Health and fitness coaching will mostly be handled by artificial intelligence in the next five years.
How is that possible when the experience of health coaching is so personal? People who need health coaching come from a variety of backgrounds and deal with a wide range of health issues.
But we posit that Artificial Intelligence, done well, means better health coaching, for both the coach and the clients, at scale.
Here’s something most people don’t know:
Health coaching today is repetitive.
The vast amount of “coaching labor” is information delivery.
Much of the work that really, really effective coaches do can be boiled down to:
- Checking for understanding.
- Asking for clarification.
- Validating participants’ emotional expression.
- Eliciting ideas from the participant.
In 90% of interactions with a participant, the coach doesn’t offer their own opinion or dictate a next step.
Across all the domains and iterations of helping people change their behavior, the day-in, day-out of coaching can be boiled down to:
- sending the right message
- to the right people
- in the right context
…for a long enough time that they reflect, learn, and make new habits.
To any data scientists reading, this is a classic “structured learning” problem. Given a robust enough dataset, developers can teach a computer to do this and it will likely better than a human at virtually no cost. Which means scale and accessibility to anyone who needs health coaching.
What AI engineers don’t yet know about health coaching that coaches do
What’s currently missing from the data-science toolkit is the framework around which to optimize. Most data scientists are not yet familiar enough with the “how” of coaching to get machines learning in the right direction.
Currently, if we feed a computer all the variables that the machine knows and the layperson thinks is important (the words in the messages and the metadata like the times the messages were sent, the medium by which they were sent, who they were sent to, how much that person weighed at time t, etc.) and optimize for “weight loss” or “number of posts and likes,” the result will almost certainly be an AI that most consumers would call “a jerk.”
And they would be right, because that AI would be trying to control them.
Self-Determination Theory, the leading theory of behavior change (because it’s supported by 40 years and over 7,000 studies) solves the control-freak AI problem. How? With an evidence-based framework and robust psychometrics. By modeling after Self-Determination Theory, data scientists can structure machine learning using performance metrics that have decades of research validity.
The result could very well be the first coaching AI that does not come across as controlling, and instead seems genuinely interested in supporting our Basic Psychological Needs.
The day that I meet a data scientist who knows what the BREQ-2R is, I’ll know the clock is ticking.
(Side note, when we were doing customer development on Habitry iOS, we optimized for perceptions of autonomy support. Sure enough, coaches using our app have yet to lose any participants because the app annoyed them.)
Better, cheaper, kinder
There’s a funny thing about AI: it might be clunky (at first), but it never judges you. It never has a bad day. It just keeps listening and being tirelessly helpful.
Think about when you’re using Google Maps and make a wrong turn. What does the robot voice say? Does she calmly redirect you to the next best route, or does she ask you, “what the hell were you thinking?” No, she just keeps gently nudging you. And when you tell her to shut up, you don’t feel bad because she’s not real. You just reactivate her when you have the bandwidth to work on the problem again.
To most lay-people with most health-related problems, that sounds like an ideal coach. There when health behavior change is a priority, gone the moment other things need your attention, and back as soon as you’re ready for her.
What’s more, Lucas, Gratch, King, & Morency (2014) showed that “participants who believed they were interacting with a computer reported lower fear of self-disclosure, lower impression management, displayed their sadness more intensely, and were rated by observers as more willing to disclose. These results suggest that automated [health coaches] can help overcome a significant barrier to obtaining truthful patient information.”
So not only do participants perceive AI as better and cheaper, they also think of it as potentially kinder than a potentially expert human. Yikes.
AI turns coaches into awesome, scaling cyborg SuperCoaches
As a consumer, I want a car when I want a car, a burger when I want a burger, and an insightful, thought-provoking question about why I am pursuing my health goals at the moment when reflecting on it will do me the most good.
But what’s interesting about AI to me as a health coach is the stuff that it means I can focus my energy on doing the stuff I love doing as a coach and leave all the boring, repetitive stuff to a CoachBot. Maybe even a “coach’s assistant” (CoachBot!) that I’m using as an assistant so that I can help the most people possible with my unique skillset.
To steal a metaphor from The Inevitable, imagine a slider on a personal CoachBot that controlled how much the bot intervened. One one side is “back off, I got this” and on the other side is, “I’m going for a walk; ping me if anyone freaks out.” All along the middle, CoachBot is offering helpful suggestions and reminders about who I’m talking to and what is going on in their life. I have to spend way less energy on remembering stuff (“when is this person’s birthday? How much do they usually eat at Thanksgiving? Why do they say they want to be pursuing this goal?”) and can focus on the parts the of job that I love: creating genuine connections with a community of people who are trying to get a little bit better at something.
My time is spent having great conversations with people who are ready to open up and reflect, and coming up with insightful metaphors based on what I have learned about the way a person thinks. If an AI can handle all the boring things so I could do more of the stuff I love, then I would enjoy my job a hell of a lot more and benefit from the autonomy of being able to set my own hours and work from anywhere.
AI also means better collaboration between coaches, which means even more scaling
Ever since the first Motivate Summit in 2014 (and the 8 more since then), Habitry has taken note of the largest hole in most health coaches’ professional lives: collaborating with peers. Currently, it is almost impossible to work on client problems together with other health coaches. We are all working in silos of our own styles, jargon, rapport, and yes, egos. I can discuss clients with a few of my peers, but not in real time. And I certainly can’t get a few hundred of them to give me feedback. But with AI, this is all possible.
A well-designed CoachBot could make suggestions as I’m typing to a client and highlight what things my co-workers have said in similar situations. It could even tell me what Omar Ganai or Dan John would say if he was talking to this client. A CoachBot could tell me when I’ve “unlocked” a new level of perceived autonomy or competence support and ask me to update a common Wiki to teach my fellow coaches how I did it. Or it could simply let a trusted peer interact with my clients for me while I was on vacation without ever revealing their privacy information.
What’s missing? Again, we at Habitry believe that a common framework, a shared language about the “role” and “goal” of coaching, is what is currently holding back this level of collaboration between coaches and a damn fine CoachBot. Once again, better dissemination of what 40 years of Motivation Science has taught us and how it can be used to optimize the coaching experience is the answer and to what Habitry has dedicated our time and energy.
With a common language, common psychometrics, and common understanding of the problems we solve and the levers we have at our disposal to solve them, there can be true “collaboration” and scale the things that only humans can do.
How Habitry is working towards this
The reality is, Habitry does not have a data science team (day-to-day, it’s just Stevo, Omar, and Vanessa). Our superpower is not engineering, it’s community organization and education. We are proud of the 600 members of the Motivate Forum and the Habitry Professionals who are working every day to create the education and peer relationships necessary to bring the next generation of coaches up in a world where Motivation Science is ambient knowledge and technology is a tool that is embraced instead of feared. The reality is, the apps are coming. The FitBits are here. And the data science teams at the largest companies in the world are already trying to get past the controlling “jerk app” stage in User Experience design and actually send messages that inspire reflection instead of reactance.
Next year, we’ll be partnering with the best SDT-based design firm in the world, mad*pow, to host the first SDT UX Unconference for User Experience Designers, data scientists, and yes, coaches to meet each other and begin collaborating on the next generation of intelligent coaching software. Who knows; maybe the people who give us CoachBot will meet there over beers.