Eugenia Rho, an assistant professor in the Department of Computer Science, has conducted extensive research at the Society + AI & Language Lab, revealing that police language plays a significant role in predicting violent interactions with Black motorists. She has also highlighted the dangers posed to American democracy by broadcast media bias and social media echo chambers. Now, Rho’s team has shifted their focus to examining the impact of social media rhetoric on COVID-19 infection and death rates in the United States, with the aim of providing insights for policymakers and public health officials.
During the COVID-19 pandemic, social media became a breeding ground for opposition to public health guidance, leading to widespread misinformation and a disregard for preventive measures such as mask wearing, social distancing, and vaccines. This resulted in soaring infection rates, overwhelmed hospitals, healthcare worker shortages, preventable deaths, and economic losses. A study published in the Yale Journal of Biology and Medicine reported over 692,000 preventable hospitalizations costing $13.8 billion among unvaccinated patients between November and December 2021.
Rho’s team utilized the chatbot GPT-4 to analyze posts in banned subreddit discussion groups that opposed COVID-19 prevention measures on Reddit. They developed a technique grounded in Fuzzy Trace Theory, pioneered by collaborator Valerie Reyna, to identify cause-and-effect relationships in social media discourse. By tracking gists across social media discussions, the researchers aimed to predict patterns in online engagement and national health outcomes related to COVID-19.
The research identified significant links between social media posts containing gists and real-world public health trends, showcasing the potential of large language models to enhance public health communication strategies. AI-driven analysis of linguistic patterns in social media discussions allowed for the prediction of total and new daily COVID-19 cases in the U.S. based on the volume of gists in banned subreddit groups. This groundbreaking study opens up new possibilities for leveraging AI technology to inform health communication strategies.
Rho hopes that this study will inspire other researchers to apply similar methods to address pressing issues. The team plans to release the code used in the project for free when the paper is published, making it accessible to a wider audience. By presenting their findings at the Proceedings of the Association of Computing Machinery Conference on Human Factors in Computing Systems, Rho aims to spark further discussion and exploration in the realm of AI-driven research for public health.
Beyond academia, Rho also hopes that this research will prompt social media platforms and other stakeholders to find alternative approaches to addressing controversial topics online. She warns against simply deleting or banning groups discussing such topics, as this may drive individuals deeper into conspiracy theories and less moderated platforms. By working collaboratively with public health officials and organizations, social media companies can better engage with and understand the public’s perspectives during public health crises, ultimately enhancing communication and response strategies.