AI Corrects Medical Misinformation Best With Right Tone

Recent research from Washington State University demonstrates that artificial intelligence can effectively counter medical misinformation, but success hinges on matching the correction tone to the recipient's perception of AI. The study, published in the International Journal of Human-Computer Interaction, was led by communications professor Porismita Borah alongside co-authors Ziyao Zhang, Xiaohui Cao, and Danielle Ka Lai Lee. To evaluate the efficacy of AI-driven corrections, the research team conducted a randomized online experiment involving 857 parents of children eligible for the human papillomavirus vaccine. Participants were exposed to simulated social media threads containing false claims about HPV vaccine safety. An AI agent responded with either a neutral, fact-based tone or a warmer, empathetic tone. Before interacting with the agent, users were assessed on their level of anthropomorphism, which measures the tendency to attribute human characteristics to nonhuman entities. The findings revealed that tone alignment with user expectations consistently outperformed the identity of the correction source. Participants who viewed AI strictly as a technical tool responded best to direct, unembellished factual statements. Conversely, individuals who perceived AI as humanlike were more persuaded by empathetic, understanding language. The research team concluded that the psychological framing of the AI interaction matters more than whether the information originated from a human or an algorithm. These results carry significant implications for digital platforms, government health agencies, and media organizations currently deploying automated fact-checking systems. The authors recommend implementing brief onboarding assessments to gauge users' anthropomorphic tendencies. Based on these profiles, AI moderation tools could dynamically adjust their conversational parameters to maximize persuasiveness and minimize defensive reactions. This adaptive approach addresses a longstanding challenge in combating online falsehoods: recipients often react defensively to corrections, questioning the source or feeling insulted, which reinforces their initial misconceptions. While prior studies have produced conflicting conclusions regarding the utility of empathy in correction messaging, this research clarifies that effectiveness is highly contextual. The study underscores the complexity of digital persuasion, highlighting that race, gender, cognitive biases, and anthropomorphic beliefs all intersect with AI communication strategies. As medical misinformation continues to proliferate across social networks, algorithmic interventions must evolve beyond binary fact-checking to incorporate nuanced, psychologically aware communication frameworks. Platform developers and policy makers are encouraged to prioritize adaptive tone modulation in next-generation AI moderation tools to sustain public trust and improve health literacy outcomes.