In-context learning for propaganda detection on Twitter Mexico using large language model meta AI
This study explores the application of Large Language Models (LLMs) for detecting political propaganda on Twitter, focusing on manipulative political narratives during the 2018 Mexican presidential election. Using LLaMA 3.2, we implement a few-shot prompting within an In-Context Learning framework t...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-09-01
|
| Series: | Telematics and Informatics Reports |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2772503025000465 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This study explores the application of Large Language Models (LLMs) for detecting political propaganda on Twitter, focusing on manipulative political narratives during the 2018 Mexican presidential election. Using LLaMA 3.2, we implement a few-shot prompting within an In-Context Learning framework to classify tweets as propagandist or non-propagandist. Using a dataset of over 800,000 tweets mentioning the leading candidate, our model identifies linguistic patterns, sentiment dynamics, and adversarial tactics, including emotive language, personal attacks, and the strategic use of hashtags. Results indicate that 58.4 % of the analyzed tweets exhibited propagandist characteristics, with a predominance of negative sentiment and aggressive tone, indicating a significant presence of information manipulation and polarization. Hierarchical clustering and word frequency analyses reveal coordinated messaging patterns, reinforcing the role of social media as a tool for political discourse manipulation. Our findings demonstrate the performance of LLMs in automating misleading communication detection and provide a replicable AI-driven framework for analyzing manipulative political narratives in digital environments. |
|---|---|
| ISSN: | 2772-5030 |