Fueled by the massive popularity of messaging platforms such as WhatsApp, “conversational UI” is providing an emerging generation of chat-based digital services that may be the next thing in consumer technology. Instead of manipulating a graphical interface, users have a conversation with a chat-bot: software that is able to understand and respond to natural language inputs. The pace of technical advances combined with a shift in cultural norms is making AI conversations feel normal for increasing numbers of people.
The idea of a “computer you can talk to” has captured the imagination of the computer science community, and the general public, for decades. Great science fiction characters such as “HAL 9000” from 2001:Space Odyssey and “Samantha” from Her have explored the promise and the risks of this idea. But real-life consumer experiences have tended more towards the comical, perhaps best represented by the “clippy” assistant for Microsoft Office. However, in the last five years leading technology brands have driven a broader acceptance of conversational interfaces within everyday life: Apple was an early mover with Siri; Google followed with Google Now; Amazon took a slightly different approach with the Echo.
Photo by Facebook
More representative of the current crop of text-based assistants is Facebook M, a general-purpose virtual assistant that “lives” in Facebook messenger and is currently in beta-testing in the US. Unlike Siri, M is not voice-powered, but it is able to sustain extended conversations and perform complex tasks, like purchasing products online. Facebook isn’t releasing details about how the technology works, but admits that the AI is still being “trained” by human operators.
There are also many simpler and more specialised chat-based services – and as one might expect, there are a number of companies working on the “AI doctor”, including Babylon Health in the UK.
To a surprisingly large degree, poor health depends on the choices we make in our daily lives – the excess of food, drink, and drugs we consume, insufficient physical activity and sleep, high stress and low attention to our wellbeing. These are the biggest risk factors for non-communicable diseases (NCDs); they are the main drivers of ill-health and healthcare costs in high and middle-income countries. Unfortunately, simply explaining this to people is not an effective way to change behaviour.
In designing solutions to help people improve and manage their health, we consistently find that people’s actions are not driven primarily by rational decision-making. Whether you are a market-farmer in Malawi or a bus-driver in Liverpool, the behaviours that impact your health are largely driven by feelings, beliefs, social and physical context, habits and addictions, and irrational ways of thinking about risks and benefits. Only by addressing all of these layers, we can achieve lasting and scalable behavioural change.
One of the most powerful methods to do this is through coaching or motivational interviews. Building on psychological theories of behaviour change, these techniques rely on a sequence of conversations that gradually unlock an individual’s self-belief, commitment, and capability to change. This process has been proven to work across diverse fields from weight-loss to alcoholism, but it can take months or years to work and is typically mediated by specialised counsellors or coaches. Scaling this method to millions of people is prohibitively expensive.
According to Public Health England, “Known potentially preventable risk factors taken together explain 40% of ill health in England. If you examine the impact of specific risks on the overall disease burden, unhealthy diet and tobacco are the two largest contributors ‡ (diet accounts for 10.8% of total disease burden and tobacco 10.7%).” Sources: Gov.Uk and Health Data.
This is where emerging AI technologies and user-interfaces that enable realistic and sophisticated conversations with computers provide a specific opportunity. It is an opportunity to leverage what we already know and accept about behaviour change and apply it together with AI to engage people in meaningful conversations about their health. It is also an opportunity to help people live healthier lives on a scale that has never before been possible.
Burden of Disease
You probably already know Jane. She lives in the London suburbs and commutes on the tube for an hour to arrive at her job as an administrative assistant. When she gets home, she orders dinner from a delivery service or prepares something quick in the frying-pan for her kids. Once they’re in bed, she can finally unwind with a bottle of wine and a cigarette before snatching a few hours rest to then start the cycle once again.
The last thing anyone should do is pass judgement on Jane, yet she’s bombarded with judgemental messages. “Jane, you should lose some weight”, “Jane, you should get some exercise”, “Jane, you should cut back on the drinks.” While these messages may be well-intentioned, they undermine her self-compassion and do nothing to help her take concrete actions to improve her health.
One evening, as Jane is catching up with friends online, she’s attracted by a different kind of message in her Facebook feed. It emphasises her importance as the epicentre of her family, resonating with her deep sense of responsibility as a mother. Unbeknownst to her, this message had been tailored and presented to her by an algorithm based on her browsing history.
She clicks to see what it’s about. Instead of the usual click-bait, this initiates a chat with a health-bot, Sara, who starts by asking her how she feels.
“Wow, how do I feel?” she wonders. “How long has it been since someone asked me that and was really interested in the answer?”
Somehow, the fact that Sara is a bot doesn’t inhibit the conversation. In fact, Jane feels freer to be honest about her feelings and aspirations knowing that a bot can’t pass a moral judgement.
Behind the scenes, Sara is following a dynamic conversation script based on proven behaviour change models. First, it focuses on Jane’s aspirations. Then it helps her to reflect on how her lifestyle impacts her wellbeing. Finally, it encourages her to choose some simple actions that she feels can be successfully integrated into her daily routine. Jane chooses to substitute deserts with fresh fruit, have two alcohol-free days a week and go for a bike-ride on the weekend. These are small steps. On their own they might not radically change her health risk, but achieving them will do wonders for her self-esteem, mental wellbeing, and sense of control – all of which are essential for larger changes to come. Jane leaves the chat feeling energized and committed to change.
Over the next days and weeks, Sara stays in touch via messenger, checking in to see how Jane is doing, providing tips and encouragement. After each success, Sara gently suggests a next step. And after each set-back, Sara is infinitely understanding and kind. Through this process, Jane is also learning to be kinder to herself – and the biggest surprise for her is how much better she feels, mentally and emotionally.
Jane is on a new trajectory that significantly reduces her risk of stroke, heart attack, diabetes, and depression, but fear and risk were never part of the conversation.
Jane is real. We’ve met her – and many others like her – while conducting design research around the world. She possesses the capability to change her own life, and we have witnessed first-hand how these capabilities can be unlocked through this kind of conversational intervention.
Sara is real too. She is to be found in the many nurses, community health workers and wellness coaches who, whether through formal training or trial and error, have found the right conversational techniques to make a connection with people like Jane.
The onus is now on us, the community of digital health innovators, to seize the opportunity provided by AI and conversational interfaces and deliver these kinds of interventions at scale. A handful of trailblazers are leading the way and achieving remarkable clinical results – from Pear Therapeutics for addiction therapy, to ginger.io for depression, to AmicoMed for hypertension. Meanwhile the heavyweights of AI, Google and IBM Watson, are collaborating with telehealth companies like Babylon and Talkspace to “train” their algorithms based on thousands of chat records with real-life coaches. However, so far we are only scratching the surface.
The techniques described here, if put to use with responsibility and compassion, can have an incredibly positive impact. However, much harm can be done by these same techniques in the hands of the cynical, the exploitative, or the merely careless. In recognition of this paradox, a group of futurists led by researcher Samuel Woolley recently published a “botifesto” posing some key questions about ethical bot design and regulation.
“Excitement about bots, from Silicon Valley to the academy, is palpable at the moment. This is partially because the creative, political, legal and ethical futures of this technology are so open. Semi autonomy, the dual human-computer nature of bots, provides fodder for all sorts of revolutionary production and reception. We must consider, however, the fact that this technology will continue to evolve and become more sophisticated. To what ends will this evolution lead?”
As designers, technologists, and business leaders, the future is not just something that happens to us; it is something that we actively shape and create. Therefore, it is our responsibility to actively shape the future of AI tools to augment and enrich our humanity – and where else should we start if not in healthcare, the most quintessentially human of activities?
This article was originally published in the Journal of mHealth: The Global Voice of mHealth