Three Ways Machine Learning Will Drive Behavior Change in Mobile Health
Today’s most prevalent diseases/conditions are both preventable and manageable. In most cases, heart disease, type 2 diabetes, hypertension, obesity, and other conditions can be avoided and offset by healthy living. A healthy lifestyle often requires people to change certain behaviors. And, as we all know, behavior change is hard.
People are adopting mobile health apps and Internet-connected medical devices to help them in this effort. New machine learning technology will greatly augment the ability of these apps and devices to spur the desired behavior changes.
In this case, I’m referring to machine learning algorithms that create behavioral profiles of end-users based on their actions in the physical world. These algorithms convert raw sensor data from a user’s phone and connected devices into insights about their habits, activities, and meaningful locations.
Digital health apps can draw upon these AI-driven user profiles to influence behavior change in the following three ways.
1. Reaching people at optimal moments.
One way health apps try to influence people’s behavior is through notifications and alerts. For example, a wellness app may prompt a user to go for a walk each day. Or, a medication adherence app may issue alerts telling a user to take their blood pressure medicine. These apps either rely on users to set a time for an alert to go off or blindly guess the best time to issue a notification.
Ill-timed alerts hurt the end goal of fostering healthy behaviors.
It’s easy to imagine how ill-timed alerts could hurt the end goal of fostering healthy behaviors. If the wellness app regularly sounds off while the user is driving, it not only reduces the odds of the user addressing their personal wellness, it increases the odds of them crashing their car. Likewise, if the medication adherence app regularly wakes the user from a deep sleep in the morning, the user will likely delete it. A deleted app is not quite as effective as an active one.
“User-aware” apps, however, fuel positive behaviors by reaching users at optimal moments. Because they know a user’s habits, activities, and meaningful locations, they issue “moment-based” notifications. For example, the wellness app from above could prompt the user to go for a run as soon as the user arrives home from work. The app catches them at a juncture when they’re most likely to engage. Similarly, the medication adherence app could detect the exact moment when the user wakes up in the morning and issue the reminder on the spot. The app communicates with the user at a convenient time, thus increasing the likelihood that the user will take the desired action.
“User-aware” apps fuel positive behaviors by reaching users at optimal moments.
2. Giving people self-knowledge that inspires behavior change.
The fact that mobile apps and fitness trackers automatically log people’s activities is an important step in engaging people in their own health. Indeed, a well-publicized Walgreens study reveals the power of automation in influencing people’s behavior in this way.
But again, AI-driven behavioral profiles will multiply the benefits of lone fitness trackers and apps. Digital health tools, enhanced by machine learning, have a holistic view of a person’s physical life, not just of their daily step count.
AI-driven behavioral profiles will multiply the benefits of lone fitness trackers and apps.
For example, a user-aware wellness app can share detailed activity dashboards that show a user’s total wellness profile — exercise habits, sleeping patterns, time spent sitting in the car, and so on — and offer health tips based on that profile. Or, an app that helps people manage a pulmonary condition can give users insight into how certain actions affect their breathing.
The more context an app gives about a person’s health, the more compelling the information will be to the user. And the more compelling the information, the more likely it is the user will address some of their unhealthy behaviors.
3. Sharing personalized resources that resonate with people.
Apps that share overly general health resources with their users eventually descend into irrelevance. User-aware apps, on the other hand, can tailor their messaging to individual users based on their behavioral profiles.
Apps that share overly general health resources with their users eventually descend into irrelevance.
For example, a dieting app, sensing that a user has poor sleeping habits, can proactively share resources on sleep and obesity. In another case, a diabetes management app, knowing that a user is a workaholic, may inform them of how overworking can rattle a person’s glucose balance.
This has a cascading effect. If an app consistently sends the user engaging messages, then the user is more likely to read the content of those messages. As they read more content tied to their health, these ideas become more deeply embedded in the user’s mind. These ideas then manifest themselves in the user’s actions in the physical world.
Behavior change is a “game of inches.” Advancements in machine learning just may turn it into a game of yards.
About the author: I’m the content writer at Neura, Inc, a startup whose personal AI service drives engagement for mobile apps and IoT devices.
