[Paper Review] Combating Misinformation in the Age of LLMs: Opportunities and Challenges

Canyu Chen, Kai Shu|arXiv (Cornell University)|Nov 9, 2023

Misinformation and Its Impacts20 citations

TL;DR

A comprehensive survey on using large language models (LLMs) to combat misinformation, outlining opportunities, challenges, and future directions for detection, intervention, and attribution of both human- and LLM-generated misinformation.

ABSTRACT

Misinformation such as fake news and rumors is a serious threat on information ecosystems and public trust. The emergence of Large Language Models (LLMs) has great potential to reshape the landscape of combating misinformation. Generally, LLMs can be a double-edged sword in the fight. On the one hand, LLMs bring promising opportunities for combating misinformation due to their profound world knowledge and strong reasoning abilities. Thus, one emergent question is: how to utilize LLMs to combat misinformation? On the other hand, the critical challenge is that LLMs can be easily leveraged to generate deceptive misinformation at scale. Then, another important question is: how to combat LLM-generated misinformation? In this paper, we first systematically review the history of combating misinformation before the advent of LLMs. Then we illustrate the current efforts and present an outlook for these two fundamental questions respectively. The goal of this survey paper is to facilitate the progress of utilizing LLMs for fighting misinformation and call for interdisciplinary efforts from different stakeholders for combating LLM-generated misinformation.

Motivation & Objective

• Provide a historical overview of misinformation mitigation before LLMs, with a focus on detection.
• Analyze opportunities and challenges of employing LLMs to combat misinformation.
• Explore LLM-assisted misinformation intervention and attribution, and potential multimodal/human collaboration avenues.
• Identify risks and necessary interdisciplinary actions to address LLM-generated misinformation.

Proposed method

• Systematic literature review of misinformation detection methods prior to LLMs across linguistic, neural, social-context, external-knowledge, generalization, supervision, multilingual and multimodal approaches.
• Synthesis of how LLMs can enhance detection, intervention, and attribution, including the use of external knowledge, tools, multimodal data, and autonomous agents.
• Discussion of trustworthiness, robustness, explainability, fairness, privacy, and transparency in LLM-enabled detectors.
• Analysis of LLM-generated misinformation, its characterization, threats across fields, and countermeasures.

Experimental results

Research questions

• RQ1 Can LLMs be effectively utilized to combat misinformation?
• RQ2 What are effective strategies to mitigate LLM-generated misinformation?
• RQ3 How can LLMs enhance misinformation detection, intervention, and attribution in a trustworthy manner?
• RQ4 What is the role of multilingual and multimodal LLMs, LLM agents, and human-LLM collaboration in this domain?
• RQ5 What interdisciplinary measures are needed to address emerging risks of LLM-generated misinformation?

Key findings

• LLMs offer strong world knowledge and reasoning that can bolster misinformation detection and verification.
• LLMs augmented with external knowledge, tools, and multimodal inputs can mitigate hallucinations and improve factuality.
• There are emerging opportunities for LLMs in intervention and attribution, not just detection, along with human-LLM collaboration.
• Trustworthiness concerns remain central, including robustness, explainability, fairness, and privacy, requiring new evaluation approaches for LLM-based detectors.
• LLM-generated misinformation poses real-world threats across journalism, healthcare, finance, and politics, necessitating proactive countermeasures and interdisciplinary coordination.