Will AI (Artificial Intelligence) Provide Your Own Personal Health Coach?

Adhering to the principle that health improvement is based on sustained behavior change, and that behavior change is based on a profound intervention by health care providers in a patient’s daily activities, a certain fantasy has made the rounds of the health care industry. In this article I’ll describe this fantasy and a product by Lark Technologies that starts to realize the fantasy in real life.

First, a bit of comparison. Some 15 years ago I shed a lot of weight (and kept it off) through firm but supportive monitoring by the health care profession. I went to my primary care physician every three months, and visited a nutritionist twice. I also exploited my personal network, conveying my goal to all my friends and lining them up to support eating habits that would lead there.

This model doesn’t scale well. Furthermore, a visit every 90 days is no match for the temptations that scream at me from the billboards and restaurant windows. (Believe me, as a chronic dieter I am very aware of the food industry’s marketing techniques.) So numerous technologists have imagined virtual assistants that follow you around and act like an intrusive Mom, asking you why you’re buying that donut or whether you’ve signed up for the health club yet.

These assistants would have to be subtle and very well tailored to your personal style to be affective. At the recent Connected Health Conference, MC Joseph Kvedar laid out a requirements list for such assistants. They must be:


To serve billions of people, these systems can’t depend on constant communication with health professionals. Somehow, software must be observer of your habits, come to know you from your demographics and health conditions, and intervene at appropriate moments with messages that have a chance of getting through the armor of your established routines.


This fancy word just reflects the kind of empathetic adaptions each of us does all the time to reflect the situation we’re in. Just as we would shout “Stop” to someone about to step in front of a trolley but “Excuse me, did you want this trolley?” to someone absorbed in her cell phone, contextual software understands that you like donuts (but would enjoy a good fruit salad if offered one), that you like to exercise before work instead of at lunch time, and so on. The interventions it makes for each person would be unique.


Positives work better than negatives in getting people to go along with suggestions. “Did you know that another round on the track will put you ahead of your walking record for yesterday?” works better than “Hey, you’ve been sitting for two hours–get up!”


If a user doesn’t like an app, he always has the option of turning it off. Therefore, a health app must reflect the user’s goals, not the goals of hie doctor, his daughter, or the Centers for Medicare & Medicaid Services. Empowering software will ask you what matters to you–for instance, being able to play with your grandchildren or stay in your third-story apartment–and remind you of these goals as a way to persuade you to stay on track.


I find this trait a bit redundant, if software is empowering. Dr. Kvedar suggested that people using this kind of personal agent get a discount on their health care premiums. I’m a fan of intrinsic rewards, myself. But the distinction can be hard to make. If an app sends you a message from your wife saying, “So proud that you lost five pounds this week!” is it an intrinsic or extrinsic reward?

At the conference I had the privilege of meeting with Julia Hu, cofounder and CEO of Lark Technologies, who showed off their personal weight loss coach, Lark Chat (available for download for Apple and Android). It was amazing how closely this software–available since this past April–matched the simulation that Dr. Kvedar showed off in his opening talk.

Lark Chat uses Siri software to accept voice input or a text message, which is then submitted to artificial intelligence software to respond appropriately to the user. When I told it what I (pretended I) had for lunch, the software readily understood french fries and salad, and made a comment on each. It did not understand what to do with breaded, fried fish, which ought to have triggered a warning. But it has been trained to understand a number of different foods enjoyed by different ethnic groups. Users can also opt into sharing the data collected by Lark so that it can run analytics and improve its interventions.

The interface is enjoyable and popular. According to Hu, “Over the last four months, the Lark coach and its users have text messaged each other 350,000,000 times. Based on a typical chronic disease case manager’s load, that’s equivalent to 25,402 full time nurses and coaches.” This adds up to the longest user engagement record of any interactive apps in the weight loss space.

The industry has been taking notice. Business Insider recently named Lark one of the 10 most innovative apps in the world, and Apple once featured it as the “Best New App” in their App Store. Forrester Research named Lark the “Most Innovative Digital Health Product of the Year” in 2015 and published an exclusive report on it.

People are getting accustomed to apps such as Foursquare and interfaces such as Siri that in previous ages might have been seen as annoyingly intrusive. As our relationships to devices and software evolved, we may find apps such as Lark Chat the perfect support for behavior change. And we may all become better people as a result. If only Mom could have created an app for me.

About the author

Andy Oram

Andy Oram

Andy Oram is a writer and editor at O'Reilly Media, a highly respected book publisher and technology information provider. His editorial projects have ranged from a legal guide covering intellectual property to a graphic novel about teenage hackers. Andy also writes often on health IT, on policy issues related to the Internet, and on trends affecting technical innovation and its effects on society. Print publications where his work has appeared include The Economist, Communications of the ACM, Copyright World, the Journal of Information Technology & Politics, Vanguardia Dossier, and Internet Law and Business. Conferences where he has presented talks include O'Reilly's Open Source Convention, FISL (Brazil), FOSDEM (Brussels), DebConf, and LibrePlanet. Andy participates in the Association for Computing Machinery's policy organization, USTPC.


Click here to post a comment