Public Opinion can be Estimated and Influenced via Social Media Platforms like X
Speaker: (opens in a new window)Przemyslaw Grabowicz (University College Dublin)
Wednesday, February 19, 14:00–14:45 (Irish time)
Please register (opens in a new window)here to receive the link and password to the online meeting and information on the room at UCD.
Abstract: Informal political polls have surged in popularity on X (formerly Twitter). In the lead-up to the 2024 US presidential election, Donald Trump and Elon Musk prominently featured the results of such polls, likely to project an image of Trump's overwhelming popularity. Our platform, (opens in a new window)socialpolls.org, tracked and corrected biases daily in thousands of US presidential election polls published on X. Remarkably, the platform's forecasts of the election outcome achieved a 1% error at the national level, outperforming traditional polling methods. Our bias-corrected time series of public opinion reveal significant trends and provide insights into major shifts, e.g., highlight a potential reason for Joe Biden's decline in the presidential race. To this end, we developed and applied advanced AI tools, including a classifier based on GPT-4o identifying relevant polls and a multimodal demographic inference model. Our findings underscore the value of X as a distinctive source of public opinion data. In addition, I will present preliminary evidence suggesting that X not only reflects but also influences public opinion, and describe briefly political biases on X before the upcoming German federal election.
About the speaker: Przemek is an Assistant Professor of Computer Science at the University College Dublin and an Adjunct Professor of Computer Science at the University of Massachusetts Amherst. He heads the (opens in a new window)SIMS research lab, which focuses on research at the intersection of technology and society. Among others, the SIMS lab develops novel methods for public opinion mining based on social media polls and advanced AI (see (opens in a new window)socialpolls.org). Overall, Przemek's research is focused on responsible AI, computational social science, social media, news media, applied NLP, and open-world learning.