Voice assistants could be the newsreaders of tomorrow. And you thought Facebook’s newsfeed presented challenges.
By Rebecca Heilweil
In the past, we’ve yelled at the news without expecting any response. With the rise of news content delivered by voice assistants and smart speakers, it’s increasingly common for the news to talk back.
In 2017, nearly half of all Americans were using some kind of voice assistant, usually on their mobile phones. Since then, the penetration of home smart speakers — led by Amazon’s Alexa and Google Home with Google Assistant — has more than tripled. While this burgeoning medium for news consumption is accessed and navigated by the spoken word of the user, it is mediated by the major tech companies that dominate the market. For this reason, voice assistants will not escape the information crisis plaguing every other internet-based content platform, most famously the headache that is Facebook’s newsfeed. While offering a new level of convenience, personalization, and futuristic novelty, they will present the same questions and problems — about fake news, content censorship, quality control, the distinction between platform and publisher — in new and more complicated forms.
News-content is not driving the voice assistant market, but it has become a compelling feature for users. According to a report conducted by NPR and Edison Research last year, 73 percent of smart speaker owners surveyed are interested in using their voice assistants to listen to news and learn about current affairs. The Reuters Institute for the Study of Journalism reports that just under half of smart-speaker owners regularly make use of their “news briefings” option — offering short, personalized updates that summarize current events — and that nearly 20 percent of surveyed users listened to these briefings daily.
On Amazon, “news” briefings (or “flash briefings,” as the e-commerce giantcalls them) cover a wide gamut of content, from Washington Post politics headlines, to sleep advice from the self-described “Sleep Doctor” Michael Breus. There are currently more than 9,000 flash briefings available on Amazon’s platform for Echo and Alexa. Some of those news briefings are produced by legacy news institutions such as the BBC and the New York Times; at least 1,000 appear to come from Patch.com, a hyperlocal news platform (a fact previously highlighted by Frank Catalano in GeekWire).
Meanwhile, searching the Amazon voice assistant platform for “news” yields more than 5,000 applications, which include but are not limited to flash briefings.
Amazon says flash briefings are for “news and content from popular broadcasters, local weather information, and more.” But the company does not gate-keep who can produce these updates. Beyond those that are news-oriented, available flash briefings include a mix of fun facts, advice, and news for small communities. Subjects are as varied as upcoming furry conventions, tips for taking care of pets, and the latest presidential tweets. One of the most popular flash briefings is “Word of the Day,” a tool that helps people improve their vocabulary. Then there is the slew of “unofficial” versions of media outlets, each created by individual developers featuring content from taken from myriad sources, from Breitbart to ESPN.
As with Facebook, the ability of almost anyone to develop a briefing creates a quality-control challenge, one that is intensified by a competition between news outlets and non-journalistic content in the same space. Amazon has set rules that developers must follow in producing their briefs (known in Amazon lingo as “skills”) and conducts audits to make sure content is in line with the company’s policies. Its guidelines limit or forbid some skills that promote products or services, or include pornographic or violent material. Amazon also warns that developers cannot create a skill that promotes hate speech, provides “out-of-context quotations,” or “contains or references Nazi symbols or other symbols of hate.” Users concerned with particular content can report it to customer service, Amazon says.
Moreover, some flash briefings include a guidance suggestion: content deemed “not suitable for all ages” can be rated as “Guidance Suggested” or “Mature.” Amazon also notes that the company may choose to remove content based on whether it “would violate that country’s or region’s laws, cultural norms, or sensitivities.” None of these limitations are nefarious on their face, but they do raise questions of how comfortable news outlets should be with Amazon — which has a business of its own to protect — as arbiter.
“Many of the issues are exactly the same as we’ve seen with Facebook and Google: prominence, fairness, transparency,” said Nic Newman, a journalist and fellow at the Reuters Institute for the Study of Journalism. “It’s even worse with voice, because the discovery problems are considerable. It’s very hard to find things.”
Because of its origins as a search engine, Google takes a different approach to news updates. News is aggregated on Google Assistant in a way that’s similar to how Google News curates a selection of articles on one’s computer or phone. Instead of producing a flash briefing, publishers submit their content to be indexed by Google’s systems. That content is subject to Google’s policies, which include regulations on advertisements, harassment and “deceptive behavior.”
That approach makes Google’s platform less oriented to specific outlets, though users can indicate or ask for preferred sources, or a favored topic to focus on. Both Amazon and Google are also interested in expanding their news briefings to include more in-depth and dynamic content. Google is now developing a prototype for audio news, which imagines a personalized radio station of sorts that combines more in-depth stories with general breaking news headlines. Amazon has revealed a tool for long-form news coverage that’s already available with major outlets like Bloomberg, Fox News, and NPR. For outlets, those expanding capabilities mean accepting that the companies will wield an even greater influence on the curation and delivery of their content.
But curation isn’t the only tricky issue raised by journalism delivered by voice assistant. A bigger challenge could be the voice itself.
If artificial voice technology gains a larger presence in our media ecosystem, one challenge will be maintaining distinctions between the literal “voice” of the publisher and the voice of the platform, a question that’s long haunted social media feeds. While Google Assistant’s voice generally introduces the voice of an outlet, the search engine’s artificial voice also has the capacity to convert text to speech (based on passages suggested by publishers)that can be used in response to queries.
Alexa distinguishes between itself and the outlet by speaking the name of the news source before a recording of journalists reading an update begins. But even this system leaves some room for confusion. A Reuters Institute study found that nearly a quarter of users surveyed in the U.K. couldn’t identify which outlet had provided them news through Alexa (though that number was just seven percent in the U.S.). Things might get more confusing as Amazon ditches Alexa’s artificial timbre for something that more accurately mimics a human newscaster.
In a future of voice-activated and voice-delivered news, trust-levels may be less about brand identity — of familiar fonts and logos — than about how an outlet or presenter sounds. Judith Donath, a fellow at Harvard’s Berkman Klein Center, notes that this problem has one foot in the future, and one in the past. “In some ways, it’s the same problem that they’ve had with T.V. programs or radio programs,” she said. “If you have an announcer who then also does commercials, you have a particular type of trust that’s built up in them for one type of information they’re providing and then they’re providing a very different type of information that has a very different trustworthiness profile. I think itThatbecomes even stronger with something like a voice assistant, where they’re designed to speak very companionably to you as if they’re a friend of some kind.”
She adds that the ability to easily modify voice could further complicate how audiences perceivean outlet’s or voice assistant’s identity. Consider the example of the same story delivered in a dozen personalized dialects, depending on the location and gender of the listener. Donath warns that the mutability of voice technology could “make people feel there’s a level of familiarity and of empathyto their situation that just comes from being able to tweakswitch[ing]the parameters of a voice.” A news voice could be personalized to sound more comforting — or authoritative — with no difference in the delivered content.
As Gabe Bullard notes in the Nieman Report, smart speaker usage could also pave the way to deep-faked audio. Indeed, widespread use of this technology may not be too far off. Canny AI, the Israeli ad-tech company that helped produce deep fakes of Mark Zuckerberg and Kim Kardashian, already offers a service to localize any ad in multiple languages.
Then there is the substance of what voice assistants actually say. The Reuters Institute found that voice assistants are most oftenused for news briefings in the morning and before going to bed, while news “search” is used throughout the day. These informational search requests require a single answer in the form of a single piece of content, selected by the platform and possibly influenced by user preference. “The danger is that when you’re asking a question, voice is a platform that doesn’t give you the option of 25 different [answers, as does a web search]. It demands but only one, and that gives more power to the platform,” said Newman.
Before we ask a voice assistant, “Where did Iran shoot down that drone?” we should remember to ask ourselves what, and who, we’re trusting to select and deliver that single answer.
Rebecca Heilweil is a journalist and researcher based in New York. She has written for Wired, Slate, and the Wall Street Journal. Find her on Twitter @rebeccaheilweil
Production DetailsV. 1.0.2
Last edited: July 15, 2019
Author: Rebecca Heilweil
Editor: Alexander Zaitchik
Artwork: Photo by Jan Antonin Kolar on Unsplash