Towards an AI-enabled and Emotion-Adaptive Persuasive Technology for Improving Mental Health and Well-being

Abstract

Emotion regulation (ER) is crucial for psychological well-being, as negative emotions can lead to mental health issues. However, little is known about personalizing persuasive technologies (PTs) to deliver just-in-time ER interventions based on individuals' emotional states. To address this gap, I conducted systematic reviews of literature and apps to examine the implementation of persuasive strategies (PS) in research and real-world health apps, the AI models used in adaptive health systems, and ER strategies for digital interventions. I then performed two large-scale empirical studies (N=568 and N=660) to establish guidelines for tailoring PTs by selecting appropriate PS to enhance effectiveness and support ER. Additionally, I developed and compared 21 machine learning models for detecting emotional states in daily journals, with the MCBiLSTM model achieving the best performance. Based on these findings, I designed and developed "Recilify," an AI-enabled, emotion-adaptive PT that delivers personalized ER interventions to improve mental health and well-being.