Using Adaptive Bandit Experiments to Increase and Investigate Engagement in Mental Health

• DOI: 10.1609/aaai.v38i21.30328
• Publication Date: 2024-03-25T12:49:54Z
• Authors: Harsh Kumar; Tong Li; Jiakai Shi; Ilya Musabirov; Rachel Kornfield; Jonah Meyerhoff; Ananya Bhattacharjee; Chris Karr; Theresa Nguyen; David Mohr; Anna Rafferty; Sofia Villar; Nina Deliu; Joseph Jay Williams
• Keywords: Testbed

Digital mental health (DMH) interventions, such as text-message-based lessons and activities, offer immense potential for accessible support. While these interventions can be effective, real-world experimental testing further enhance their design impact. Adaptive experimentation, utilizing algorithms like Thompson Sampling (contextual) multi-armed bandit (MAB) problems, lead to continuous improvement personalization. However, it remains unclear when simultaneously increase user experience rewards facilitate appropriate data collection social-behavioral scientists analyze with sufficient statistical confidence. Although a growing body of research addresses the practical aspects MAB other adaptive algorithms, exploration is needed assess impact across diverse contexts. This paper presents software system developed over two years that allows text-messaging intervention components adapted using while collecting side-by-side comparison traditional uniform random non-adaptive experiments. We evaluate by deploying DMH 1100 users, recruited through large non-profit organization, share path forward this at scale. not only enables applications in but could also serve model testbed experimentation domains.