Abstract: In the US, unhealthy behaviors—such as sedentary lifestyle, overeating, substance use, and tobacco use—account for approximately 40% of the risk of premature deaths. While successful changes to these unhealthy behaviors can mitigate the risk of harm, behavior change is often difficult because of the high self-management burden. Fortunately, mobile phones and wearable devices may reduce the burden of self-management. Apps on phones or wearables can continuously monitor user behavior and use onboard computation to deliver the right intervention at the right time to make self-management easy.
Despite their promise, health apps have largely been ineffective at driving meaningful behavior change. This shortcoming partly stems from the research community’s emphasis on predicting health and well-being indicators without modernizing the interventions themselves. As a result, most interventions remain generic, non-adaptive, and ineffective. My research addresses this gap by modernizing interventions in three key ways. First, I design personalized intervention content that leverages dense mobile data and AI to generate actionable suggestions aligned with a user’s health goals and daily routines. Second, I apply sequential experimentation and reinforcement algorithms to adapt interventions to the right context. This approach ensures interventions are delivered when effective, while also revealing gaps that prompt the testing of new strategies. Finally, I integrate these interventions into healthcare delivery systems, where I work directly with patients, provide support between clinic visits, reduce clinicians’ workload during appointments, and deliver valuable insights to inform clinical decision-making. In this talk, I will present two projects on personalized intervention content generation. The first, MyBehavior, provides individualized, low-burden suggestions for improving physical activity and dietary intake. The second, the sub-goal app, breaks down daily goals into personalized sub-goals and suggests changes during periods when users are typically active. I will also briefly discuss several novel interventions developed and evaluated through sequential experimentation, along with a dashboard that integrates visualization and AI to support reliable, low-burden clinical decision-making. I will conclude by outlining my research agenda for mobile health interventions and how these approaches can be effectively integrated into traditional models of care..
Bio: Mashfiqui Rabbi is an assistant research professor at the Siebel School of Computing and Data Science. Previously he was a senior research scientist at Optum AI under UnitedHealth Group, the largest payer and one of the largest providers in US healthcare. Before Optum, Mashfiqui was a research scientist at Apple Health AI and a postdoctoral fellow at Harvard University, where he worked with Professor Susan Murphy. Before his postdoc, he received a Ph.D. in Information Science from Cornell University. His Ph.D. advisor was Professor Tanzeem Choudhury. Mashfiqui’s Ph.D. thesis led to the creation of the MyBehavior app, the first mobile recommender system to automatically generate personalized physical activity and food suggestions from mobile phone data. In his postdoc at Harvard, Mashfiqui created the first just-in-time intervention for improving health app engagement. This engagement intervention was later adopted by multiple NIH-funded grants focusing on youth substance abuse, cancer rehabilitation, and sickle cell disease. Mashfiqui also helped establish the AI-based intervention group at Apple’s Health AI team, and his sub-goal app was adopted for the customized plan feature in Apple’s Fitness+ app, which is currently used by millions of iPhone users. Mashfiqui’s work has also been featured in MIT Technology Review, New Scientist, the Economist, Mashable, and the NY Times.