Bias and Justice in Language Models: Evidence from a Linguistically, Socially, and Culturally Asymmetric Literature

2026-04-12 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Advanced, quick

Summary

A systematic literature review of 60 studies published between 2020 and 2025 reveals significant biases and gaps in research concerning Large Language Models (LLMs). The review highlights a strong predominance of evaluations conducted in English, a disproportionate focus on gender bias often treated as binary, and a greater emphasis on diagnosing bias rather than implementing mitigation strategies. Key findings also indicate a scarcity of intersectional, multilingual, and real-world use case analyses, pointing to substantial methodological and sociocultural deficiencies within the current academic discourse on LLM fairness and justice. This asymmetry in research priorities raises concerns about the generalizability and applicability of current bias mitigation efforts.

Key takeaway

For research scientists developing or evaluating LLMs, recognizing the current literature's biases is crucial. You should prioritize designing studies that incorporate multilingual datasets, intersectional analyses of discrimination, and real-world usage scenarios to ensure more robust and equitable model development. Shifting focus from mere diagnosis to effective mitigation strategies will lead to more just and globally applicable LLMs.

Key insights

LLM bias research is skewed towards English and binary gender, lacking intersectional and multilingual real-world analyses.

Principles

• Research disproportionately diagnoses bias over mitigating it.
• Bias evaluations are predominantly English-centric.
• Gender bias is often treated as a binary issue.

In practice

• Prioritize multilingual LLM bias evaluation.
• Integrate intersectional bias analysis.
• Focus on real-world LLM use scenarios.

Topics

• Large Language Models
• Algorithmic Justice
• Social Bias
• Systematic Literature Review
• Gender Bias

Best for: Research Scientist, AI Scientist, AI Ethicist, NLP Engineer

Related on AIssential

• A Systematic Analysis of Biases in Large Language Models — LLMs, despite alignment efforts, inherit and perpetuate diverse biases from human-generated training data.
• Taken into Consideration: An Evaluation of Estimation Biases by Race, Gender, and Region in Large Language Models in Brazilian Portuguese — Large language models exhibit systematic esteem biases related to gender, race, and region in Brazilian Portuguese.
• Why are all LLMs Obsessed with Japanese Culture? On the Hidden Cultural and Regional Biases of LLMs — LLMs exhibit a surprising bias towards Japanese culture, emerging post-supervised fine-tuning.
• Artificial Intolerance: Stigmatizing Language in Clinical Documentation Skews Large Language Model Decision-Making — LLMs inherit and propagate human cognitive biases from stigmatizing language in clinical notes, leading to less aggressive patient management.
• Investigating Counterfactual Unfairness in LLMs towards Identities through Humor — LLMs exhibit counterfactual unfairness, applying stricter safety policies and attributing more malicious intent based on speaker-target identity.
• Extending the Contact Hypothesis: Cross-Linguistic Evaluation of Religion and Nationality Bias When Prompting LLMs in German and Icelandic — LLMs exhibit religion and nationality biases, requiring cross-linguistic evaluation for fairness.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.