Developers talk promise, pitfalls of voice assistants in healthcare

By Jonah Comstock
Share

While its easy to think of voice apps as just another platform to develop apps for, voice interfaces offer a whole new way of interacting with consumers and present their own host of design challenges.

At the Connected Health Conference in Boston last week, three healthcare developers talked about their experience developing voice apps in healthcare, and where they think the technology is going.

“Voice as an interface is compelling because if we all think about it, it’s a much more natural way of engaging with technology,” Nathan Treloar, president and COO of Orbita, said. “In a perfect world we’d all harken back to the old Star Trek episodes where you could just speak to the computer and ask it any question you want. And a more conversational style is more natural in healthcare settings.”

Anne Weiler, cofounder and CEO of Wellpepper, which recently won the Alexa Diabetes Challenge, said a combination of factors is leading to voice’s moment in the sun.

“I think what we’re seeing is the confluence of ease of use, price point, and the ubiquitous technology,” she said. “I don’t know that we could have had this voice moment if we hadn’t had our mobile moment first, which was the widespread adoption of mobile devices that we’re trying to get to do everything for us. And there comes a point where, trying to get the mobile device to do certain things, you actually do want that voice interaction.”

Weiler agreed with Treloar’s assessment that voice can be a particularly good fit in healthcare, where patients have a strong need to be heard.

“One of the things is the people we tested it with, they told us they felt cared for and they felt that empathy was delivered,” she said. “We try to do that in apps as well, but the power of voice to do that is much greater. The healthcare experience is very natural; it feels like a conversation with a healthcare professional.”

Andy Kropa, founder of SoundMind Intelligence, also pointed out that voice has great potential for data collection, which is also something that more healthcare organizations are looking for.

“What’s really interesting is you’re able to know what people are asking for in a more direct way because with the computer, you can ask anything,” he said. “You’re able to learn and respond much faster when you see there’s a certain demand for things. We definitely count interactions on a daily basis and rely on surveys and that kind of reporting to deliver the new technology into use.”

However, there are some challenges and cautions for building voice apps in healthcare. One is that developers have to recognize that cut-and-paste approaches don’t work.

“You can’t just take a web page of 5 pages of content and stick that into a voice app. It’s not going to work,” Treloar said. “There are patterns for how you make longform text voice-ready, and one of the patterns is to make it more interactive and conversational. If you were talking to a human, a human’s not going to talk at you for five minutes, they’ll say ‘what else do you want to know?’ And that’s the nature of voice — it makes you think about those activities differently.”

Panelists also expressed the clear difference between intelligent voice assistants and phone trees as an example of what the experience can be like if you don’t put enough work into developing it. Treloar also gave another example from his company’s work in healthcare.

“We put an assessment survey for rheumatoid arthritis into a voice experience,” he said. “It’s three questions, but it’s not really three questions because the third question is ‘Which of the following symptoms do you have?’ and it’s 50 checkboxes. How do you put that in a voice experience? ‘Stop me when I list one that you have?’ It’s not going to work. So you have to rethink the way you deliver these text-based experiences.”

Weiler said some other challenges are getting voice apps to recognize things like colloquialisms or mispronounced drug names. With seniors, Kropa said, users can have too much empathy and forget they’re talking to a machine.

“If we have a session where we’re showing different capabilities, at some point we have to stop to remind people that they’re talking to a computer because that realization doesn’t enter their heads,” he said. “We see seniors being super polite to Alexa, so at some point we might say ‘You don’t need to say please but it’s great that you do.'”

One of the biggest challenges in developing for voice right now, however, is developers doing it for the wrong reasons.

“With voice having a moment, there are an awful lot of skills out there, ... it’s exactly like the early days of the app store with apps that weren’t actually apps, they were features,” she said. “‘Just go build a skill’ can be fun, but if you’re looking to make an impact in healthcare you have to ask what is the problem you’re trying to solve, who are you trying to help, and how does this take you further than what you can do today?”

One thing the panelists agreed on: Voice technologies are not going away. They’ll keep getting more advanced and more ubiquitous, so the time to start making them work for healthcare is now.

“I think what we’re seeing with voice is an interesting home technology that is inexpensive enough that if all I do is have it play music, I’m satisfied as a consumer,” Weiler said. “But then once it’s in the home there’s an opportunity to do a whole lot more.”