Expert, Peer, or Future-Self? Exploring Relational Roles in Smoking Cessation Conversations

Abstract

Engaging smokers with smoking cessation support remains challenging, particularly during early stages of readiness to quit, where psychological, social, and structural frictions often prevent participation in existing interventions. Inspired by prior work on face-to-face brief advice delivered at smoking hotspots, this exploratory study investigates how conversational agents powered by large language models (LLMs) could support such smoking cessation engagement. We evaluate three relational agent roles (expert, peer smoker, and future-self) delivering reflective, non-clinical smoking cessation conversations inspired by principles of the 5A framework. Early-stage evaluations were conducted with four healthcare professionals and six smokers to assess perceptions of empathy, credibility, engagement, and appropriateness across the agents. Preliminary findings highlight divergent preferences regarding relational framing, along with challenges related to role consistency and emotional resonance. We conclude with design implications for future situated LLM-based smoking cessation interventions.