Who Believes and Who Shares Fake News: Experimental Misinformation Research with LLM-based Agents
Speaker: Linette Lim (University College Dublin) with (opens in a new window)Yen-Chieh Liao (University of Birmingham)
Wednesday, March 5, 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: In the age of digital communication, the lines between factual reporting and misinformation have increasingly blurred, providing openings for nefarious actors to spread disinformation. In this paper, we implement multi-agent simulation with large language models (LLMs) to study decision-making under conditions of misinformation exposure in the election context of Taiwan. Using pre-registered surveys with LLM-powered agents, we examine two research questions: (1) who believes and shares fake news, and (2) what is the impact of fact-checking interventions on misinformation spread and agent perceptions? Drawing on existing research, we focus on how agents' characteristics, particularly their education levels, economic status, media literacy, and partisanship, affect their susceptibility to misinformation, and propensity to disregard fact-checking interventions and share false content. We implement a cutting-edge multi-agent simulation framework using AutoGen and powered by both open-source and commercial LLMs, establishing a novel prototype that sets the foundation for future experimental survey methodologies. This paper combines survey experiments with multi-agent simulations to examine the effectiveness of fact-checking interventions and how individual characteristics and agent interaction shape misinformation dynamics.
About the speaker: Linette Lim is a PhD student at the School of Politics and International Relations. Her research encompasses authoritarian politics and media control in Asia, and Chinese nationalism. Prior to coming to UCD, she was the Shanghai-based correspondent with the Singapore public broadcaster, covering economic and political news in Greater China. Key assignments included the annual "Two Sessions" meetings in Beijing, the anti-government protests in Hong Kong in 2019, and the Taiwan elections in 2020. Outside of journalism, she has held roles in corporate communications, dealing with media and government officials. She completed her undergraduate and postgraduate studies at the Singapore Management University and King’s College London respectively.
Yen-Chieh Liao is a Research Fellow in Natural Language Processing at the Centre for Artificial Intelligence in Government (CAIG) and the School of Government at the University of Birmingham. Before joining Birmingham, he was a postdoctoral researcher at the School of Politics and International Relations at University College Dublin and a Pre-doctoral Researcher at the Chair of Comparative Politics at the University of Bamberg. He earned his PhD in Politics from the University of Essex in January 2023, after obtaining a BA and MA in Public Administration & Policy from National Taipei University in Taiwan.