CoCoA: Context-Conditional Cultural Alignment for Large Language Models
Computation and Language
Summary
The authors found that large language models often prefer Western examples regardless of cultural context. They created CoCoA, a method that helps models choose culturally relevant examples when given cultural clues, but stay neutral otherwise. By training with and without cultural hints on the same data, CoCoA better aligns models to culture-specific expectations. Testing showed CoCoA reduces cultural bias while keeping overall performance stable. This work suggests that handling cultural bias needs context-aware approaches, not one-size-fits-all fixes.
Authors
Kyungdon Lee, Wei Xu, Alan Ritter, Dong-Ho Lee, JinYeong Bak
Abstract
Large Language Models (LLMs) often favor Western-associated entities across cultural contexts. Conventional debiasing methods aim for uniform neutrality, but cultural bias mitigation demands context-conditional behavior, preferring culturally appropriate entities when cultural cues are present and remaining neutral when they are absent. We propose CoCoA (Context-Conditional Cultural Alignment), a framework that learns this behavior through dual-context training on the same entity pairs under contexts with and without cultural cues. CoCoA combines a contrastive alignment objective with calibration and drift regularization, optimized through goal-aware gradient reconciliation. We evaluate CoCoA on CAMeL and Camellia, two entity-centric cultural bias benchmarks, across ten language settings and four LLMs. CoCoA reduces the Cultural Bias Score from 43 to 24 on average while maintaining near-neutral preferences at 50.2, with minimal impact on general performance across five standard benchmarks. These findings highlight that effective cultural alignment requires context-conditional modeling rather than uniform debiasing, and establish a new direction for mitigating entity-centric cultural bias in LLMs.