A Motivational Interviewing Chatbot With Generative Reflections for Increasing Readiness to Quit Smoking: Iterative...

Area of Science:

• Digital Health
• Behavioral Science
• Natural Language Processing

Background:

• Motivational interviewing (MI) is effective for smoking cessation but limited in accessibility.
• Text-based chatbots traditionally use scripted responses, reducing MI effectiveness.
• Natural language processing (NLP) enables generative language models for personalized chatbot responses.

Purpose of the Study:

• To design, evolve, and assess an MI chatbot using generative reflections for ambivalent smokers.
• To guide smokers toward quitting by leveraging NLP for tailored conversational support.

Main Methods:

• Four MI chatbot versions were developed and tested on 349 ambivalent smokers.
• Chatbot versions varied in question-asking, reflection generation (basic to advanced), and interaction length.
• Participant readiness, confidence, and importance to quit were measured pre- and post-conversation; quit attempts and perceived empathy were also surveyed.

Main Results:

• Generative reflection versions showed increased confidence compared to the non-generative version.
• The chatbot with enhanced reflections and extended interaction significantly improved readiness and importance to quit.
• Higher perceived empathy was reported for chatbots with enhanced reflections and extended conversations.

Conclusions:

• Generative reflections enhance chatbot conversations for smoking cessation, increasing readiness to quit.
• Even basic chatbot interactions without reflections show some positive impact.
• Further development of generative AI chatbots is warranted for smoking cessation interventions.