AI, Health Coaches, and AI Health Coaches—Oh My!

Health Coaches

Health coaching, when properly utilized, maybe a helpful tool. You must understand the profession: what it is, what it requires (or doesn’t), what a health coach can and cannot do (e.g., cannot prescribe drugs or diagnose an illness), and how to vet information. 

Health coaches identify as someone who can help you with your health challenges and goals. Often, they advertise as specializing in nutrition, chronic disease management, and smoking cessation and generally promote more healthful behaviors. They’re not regulated. There are no standards for education, training, or experience. Anyone can identify as a health coach, meaning the information they provide varies wildly. 

While health coaching may sound like a promising idea, with no regulations, misinformation, under the guise of expertise, can easily run rampant in this environment. This can endanger the health of those seeking their guidance.  

AI

Like health coaching, AI can be a fantastic tool when used properly and responsibly. This involves understanding that AI is prone to bias and errors (hallucinations). As such, it’s essential to consider what you really know about the algorithm and what it returns to you. For example, what do you know about how the algorithm works? Where does it get its information from? What are the limitations of the training dataset? Also, consider the results it returns for you: have you double and triple-checked the responses? Are you able to find those same responses from reputable sources? 

The information that AI provides should be considered less than reliable. It needs to be thoroughly vetted and investigated, which needs to be done every time you ask for information. Some questions are lower stakes than others, such as asking when a specific holiday falls. It’s important to understand the types of assumptions that AI may have to make, even in the low-stakes questions. For example, what year? According to what calendar? Are you asking about a specific event such as a holiday that goes from sunset to sunset, in which case, what day does it begin? Or end? Those questions can be quickly checked, and the risk of getting them wrong is relatively low, all things considered.

Now consider a more nuanced question like “Is sunscreen more likely to cause cancer than the sun?” The answer AI provides will solely depend upon where it gets its information and the weight it gives to specific types of sources. That depends upon the programmer, the algorithm they’ve designed, what access to sources and information the algorithm has, and what sources are specifically targeted. The algorithm will reflect the reality of those programming it: their beliefs, knowledge, and ability to vet sources and account for new information. It’s not hard to imagine a world where one AI is popular with a specific group of people and another AI is popular with a different group of people because each algorithm is trained on and reflects their respective world views. It’s easy to imagine how this would exacerbate the echo chambers and confirmation bias that seem to fuel the internet. 

The Combination

Arianna Huffington and Sam Altman coauthored an article announcing their collaboration on an AI health coach. According to the duo, 

“It will be trained on the best peer-reviewed science as well as Thrive’s behavior change methodology—including Microsteps, which are tiny daily acts that cumulatively lead to healthier habits. And it will also be trained on the personal biometric, lab, and other medical data you’ve chosen to share with it.”

This sounds fine, in theory. However, all the problems identified with the health coaching profession and AI still apply. Plus, there’s the question of what happens to your health data once you’ve entered it into the algorithm. Where does it go? Most importantly, who owns it, and who has access to it? 

HIPAA  and the HITECH Act protect your health data,  unless you willingly give information away. The health information you willingly shared with the AI algorithm could be used in myriad ways. Insurance companies would be very interested in getting their hands on those data, and nothing is stopping Thrive AI Health from selling your data. 

Aside from a significant privacy concern, there’s the question of how much of an impact this coach will have on your health. The theory behind this AI health coach is that individual, prescriptive behavioral change will make people healthier, leading to fewer chronic diseases maybe even reversing some. However, there is reason to suspect that an AI health coach will not solve the problem of chronic diseases because it completely ignores any societal factors. Writing for The Conversation, Jathan Sadowski stated, 

“But individual lifestyle choices aren’t everything. In fact, the “social determinants of health” can be far more important. These are the social conditions that determine a person’s access to health care, quality food, free time, and all the things needed to have a good life. Technologies like Thrive AI Health are not interested in fundamental social conditions. Their “personalization” is a short-sighted view that stops at the individual.”

In health, and especially in public health, every little bit helps; I don’t wish to devalue that. However, I suspect the AI health coach will be far from a health revolution. Also, as I’ve said before, individual solutions to primarily systemic problems are not solutions. The people who would most likely use this are probably not those in most need of it. Also, those in greatest need may likely benefit more from societal-level changes. If you don’t have enough money to buy mostly “healthy” foods or live in a food desert, you have to ignore the behavioral nudge; if you don’t have enough time in the day to exercise, then you will ignore that nudge.

I don’t wish to sound cynical or immediately dismiss a new technology simply because it’s new. However, there are some critical unanswered questions regarding this initiative. In the world of health and healthcare, “Go fast, break things” gets people seriously injured and killed. Therefore, I recommend that you “Go slowly and thoroughly test.”