Is Lying Only Sinful in Islam? Exploring Religious Bias in Multilingual Large Language Models Across Major Religions

2025-12-03

AI Summary

This study systematically investigates religious bias in multilingual large language models (MLLMs) across Buddhist, Christian, Hindu, and Islamic contexts, revealing systematic, cross-lingual inconsistencies—particularly pronounced negative stereotyping of Islam. To address this, the authors introduce BRAND, the first bilingual, auditable dataset for South Asia’s four major religions, comprising 2,400+ English–Bengali samples. They design diverse prompt templates for controlled comparative experiments. Quantitative and qualitative analyses demonstrate that state-of-the-art MLLMs consistently underperform in Bengali relative to English and perpetuate anti-Islamic bias even in ostensibly religion-neutral queries. This work is the first to uncover implicit structural imbalances in religious representation within multilingual LLMs, establishing a novel benchmark and methodological framework for assessing religious fairness in AI systems.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy
• Philosophy and Ethics of AI: Bias, Fairness & Equity

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

While recent developments in large language models have improved bias detection and classification, sensitive subjects like religion still present challenges because even minor errors can result in severe misunderstandings. In particular, multilingual models often misrepresent religions and have difficulties being accurate in religious contexts. To address this, we introduce BRAND: Bilingual Religious Accountable Norm Dataset, which focuses on the four main religions of South Asia: Buddhism, Christianity, Hinduism, and Islam, containing over 2,400 entries, and we used three different types of prompts in both English and Bengali. Our results indicate that models perform better in English than in Bengali and consistently display bias toward Islam, even when answering religion-neutral questions. These findings highlight persistent bias in multilingual models when similar questions are asked in different languages. We further connect our findings to the broader issues in HCI regarding religion and spirituality.

Problem

• Detect religious bias in multilingual large language models
• Address misrepresentation of major South Asian religions
• Compare model performance across English and Bengali prompts

Innovation

• Created BRAND dataset for religious bias detection
• Used bilingual prompts in English and Bengali
• Focused on four major South Asian religions