Generative Experiences for Digital Mental Health Interventions: Evidence from a Randomized Study
Researchers have introduced a new paradigm called 'generative experience' for digital mental health (DMH) tools, focusing on how support is experienced rather than just what content is provided. This approach is instantiated in GUIDE, a system that dynamically composes personalized intervention content and multimodal interaction structures at runtime using rubric-guided generation. A preregistered randomized study involving 237 participants demonstrated that GUIDE significantly reduced stress levels (p = .02) and improved user experience (p = .04) compared to a control group using standard LLM-based cognitive restructuring. The system supported diverse forms of reflection and action through varied interaction flows, although it also revealed tensions regarding personalization across interaction sequences. This work establishes a foundation for DMH interventions that can dynamically shape the enactment and experience of support in digital settings, addressing the limitation where well-matched content fails due to misaligned interaction formats. The study highlights the potential of generative AI to enhance the efficacy and engagement of digital mental health solutions by adapting not only the message but also the mode of delivery to individual user needs.
Wire timeline
Generative Experiences for Digital Mental Health Interventions: Evidence from a Randomized Study
Researchers have introduced a new paradigm called 'generative experience' for digital mental health (DMH) tools, focusing on how support is experienced rather than just what content is provided. This approach is instantiated in GUIDE, a system that dynamically composes personalized intervention content and multimodal interaction structures at runtime using rubric-guided generation. A preregistered randomized study involving 237 participants demonstrated that GUIDE significantly reduced stress levels (p = .02) and improved user experience (p = .04) compared to a control group using standard LLM-based cognitive restructuring. The system supported diverse forms of reflection and action through varied interaction flows, although it also revealed tensions regarding personalization across interaction sequences. This work establishes a foundation for DMH interventions that can dynamically shape the enactment and experience of support in digital settings, addressing the limitation where well-matched content fails due to misaligned interaction formats. The study highlights the potential of generative AI to enhance the efficacy and engagement of digital mental health solutions by adapting not only the message but also the mode of delivery to individual user needs.
cs.AI updates on arXiv.org