Acceptability of Using Large Language Models to Support Smoking Cessation Attempts: a Qualitative Study Among African...

Area of Science:

• Digital Health
• Health Disparities
• Artificial Intelligence in Healthcare

Background:

• African Americans experience significant disparities in tobacco-related diseases and smoking cessation resources.
• Digital health interventions, particularly those using large language models (LLMs), show potential for improving engagement and quit attempts.
• There is a lack of understanding regarding the acceptability of LLM-based interventions among African Americans, highlighting the need for tailored approaches.

Purpose of the Study:

• To evaluate the perceptions of African Americans regarding LLM-based interventions.
• To inform the development of an LLM-based chatbot designed to support smoking cessation attempts within this community.

Main Methods:

• Remote focus groups were conducted with adult African American cigarette smokers in Minnesota.
• Discussions explored the acceptability of LLM-based health interventions and chatbot functionalities.
• Focus group transcripts were independently coded by two analysts using a shared codebook.

Main Results:

• Twenty-one adults participated, with 57.1% identifying as female and a mean age of 45.6 years.
• Key themes included limited familiarity with LLMs/AI, a willingness to use LLM apps for quitting, and specific feature perceptions influencing chatbot acceptability.
• Participants expressed a need for user-friendly and trustworthy AI health tools.

Conclusions:

• African Americans hold nuanced views on LLMs for health interventions, recognizing potential benefits alongside knowledge gaps and concerns.
• Building trust and acceptability requires prioritizing community engagement and ethical considerations in LLM intervention development.
• LLM-driven health solutions must be inclusive and effective across diverse communities.