When AI Joins the Social Media Conversation: Exploring the Impact of Simulated AI-Assisted Comments on Health Risk Perceptions and Behaviors.

Abstract

Misinformation on social media has long been a pressing concern, and its spread is often exacerbated by social endorsement metrics such as likes and shares. While prior research has explored artificial intelligence (AI)-based strategies to combat misinformation, little attention has been given to how AI-powered functions (e.g. automated comments) shape user perceptions and behaviors in the context of social media misinformation. This study employs a 2 (misinformation social endorsement: high vs. low) × 3 (AI-suggested comments: approve misinformation vs. disapprove misinformation vs. no AI comments) between-subject experimental design, focusing on a health issue heavily affected by misinformation - HPV vaccination. The findings reveal that misinformation posts perceived as highly endorsed, either through social endorsement or AI-generated supportive comments, are viewed as most credible. Additionally, AI comments endorsing misinformation evoke greater fear under conditions of high social endorsement compared to the absence of AI comments. Conversely, AI comments challenging misinformation reduce the credibility of misinformation and mitigate fear surrounding HPV vaccination, but only under conditions of low social endorsement, relative to no AI comments. Credibility and fear emerge as critical mediators, influencing subsequent perceived risk of HPV vaccination and vaccination intentions. Theoretical and practical implications of these findings are discussed, shedding light on the nuanced interplay between AI interventions and social endorsement in addressing misinformation.