OK Google. Where do you get your information from?
Already, smart speakers are dispensing answers to millions if not billions of questions every month. But where do these internet-connected devices get their information from? And what does it mean for public health?
It’s just three years since the Amazon Echo arrived, pioneering what would become known as the smart speaker market. While internet-connected speakers had been popular with audiophiles for some time, Amazon’s Echo speaker had something new: the additional ingredient of Alexa, Amazon’s voice-activated digital assistant.
Of course, chatty digital assistants had been slowly making their way into the collective cyber-consciousness since 2011 when Apple embedded Siri in the iPhone 4S. But taking that human-machine interface and lodging it inside internet-connected speakers – making it part of the furniture – that was new.
Google was quick to match Amazon, porting its own voice-activated Google Assistant into the Google Home range. Apple trailed behind, eventually releasing the music-focused HomePod speaker, with Siri on board, in February 2018.
Voice-activated smart speakers are proving remarkably popular. In the first quarter of 2018, estimates from technology research group Canalys showed a 210% year-on-year market growth and Google as the top selling vendor with 9 million units of its Home devices shipped worldwide.
In Australia the statistics are a little more opaque but Google, the dominant seller locally, claims to have seen sharp growth with the number of people using a Home product tripling during December 2017 alone.
These devices aren’t just sitting around gathering dust. People are using them to listen to music and podcasts, of course. But they are also using them to get answers to questions.
Voice platform developers Alpine.AI estimate that one billion voice searches occur each month, not just on smart speakers, but across all devices with voice assistant software built-in, from smartphones and wearables to smart cars and even smart refrigerators). Media measurement agency comScore has estimated that, by 2020, 50 per cent of all search queries on the internet will be done via voice.
There are more than a few hurdles for companies, brands and organisations when it comes to voice searches.
‘These voice-activated devices present a whole load of new challenges, even a whole new landscape, for the way we get information across,’ says Viveka Weiley, Head of New Things with consumer advocacy group Choice.
Jerril Rechter, CEO of VicHealth agrees. ‘The potential of voice-assisted search is exciting, but there are concerns around where these voice-activated devices get their health information from, and just how transparent that will be for users.’
The single result
Ask most search engines a question using a screen and keyboard and you’ll be presented with a list of results to choose from, with the ones that are ads clearly labelled as such. But ask a smart speaker the same question and you’ll get a single answer. So who decides which answer is delivered?
Weiley says that the information presented will be ‘platform-controlled’, essentially meaning that Google, Amazon, Apple and any other players yet to come are ultimately in control of the results that are served up from voice search.
‘In the long run the information presented from a voice search is going to be reliant on deals and relationships with platform owners,’ he says.
Weiley notes that Choice, like many similar organisations, prizes its independence and would be reluctant to make exclusive deals with platforms to ensure that Choice results are foremost. The converse to this is that many brands aren’t so worried about deal-making.
One way that organisations can ensure their information is delivered accurately via voice search is to create their own apps for use on these platforms, although that adds additional pain points.
‘Having to say ‘Hey Google, ask X app the following question’ is a terrible solution,’ acknowledges Weiley. ‘It’s a stopgap at best.’
In the US these third-party apps (although ‘apps’ are referred to as ‘skills’ on the Amazon/Alexa platform, and ‘actions’ on Google Home) have been used in the unique US healthcare environment.
Health services company Cigna has a ‘skill’ for the Alexa platform called ‘Answers by Cigna’, capable of providing answers to more than 150 commonly asked healthcare questions and, arguably, making a decent contribution to health literacy in the community. Studies have shown that in the US many users of healthcare services don’t understand key terms used in health insurance, for example.
It’s important that we create voice applications that are fit for purpose.
Another Alexa skill for healthcare is KidsMD from Boston Children’s Hospital. KidsMD allows parents to quiz Alexa about common children’s illnesses with the results presented by the hospital. As of March this year KidsMD has fielded over 100,000interactions.
‘It’s important that we create voice applications that are fit for purpose,’ says Hugh Stephens, a Melbourne-based technologist and founder of Dialogue Consulting. ‘The best outcomes that can be delivered by organisations are when they use voice [technology] for the same reasons that [the human] voice is useful. In the same way that taking a pamphlet about exercise and shoving it on a website was never an effective way of delivering health promotion information via the web, we need to see what elements of voice search can be tailored to provide the most engagement.’
The new voice
This brave new voice world also requires some shifts in terms of how the information is presented. A good voice-activated device needs to be able to understand a conversational question, parse it down to its essential elements, answer it as a query and then present that information back as a conversational reply.
This is markedly different from the keyword style of traditional ‘visual’ searches. Google has released criteria for how voice search ranking is formulated, and Aisling Finch, the director of marketing for Google in Australia and New Zealand, has acknowledged that Google itself has had to hire a writer to help with some of the queries.
Brands are already experimenting
Fast food and alcohol brands are at the forefront of pushing products and services via voice. Companies such as Domino’s Pizza, Johnnie Walker and Patron Tequila have experimented with voice-activated advertising but Burger King has been particularly clever.
In April last year in the US if you were to ask Google Home what a Whopper Burger was, Google Home would simply read out the list of ingredients from the Burger King Whopper Wikipedia page. This information was edited, presumably by the company, to have a more marketing-style spin, with descriptions such as ‘a flame-grilled patty made with 100 per cent beef with no preservatives or fillers’.
Then Burger King got even more creative, running 15-second ads that included the phrase ‘Ok Google, what is a Whopper?’ prompting any Google Home device within earshot to read out the marketing spiel from an artfully edited Wikipedia listing.
Wags counter-hacked the site to list phoney ingredients including ‘cyanide’ and ‘toenail clippings’ and Google eventually blocked the ad, but not before it made global headlines. Later, the campaign was rewarded with the Grand Prix in Direct at the Cannes Lions, with one juror calling it ‘the best abuse of technology’.
Smart speakers remain a fledgling technology and many changes are likely over the next few generations.
Notwithstanding Big Whopper ambushes, Google’s technology is still widely seen as the most advanced, outclassing both Alexa and Siri with a ‘deep neural network’ approach to AI that continues to drive and improve other non-traditional search functions like Google Translate and Google Image. Google Brain, the company’s AI research unit, regularly releases papers on issues including ‘voice texture’ and ‘cross-dialect voice searches’. The company has also been experimenting with AI-supported health diagnoses.
It paves the way for great opportunities in the health space – but it also means that the health promotion sector needs to be active sooner rather than later.
One advantage that Stephens sees is a chance to bring technical engagement to an aged audience. ‘Voice search is actually helpful and effective for older people. Phone screens might be hard to read, arthritis might make it hard to type, but voice can help an older audience engage.’
Right now if you ask Google Home ‘how do I quit smoking?’ or ‘how can I stop gambling?’ the response is ‘I can’t help with that right now’. This opens up the undesirable possibility that a campaign like the Burger King one could be brought into play by the tobacco, alcohol or gambling industries.
‘People need to be paying attention to this space,’ says Weiley. ‘You have to be experimenting now even if you’re not sinking a huge amount of money into making your own voice app. You’ve got to have a little R&D. Now is the time to get on your feet because this is very, very different and the only way to take advantage of it is by trying it out.’
Stefan Grun, Executive Manager of marketing and communications at VicHealth agrees. ‘It’s easy to see the appeal of talking over typing and easy to imagine that voice-assisted search will grow rapidly in coming years. Commercial marketers have already demonstrated they have the skills and resources to make the most of this new information pathway. If those of us in health promotion or behaviour change are going to keep pace, we need to determine what these technological advances mean for our work and be proactive and positive in our response.’