Background
The rapid dissemination of news is an important factor for the global population to stay up to date with current events, or to make well-informed decisions in dynamic sectors like the stock market. Journalists typically include large amounts of quantitative information using data visualization to help enhance accessibility and comprehension of complex datasets. However, the propagation of misinformation and false news has also grown in recent years, leading to the undermining of public trust in news reporting. Current fact-checking efforts primarily focus on the factual integrity of the text in journal articles, which tends to neglect the impact of argumentation practices and the accuracy of data visualizations.
Invention Description
Researchers at Arizona State University have developed new methods and systems for detecting disinformation generated by large language models (LLMs). This technology includes the development of novel datasets and advanced prompting techniques to enhance the accuracy of disinformation detection, addressing the limitations of existing techniques. This system leverages a combination of fine-tuned machine learning models and chain-of-thought-inspired prompts to systematically identify and fact-check content elements, improving detection robustness and reducing misclassification rates.
Potential Applications:
- Enhancing disinformation detection for news & media outlets
- Improving accuracy of news information on social platforms
- Cybersecurity & public policy
Benefits and Advantages:
- Robust – employs state-of-the-art disinformation detection training and testing models
- Versatile – able to detect both simple and complex disinformation scenarios, as well as nuanced and persuasive information generated by LLMs
- Reliable – incorporates detailed analytical process that assesses the factualness and relationships between key content elements
- Easily integrated – can enhance the ability of existing platforms and services to detect and mitigate disinformation
- Scalable – can be applied across various domains
Related Publication: Disinformation Detection: An Evolving Challenge in the Age of LLMs