Job Description

Misinformation, encompassing false, inaccurate, and misleading information, profoundly impacts both individuals and society. As a longstanding focus of public discourse and scientific research, the frontiers of misinformation debunking frameworks have evolved from train-from-scratch neural detectors to powerful agents empowered by Large Language Models (LLMs). However, three key challenges persist in deploying effective misinformation debunking and mitigation systems: (1) Reliable debunking with LLM-empowered systems requires the LLM backbones to reason about statement truthfulness and flag statements lacking sufficient context, which necessitates methods to reduce LLM hallucinations and improve their uncertainty estimation capabilities. (2) Real-world misinformation varies in type (e.g., fake news, clickbait, and spam) and data modality (e.g., text, vision, and tabular data), which requires automated general-purpose debunking systems. (3) Effective misinformation mitigation in digital environments requires in-depth understanding of the social dynamics in misinformation consumption and propagation.In light of these challenges, our research vision is to enhance the factuality and reliability of LLMs, utilize them as versatile debunking agents to combat diverse misinformation types, and implement personalized mitigation strategies to foster a more trustworthy digital landscape.RequirementsThe candidate must possess a PhD in Computer Science and have experience in fake news detection. Ideally, he or she will have worked on relevant research for real-world generalizability challenges for debunking misinformation, which may include work along any of the following directions (including but not limited to):

• Generalizability across newsworthy events.
• Generalizability across social graphs.
• Generalizability across writing styles.

National University of Singapore