OCTOBER 16-18, 2019



October 16-18, 2019 | Boston, MA

Insight and Intrusion; Connectivity and Connection

Ellen Beckjord, PhD, MPH

Earlier this fall, Dr. Kate Wolin explained that context plays an important role in understanding barriers to, and facilitators of behavior change. She acknowledged the role of habit formation in behavior change efforts: most health behavior changes require stopping existing habits (e.g., sedentary behavior) and adopting new ones (e.g., a regular walking routine). But Dr. Wolin pointed out that habit change strategies alone (sometimes all that’s contained in commercially available health apps) are insufficient without acknowledgement of context (e.g., if a person doesn’t live in a neighborhood safe enough to walk in, habit formation focused on taking outdoor walks won’t have an impact).

I agree with the points in Dr. Wolin’s post and look forward to discussing context as part of how to help digital strategies to support health behavior change be successful at Connected Health this year.

The sensors embedded in a typical smartphone combined with the data we all generate through routine use of the Internet provide more than enough information to make reasonably accurate inferences about context in real- or near-real-time. Yet today, inferences about context are still largely made on the back of actively-reported data by app users (ecological momentary assessment protocols; responses to repeated measurement of contextual factors using (hopefully) validated questionnaires). One of the biggest opportunities available to digital behavior change interventionists is better leveraging of passively-sensed data to make inferences about context in order to personalize interventions. Doing so would both reduce user burden and increase opportunities to be more context aware.

But the opportunity to use more passively-sensed data to build context-aware behavior change interventions raises many questions that I have grappled with in my own work, and that are critical to address as we seek to be more context-sensitive in our efforts to support people to live as healthy and well as possible. The questions raised are ones that get to the heart of the line between insight and intrusion. On the one hand, as a behavioral scientist who 1) builds digital behavior change interventions and 2) knows that accounting for context positions any digital health intervention to have a higher probability of success, I want to use any and every piece of data to make inferences about context in as close to real-time as possible so as to make the digital interventions I build entirely context-aware and hyper-personalized. Doing so includes embedding the process of making inferences based on actively-reported passively-sensed data into a learning system, so that users have an opportunity to validate the accuracy (or error) of the algorithms that drive personalization of when and how interventions are delivered.

On the other hand, I have concerns about making use of passively-sensed data in these ways for a few reasons: does doing so promote or threaten autonomy, competency, and relatedness? These are the core tenants of self-determination theory, a theory of intrinsic motivation that highlights choice, mastery, and connection (three synonyms, respectively, for the core tenants listed above) as paramount to supporting intrinsic motivation in the service of behavior change. If passively sensed data can be used to promote user insight, then I feel more confident that we’re working towards autonomy, competency, and relatedness. If not, then the scale is tipped towards intrusiveness, and not only will the opportunity to make use of context in a behavior change intervention be squandered, worse, trust could be damaged as well.

In a world where we know the data we generate are already being used in multiple ways – and in ways that we’ve often likely consented to in any number of “Terms and Conditions” that we agreed to without reading – here are some guideposts to be responsible stewards of both actively-reported and passively-sensed data as we seek to account for context in our digital behavior change interventions:

  1. Be clear about what data are being collected and how they’re being used, and communicate that to users early and often: Co-design such efforts with your target population to help ensure that your efforts to do so are successful.
  2. Do due diligence during research and development to achieve maximum data efficiency: Collect the minimum data (actively-reported or passively sensed) needed to make the maximally accurate contextual inferences.
  3. Embed these efforts that leverage connectivity in a sense of connection: Challenge yourself to put connection at the heart of your work; meaning, never lose sight of how your digital behavior change intervention is a service – a service intended to connect a person who has a health behavior change goal to a service that has been designed to help them achieve it.

These guideposts certainly do not fully address the full spectrum of privacy issues that are necessary to account for in digital health, but we would do well do keep them in mind as we manage the tension between insight and intrusion. The imperative to improve the health of individuals demands that we probe this tension, and for behavioral scientists, our history of understanding people suggests we can be successful.

Ellen Beckjord, PhD, MPH is the Associate Vice President for Population Health and Clinical Affairs at UPMC Health Plan, Insurance Services Division.

Ellen Beckjord also examines the Anatomy of a Great Health App
In this CHC19 session, Ellen Beckjord joins panelists to discuss some of the most frequently downloaded health apps to help tease out the features that contribute to measures of success.

Hear Ellen at CHC

Secure Your Seat

Be a part of the conversation!


Register Today