Skip to main content
ABCT 2023 Annual Convention Main banner
Final Program


Program Flipbook
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Presentation Icons
Ticketed Session
Ticketed Session
Recorded Session
Recorded Session
Back
ViewAttendees 5

Symposia

Technology

Session: Using machine learning in scalable, adaptive digital mental health interventions

2 - (SYM 99) Trials and Tribulations: Ensuring the Miwaves RL Algorithm Is Ready for Implementation!

Saturday, November 18, 2023
4:15 PM – 5:45 PM PST
Location: Chiliwack (402), Level 4

Recent advances in digital technology offer much potential for promoting healthy behaviors in real-time, real-world settings.  At the same time advances in artificial intelligence algorithms offer an approach to personalizing mobile health interventions to maximize impact.  MiWaves is a digital intervention aimed at reducing cannabis use among young adults who use cannabis regularly (at least 3x/week) with some motivation to change.   MiWaves integrates empirically based intervention strategies for reducing substance use (e.g., motivational interviewing, behavioral economics) with engagement strategies grounded in psychology, human-computer interaction, and marketing.  Many of these strategies are delivered via smartphone prompts.    Since excessive delivery of prompts can induce habituation and disengagement, we are developing a MiWaves reinforcement learning (RL) algorithm that will, as each participant experiences MiWaves, personalize whether to send or not send prompts depending on each participant’s current setting (i.e., recent cannabis use, recent engagement).  Here we describe the multiple stages of development of an online RL algorithm for learning whether and under what conditions, sending a prompt benefits intervention engagement. Challenges in designing and testing an RL algorithm in these real-life health settings include developing a simulation testbed on impoverished data, ensuring the RL algorithm can learn and run stably under a variety of constraints, e.g., avoid overburdening participants and ensuring the resulting data can be used in standard statistical analyses.  We will present feasibility results from beta testing of MiWaves.   The MiWaves study, with RL algorithm, is to start in December, 2023.