Beyond Logical Forms: LLM-Extracted Patterns for Fallacy Classification

2026-06-25

AI Summary

Existing approaches struggle to simultaneously capture the abstract reasoning structures of logical fallacies and contextual linguistic cues, limiting their effectiveness in identifying complex and subtle fallacies. This work proposes a data-driven pattern extraction mechanism that leverages large language models to induce discriminative representations from fallacy instances and their explanations, integrating both logical rigor and contextual awareness—thereby overcoming the constraints of purely formal logical representations. Employing zero-shot and one-shot prompting strategies, the method consistently outperforms baseline approaches across multiple large language models and experimental settings. Furthermore, it demonstrates strong generalization capabilities in cross-dataset evaluations, significantly advancing the state of the art in fallacy classification performance.

Abstract

In today's fast-paced information era, logical fallacies, defined as defective patterns of reasoning, inevitably contribute to the growth of information disorder. However, often fallacies appear in nuanced forms that complicate automated classification. In this study, we investigate whether merging abstract logical structures with context-level linguistic cues proves beneficial for fallacy classification, developing a framework that inductively extracts such patterns from fallacious examples and their explanations using Large Language Models (LLMs). We evaluate the impact of these patterns across different LLMs and experimental zero- and one-shot configurations, showing statistically significant improvements over zero-shot baselines and outperforming competing approaches. Cross-dataset experiments validate generalization, establishing data-driven pattern extraction as an effective method for generating logical representations.

Problem

• logical fallacies
• fallacy classification
• information disorder
• automated classification
• nuanced reasoning patterns

Innovation

• logical fallacy classification
• pattern extraction
• large language models
• zero-shot learning
• inductive reasoning