IBI-001303_significativepublie
Majority Language Bias in LLMs
Biais de langue majoritaire (LLMs)
IBI Score
6.7
out of 10
Severity3_significative
Frequencysystematique
Scope groups4
Detection difficulty2
🇫🇷 Définition (FR)
Performances dégradées des LLMs pour les langues minoritaires peu représentées.
🇬🇧 Definition (EN)
Degraded LLM performance for minority languages under-represented.
TriAxis Taxonomy
Structural Classification
O
Origin
O2
Computational & Statistical
L
Lifecycle
L2
Representation
S
Specific
S22
Linguistic bias
💡
Concrete Example
GPT-4 évaluation 2023 : 85% de précision en anglais vs 56% en arabe dialectal.
📊 Detection Metrics
Demographic Parity DifferenceDisparate Impact RatioEqualized Odds Difference
⚙ Alert Thresholds
Disparate Impact Ratio0.8
Equalized Odds Difference0.1
Demographic Parity Difference0.1
📚 Sources & Scientific References
- •The State and Fate of Linguistic Diversity and Inclusion in the NLP WorldPratik Joshi, Sebastin Santy, Amar Budhiraja, Kalika Bali, Monojit Choudhury (2020)
- •Think Outside the Data: Colonial Biases and Systemic Issues in Automated Moderation Pipelines for Low-Resource LanguagesFarhana Shahid, Mona Elswah, Aditya Vashistha (2025)
- •On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?Bender, E.M., Gebru, T., McMillan-Major, A., Shmitchell, S. (2021)
🛡 Impact & Mitigation
🎯 Vulnerable groups impacted
Minority language speakers
🛠 Mitigation methodologies
- AUDIT_TIERS : Third-party Audit
🏢 Sectors concerned
Education & Training (Risk: /5)Media & Social Platforms (Risk: /5)HR & Recruitment (Risk: /5)Justice & Security (Risk: /5)Healthcare & Medicine (Risk: /5)
⚖ Applicable regulations
🌍 International
- •ISO_42001 : ISO/IEC 42001:2023
- •UNESCO_2021 : UNESCO - Recommandation sur l'éthique de l'IA
- •OECD_AI_PRINCIPLES : OCDE - Principes IA
- •COE_AI_CONVENTION_INT : Conseil de l'Europe - Convention IA
📋 Metadata
Contributor
Mohamadou SIDIBE
Institution
AI BIAIS INITIATIVE
Version
1
Created on
October 14, 2026
Updated on
November 4, 2026
UUID
b1287af9-c81e-4e49-b914-00ee992ed01f
Explore other biases
250 documented biases, classified by the TriAxis taxonomy.