Large Language Model Chatbot Conversations vs Public Health Materials and Parental HPV Vaccination Intentions: A...

Area of Science:

• Public Health
• Health Communication
• Artificial Intelligence in Healthcare

Background:

• Health systems are exploring large language model (LLM)-based chatbots for vaccine communication.
• Evidence for LLM chatbots improving lasting, behaviorally relevant outcomes beyond traditional health materials is scarce.
• Human papillomavirus (HPV) vaccination is crucial for preventing related cancers.

Purpose of the Study:

• To evaluate if brief LLM chatbot interactions increase parental intention to vaccinate children against HPV.
• To compare chatbot effectiveness against no intervention and existing government public health materials.
• To assess the persistence of any observed effects on vaccination intent and uptake.

Main Methods:

• A randomized clinical trial involving 1297 parents in the US, Canada, and UK.
• Participants were assigned to a no-message control, government public health materials, or a 3-minute LLM chatbot interaction (GPT-4o).
• Outcomes included immediate and 45-day vaccination intent and self-reported vaccination uptake.

Main Results:

• Both public health materials and LLM chatbots increased immediate HPV vaccination intent compared to no intervention.
• At 45 days, only public health materials maintained modest effects on intent; chatbots did not show sustained benefits.
• No intervention group demonstrated increased self-reported vaccination uptake.

Conclusions:

• Brief LLM chatbot interactions may offer a short-term boost in HPV vaccine intent but do not yield lasting effects.
• Standard public health messaging may be as effective or more effective than short chatbot conversations for HPV vaccine promotion.
• Further research is needed to optimize AI tools for durable behavioral health outcomes in vaccination.