Brown School Study Uses AI to Analyze Social Media Attitudes About Soda Taxes 3/23/2023 Research; Faculty Share this Story: Page Image (Photo: Shutterstock) Brown Page Content 1Researchers at the Brown School used artificial intelligence (AI) to analyze hundreds of thousands of tweets to assess people’s attitudes toward soda taxes. Results were published recently in the Journal of Public Health Management and Practice. The study was led by Ruopeng An, associate professor at the Brown School, who said the research demonstrated the value of AI technologies in analyzing large quantities of data. His team designed a comprehensive search algorithm to systematically identify soda-tax-related tweets. “AI enables computer modeling that is useful in analyzing ‘big’ data that is massive in scale and messy to work with, such as images or unstructured texts like those in tweets,” he said. An’s team identified and collected around 370,000 soda tax-related tweets posted between 2015-2022 and built AI models to classify those tweets by sentiments. They found that public attention paid to soda taxes peaked in 2016 and has since declined considerably. Soda taxes were proposed as a way to discourage soda consumption and combat obesity by taxing sugar-sweetened drinks. The finalized neural network model achieved an accuracy of 88% in predicting tweet sentiments in the test set. “Social media has the power to shape public opinion and catalyze social changes, but remains an underutilized source of information to inform government decision-making,” An added. “Modern social media sentiment analysis powered by AI technologies may inform the design, implementation, and modification of soda tax policies to gain social support while minimizing confusion and misinterpretation.” Traditional polls and surveys are costly, offer little historical data or significant time delay. Social media sentiment analysis enables collecting and analyzing large volumes of social media data in near real-time, unrestrained by geographical boundaries, and at a fraction of the cost of those traditional methods. An chairs the Artificial Intelligence Applications for Health Data Post-Masters Certificate and the Artificial Intelligence and Big Data Analytics for Public Health Certificate programs at WashU and teaches several graduate-level artificial intelligence and data science courses. The three coauthors of the study, Yuyi Yang, Quinlan Batcheller, and Qianzi Zhou, are Master of Public Health students at the Brown School.