In 2002 an MIT scientist named Timothy Bickmore recruited 91 students to a study in a field of research that he later dubbed ‘patient-centric computing.’
Over 30 days, two groups of participants interacted with a PC-based health coach called Laura, designed to engage users in a dialogue around taking more exercise. For one group these interactions were strictly functional, but for the other group Laura used more emotional language and her onscreen avatar displayed empathetic facial expressions.
Bickmore’s study was based on existing research demonstrating the importance and impact of patient-physician communication skills on satisfaction and treatment outcomes. His aim was to see if these same ‘relational’ techniques could succeed if applied by a computerised virtual coach rather than a real-life one.
The results were encouraging. Both groups reported increased levels of exercise over the study period, but those who had interacted with Laura reported a greater desire to continue with these new healthy behaviours.
From a 2018 perspective it’s easy to recognise Bickmore’s Laura as the mother of today’s health-oriented chatbots and virtual coaches. Only now, in a world of ubiquitous smartphones and comfort with messaging apps, two recent studies demonstrate that the human face, name and pseudo-realistic animation adopted by Laura are no longer necessary and that text-based conversation alone is enough to build a healthy therapeutic relationship.
The first of these studies looked at ‘Woebot’, a Facebook Messenger bot delivering text-based cognitive behavioural therapy, which successfully reduced self-reported symptoms of depression in its users.
The second looked at weight-loss coaching app Lark and showed that – if delivered properly – a chatbot style intervention could successfully support people at risk of type 2 diabetes both in losing weight and in increasing their percentage of healthy meals.
What makes Lark’s approach still more interesting is its use of artificial intelligence (AI) to help gauge its empathetic conversational prompts and responses based on user data.
The potential of chatbots and ‘conversational interfaces’ like Lark and Woebot for delivering patient support and engagement is supported further by data from research that Syneos Health conducted at the end of last year.
As part of a wider focus on global patient attitudes towards artificial intelligence in health, we asked over 650 patients across Europe with an average age of 56 about their current use of technology, and their level of comfort engaging with AI in different scenarios related to their health.
Even among this older audience, more than 60% were comfortable with the idea of an AI-powered ‘virtual nursing assistant’ to provide ongoing support or monitoring around their general health and wellness, around a particular condition, or in relation to a specific prescription product.
We also uncovered a finding that very much points to the next logical evolution of conversational interfaces. When asked about the most appealing ‘human’ characteristics of virtual nursing assistants, a ‘realistic voice’ was rated far more highly than embodied attributes such as a physical presence, a realistic face or even a human name.
AI-powered voice assistants such as Amazon’s Alexa are now a presence in an estimated 30 million homes worldwide, offering an increasingly seamless way for users – particularly those with low literacy or technical ability – to interact with digital content and services.
While these devices are still far from mainstream, and robust applications for healthcare are yet to emerge, the capability of AI to deliver increasingly seamless voice-driven interfaces will only improve over time. Indeed, now that Amazon is making its first, long-awaited moves into health, it seems a safe bet that Alexa will have a central role in its future service developments.
So, what should life science brands be doing today in order to prepare for this tomorrow?
Our survey data suggests that patients see manufacturers’ greatest ‘right to play’ for virtual assistance as being around support and monitoring for the specific products in which they’re acknowledged experts.
And let’s face it – product specific patient information is more than overdue an overhaul. A number of studies point to patient support content which is written in overly technical language or which (according to a 2017 survey by the Academy of Medical Sciences) may even put patients off taking a product before they actually start it.
So regardless of whether the age of voice arrives in two years or in ten, there’s not a single brand out there that wouldn’t benefit from some immediate speech therapy and from reimagining their content for the coming age of Laura’s grandchildren sooner rather than later.
Duncan Arbour is SVP innovation at Syneos Health Communications
