, while blacks had no racial preference. White respondents gave the white readers an average credibility rating of 4.07 while they gave the black readers a 3.60. Blacks gave nearly identical ratings for each racial group — 3.86 for the whites and 3.93 for the blacks.
Lee cautions against accepting that finding unless wider research confirms it. “This isn’t enough to make a judgment on,” Lee said.
Lee’s research was partially supported by the Medill School’s Knight Chair in Digital Media Strategy.