Beyond the Explicit: A Bilingual Dataset for Dehumanization Detection in Social Media

2025-10-21

AI Summary

Prior research on digital dehumanization primarily focuses on detecting explicit negative rhetoric, overlooking more subtle and pervasive implicit forms—those lacking overt hostility yet perpetuating bias against marginalized groups, posing greater detection challenges. Method: We introduce the first theory-driven, bilingual (English–French) dataset for implicit dehumanization (16,000 annotated instances), featuring systematic, multi-dimensional annotations at both document- and span-levels (e.g., animalization, mechanization, individuation loss). Annotation combines expert guidance with crowd-sourcing, multi-source sampling, and ML-assisted refinement. Evaluation employs zero-shot and few-shot settings. Contribution/Results: Our approach significantly outperforms state-of-the-art methods under low-resource conditions. This work establishes the first benchmark for implicit dehumanization detection, enabling reproducible cross-lingual evaluation and advancing bias-aware NLP in resource-constrained settings.

Technology Category

• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy
• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Ethics, Bias, and Fairness

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

Digital dehumanization, although a critical issue, remains largely overlooked within the field of computational linguistics and Natural Language Processing. The prevailing approach in current research concentrating primarily on a single aspect of dehumanization that identifies overtly negative statements as its core marker. This focus, while crucial for understanding harmful online communications, inadequately addresses the broader spectrum of dehumanization. Specifically, it overlooks the subtler forms of dehumanization that, despite not being overtly offensive, still perpetuate harmful biases against marginalized groups in online interactions. These subtler forms can insidiously reinforce negative stereotypes and biases without explicit offensiveness, making them harder to detect yet equally damaging. Recognizing this gap, we use different sampling methods to collect a theory-informed bilingual dataset from Twitter and Reddit. Using crowdworkers and experts to annotate 16,000 instances on a document- and span-level, we show that our dataset covers the different dimensions of dehumanization. This dataset serves as both a training resource for machine learning models and a benchmark for evaluating future dehumanization detection techniques. To demonstrate its effectiveness, we fine-tune ML models on this dataset, achieving performance that surpasses state-of-the-art models in zero and few-shot in-context settings.

Problem

• Detecting subtle dehumanization forms in social media content
• Addressing overlooked bilingual dehumanization in computational linguistics
• Identifying harmful biases against marginalized groups online

Innovation

• Bilingual dataset for dehumanization detection
• Document- and span-level annotation of instances
• Fine-tuned ML models outperform state-of-the-art