AI for Digital Intervention Optimization
Susan A. Murphy

Artificial Intelligence (AI) methods can be used in a variety of ways to contribute to digital health intervention optimization. These include online continual updating of predictions of risk and receptivity, creation and updating of digital twins for use in the early stages of intervention optimization and for continually updating decision rules that link an individual's current internal state and external context to intervention options. This presentation will touch on the former two and focus on the continual updating of decision rules with the goal of delivering intervention options when they are most effective, yet managing burden.

Personalizing just-in-time adaptive interventions and micro-randomized trials
Devin Tomlinson

A personalizing just-in-time adaptive intervention (pJITAI) is an intervention design that uses artificial intelligence (AI) to learn and update decision rules based on when participants are most likely to engage with and respond to interventions. In a pJITAI, decision rules are adapted over time based on ongoing information collected throughout the intervention to learn about an individual’s responsivity to decision rules, allowing for more effective and personalized support. The micro-randomized trial (MRT) is an experimental design that includes rapid sequential randomizations to allow investigators to answer questions about how to best deliver and adapt digitally-delivered interventions in the real world. In this talk, we will discuss the MiWaves project, a pilot MRT of a pJITAI for reducing cannabis use among emerging adults. In the MiWaves study, emerging adults were invited to complete twice-daily surveys followed by twice-daily randomizations to an intervention message (or no message), where probabilities of message delivery were determined by a reinforcement learning (RL) algorithm. The acceptability and feasibility of MiWaves, and analysis of data from the MRT, will be discussed.