Paper Type
Complete
Paper Number
PACIS2025-1486
Description
In this systematic review of large language model (LLM)-based mental health interventions, we focus on LLM’s effectiveness in promoting behavior change and improving mental health outcomes through a persuasive system design (PSD) lens. Highlighting the need for scalable, inclusive, and personalized solutions, we examine the state of research on LLM-based interventions that promote behavior change and the implementation of PSD principles in such interventions. Our findings reveal that these interventions demonstrate both direct impacts, such as symptom reduction and emotional regulation, and indirect impacts, including fostering therapeutic alliances and enhancing reflective practices. Also, studies achieving strong indirect impacts combined with clinically significant and measurable direct outcomes, on average, implemented more PSD principles than those demonstrating significant indirect impacts but with limited or moderate direct impacts. However, significant gaps remain in the representation of long-term studies, cultural inclusivity, and ethical considerations.
Recommended Citation
Giri, Sarthak; Kulasinghe, Kavinda; and Oinas-Kukkonen, Harri, "Exploring Large Language Model-Based Mental Health Interventions: A Systematic Review with a Persuasive System Design Lense" (2025). PACIS 2025 Proceedings. 9.
https://aisel.aisnet.org/pacis2025/emerg_tech/emerg_tech/9
Exploring Large Language Model-Based Mental Health Interventions: A Systematic Review with a Persuasive System Design Lense
In this systematic review of large language model (LLM)-based mental health interventions, we focus on LLM’s effectiveness in promoting behavior change and improving mental health outcomes through a persuasive system design (PSD) lens. Highlighting the need for scalable, inclusive, and personalized solutions, we examine the state of research on LLM-based interventions that promote behavior change and the implementation of PSD principles in such interventions. Our findings reveal that these interventions demonstrate both direct impacts, such as symptom reduction and emotional regulation, and indirect impacts, including fostering therapeutic alliances and enhancing reflective practices. Also, studies achieving strong indirect impacts combined with clinically significant and measurable direct outcomes, on average, implemented more PSD principles than those demonstrating significant indirect impacts but with limited or moderate direct impacts. However, significant gaps remain in the representation of long-term studies, cultural inclusivity, and ethical considerations.
Comments
Technologies