At a time of rampant medical misinformation online, artificial intelligence can do a good job of correcting falsehoods, but the tone of the correction is key, according to new research led by Washington State University.
For people who view AI strictly as a technical tool, corrections to social media posts offered in a neutral, just-the-facts tone are most persuasive. For those who believe AI can be humanlike, an empathetic, understanding tone is best.
In fact, according to the findings published in the International Journal of Human-Computer Interaction, tone was more important than the source of the correction itself.
“In multiple studies, our team has found that corrections could work most times, but the tone of the correction is important,” said Porismita Borah, professor in WSU’s Edward R. Murrow College of Communications and corresponding author of the new publication. “In this study, it did not necessarily matter whether the correction came from a human being or an AI agent. What mattered was the tone and how the tone aligned with people’s beliefs about whether AI agents should be more humanlike or more machinelike.”
The findings suggest that social media platforms, government agencies, news organizations and others could target AI fact-checking agents toward a user’s level of anthropomorphism — the tendency to assign human characteristics to animals, machines and other non-human entities. They could design a simple onboarding step for users of platforms or accounts to assess their anthropomorphic tendencies and adjust the AI’s conversational tone accordingly, the authors said.
Borah’s co-authors were Ziyao Zhang, a PhD student at WSU; Xiaohui Cao, a PhD student at the University of Wisconsin-Madison; and Danielle Ka Lai Lee, an assistant professor at Hong Kong Shue Yan University.
It takes more than facts to persuade someone that a claim is incorrect. People may feel challenged or insulted and react defensively when corrected. They may question or doubt the source of the correction. As social media has proliferated, fueling the spread of dubious health claims, researchers have increasingly sought to better understand how to effectively combat misinformation.
Borah has been studying the subject for a decade. Her current publication advances the understanding of how empathy plays a role in corrections. Other researchers have come to conflicting conclusions about whether an empathetic tone in correction messages is effective in decreasing misperceptions.
Borah’s team added another layer to the question, testing the effectiveness of tone against the expectations and beliefs of the person being corrected. They conducted a randomized online experiment with 857 parents of children in the age range recommended to receive the vaccine for the human papillomavirus, or HPV.
Related on the SLP
The virus is spread through sexual contact and can cause a variety of health problems, including cancers. The HPV vaccine is considered safe and effective, but has been the subject of widespread misinformation.
Survey participants were evaluated for their level of anthropomorphism belief, then shown a simulated Facebook comment thread that began with a false claim: “HPV vaccines increase the risk of neurological problems.” An AI corrections account engaged with the claim in the comment thread.
The neutral answers used direct, plain language: “That’s not true. Scientific studies have shown no link between HPV vaccines and any of those scary neurological conditions.”
The empathetic tone was warmer, such as: “I hear you, but scientific studied have shown….”
The correction was most effective at reducing misperceptions when the tone matched a respondents’ anthropomorphism beliefs.
The findings add to the growing understanding of the complexity of persuasion and correcting misinformation.
“The problem of misinformation is critical, and it’s not going away,” Borah said. “The effectiveness of corrections depends on a lot of factors —for example the way you talk to someone when providing accurate information — an empathetic tone may often work better than a condescending one. Race, gender, and other factors also matter. We’re ultimately trying to study humans — and humans are remarkably complex.”
