Causality Guided Representation Learning for Cross-Style Hate Speech Detection

2025-10-08

AI Summary

Existing hate speech detection models over-rely on superficial linguistic features, limiting generalization to implicit expressions (e.g., irony, stereotypes, coded language) and cross-platform stylistic variations, while remaining vulnerable to spurious correlations. To address this, we propose CADET—a novel causal representation learning framework for hate speech detection. CADET constructs a causal graph integrating contextual, motivational, intentional, and stylistic factors, and employs latent variable disentanglement, counterfactual intervention, and deconfounded representation learning to explicitly separate underlying hateful intent from surface-level linguistic style. Extensive experiments demonstrate that CADET significantly outperforms state-of-the-art methods on multi-source, cross-style detection benchmarks. It achieves superior robustness against distributional shifts, enhanced generalization to unseen domains and implicit expressions, and improved interpretability via causal attribution. To our knowledge, CADET is the first framework to formalize hate speech detection as a causal inference problem and instantiate it through principled causal representation learning.

Technology Category

• Machine Learning: Causal Learning
• Natural Language Processing: Sentiment Analysis, Stylistic Analysis, and Argument Mining
• Knowledge Representation and Reasoning: Action, Change, and Causality

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

The proliferation of online hate speech poses a significant threat to the harmony of the web. While explicit hate is easily recognized through overt slurs, implicit hate speech is often conveyed through sarcasm, irony, stereotypes, or coded language -- making it harder to detect. Existing hate speech detection models, which predominantly rely on surface-level linguistic cues, fail to generalize effectively across diverse stylistic variations. Moreover, hate speech spread on different platforms often targets distinct groups and adopts unique styles, potentially inducing spurious correlations between them and labels, further challenging current detection approaches. Motivated by these observations, we hypothesize that the generation of hate speech can be modeled as a causal graph involving key factors: contextual environment, creator motivation, target, and style. Guided by this graph, we propose CADET, a causal representation learning framework that disentangles hate speech into interpretable latent factors and then controls confounders, thereby isolating genuine hate intent from superficial linguistic cues. Furthermore, CADET allows counterfactual reasoning by intervening on style within the latent space, naturally guiding the model to robustly identify hate speech in varying forms. CADET demonstrates superior performance in comprehensive experiments, highlighting the potential of causal priors in advancing generalizable hate speech detection.

Problem

• Detecting implicit hate speech using sarcasm and irony
• Addressing spurious correlations in cross-platform hate detection
• Disentangling genuine hate intent from superficial linguistic cues

Innovation

• Causal graph models key hate speech factors
• Disentangles hate speech into interpretable latent factors
• Uses counterfactual reasoning for robust style variation detection