LLM-based Detection of Manipulative Political Narratives

2026-05-14 · Source: Computation and Language · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Social Sciences & Behavioral Studies · Depth: Advanced, quick

Summary

Researchers developed a new computational framework to detect and structure manipulative political narratives on social media, addressing the challenge of distinguishing them from legitimate critiques. The framework processes over 1.2 million social media posts, initially filtering them using a detailed few-shot prompt that combines documented campaign narratives with legitimate criticisms. This prompt enables a reasoning model to assign labels, retaining only manipulative posts. The filtered posts are then embedded and dimensionality-reduced using UMAP, followed by HDBSCAN to uncover narrative groups. This unsupervised approach identified 41 distinct manipulative narrative clusters, demonstrating its ability to discover new narrative patterns without predefined categories. A final reasoning model then interprets the narrative behind each cluster.

Key takeaway

For research scientists developing tools for social media analysis, this framework offers a robust method for identifying and categorizing manipulative political narratives. You should consider integrating prompt-based filtering with unsupervised clustering techniques like UMAP and HDBSCAN to uncover novel narrative structures in large datasets, enhancing your ability to detect emerging disinformation campaigns without relying on predefined categories.

Key insights

A framework combines prompt-based filtering with unsupervised clustering to detect and structure manipulative political narratives.

Principles

• Differentiate manipulation from legitimate critique.
• Unsupervised clustering uncovers new narrative patterns.

Method

Filter posts with a few-shot prompt, embed and reduce dimensionality with UMAP, cluster with HDBSCAN, then use a reasoning model to interpret clusters.

In practice

• Use few-shot prompting for nuanced text classification.
• Combine UMAP and HDBSCAN for unsupervised topic discovery.

Topics

• LLM-based Detection
• Manipulative Narratives
• Few-shot Prompting
• Unsupervised Clustering
• Social Media Analysis

Best for: Research Scientist, NLP Engineer, Machine Learning Engineer, AI Scientist

Related on AIssential

• LLM-based Detection of Manipulative Political Narratives — LLM-driven pipelines can effectively identify and cluster manipulative political narratives in raw social media data.
• Addressing Data Scarcity in Bangla Fake News Detection: An LLM-Based Dataset Augmentation Approach — LLM-based data augmentation significantly improves fake news detection in low-resource languages by generating diverse synthetic data.
• ttda704 at SemEval-2026 Task 6: Structured Chain-of-Thought Prompting for Political Evasion Detection — Structured Chain-of-Thought prompting significantly improves political evasion detection over fine-tuning large language models.
• Exposing biases, moods, personalities, and abstract concepts hidden in large language models — A new method can identify and manipulate abstract concepts hidden within large language models to enhance or minimize their expression.
• Your UnEmbedding Matrix is Secretly a Feature Lens for Text Embeddings — EmbedFilter refines LLM text embeddings by filtering out high-frequency token influence from the unembedding matrix subspace, enhancing semantics and reducing dimensions.
• Graph-Augmented LLMs for Swiss MP Ideology Prediction — Graph-augmented LLMs can model complex political ideologies from text and relational data.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.