The field of artificial intelligence is moving towards a greater emphasis on cultural awareness and bias mitigation. Recent research has highlighted the importance of developing AI systems that can understand and respect diverse cultural contexts, values, and norms. This includes the need for more equitable and robust performance in safety guardrails, the detection and mitigation of perspectival biases, and the development of culturally aware machine translation. Furthermore, there is a growing recognition of the need for mixed-initiative collaboration between humans and AI systems to reflect cultural knowledge and values. Overall, the field is shifting towards a more nuanced understanding of the complex relationships between culture, language, and AI, with a focus on developing more inclusive and respectful AI systems. Noteworthy papers in this area include: Evaluating Perspectival Biases in Cross-Modal Retrieval, which highlights the need for targeted strategies to mitigate bias in multimodal systems. Characterizing Selective Refusal Bias in Large Language Models, which emphasizes the importance of equitable and robust performance in safety guardrails. Why Do Multilingual Reasoning Gaps Emerge in Reasoning Language Models?, which proposes a simple yet effective strategy to bridge the multilingual reasoning gap.
Cultural Awareness and Bias Mitigation in AI
Sources
Evaluating Perspectival Biases in Cross-Modal Retrieval
Characterizing Selective Refusal Bias in Large Language Models
Why Do Multilingual Reasoning Gaps Emerge in Reasoning Language Models?
MemeArena: Automating Context-Aware Unbiased Evaluation of Harmfulness Understanding for Multimodal Large Language Models
Back to the Communities: A Mixed-Methods and Community-Driven Evaluation of Cultural Sensitivity in Text-to-Image Models
Culture Cartography: Mapping the Landscape of Cultural Knowledge
Diverse Human Value Alignment for Large Language Models via Ethical Reasoning
Do You Know About My Nation? Investigating Multilingual Language Models' Cultural Literacy Through Factual Knowledge
The Riddle of Reflection: Evaluating Reasoning and Self-Awareness in Multilingual LLMs using Indian Riddles
Surfacing Subtle Stereotypes: A Multilingual, Debate-Oriented Evaluation of Modern LLMs
Evaluating Cultural Knowledge Processing in Large Language Models: A Cognitive Benchmarking Framework Integrating Retrieval-Augmented Generation
LiveSecBench: A Dynamic and Culturally-Relevant AI Safety Benchmark for LLMs in Chinese Context
PragExTra: A Multilingual Corpus of Pragmatic Explicitation in Translation
RLHF: A comprehensive Survey for Cultural, Multimodal and Low Latency Alignment Methods
LLMs and Cultural Values: the Impact of Prompt Language and Explicit Cultural Framing
Advancing Equitable AI: Evaluating Cultural Expressiveness in LLMs for Latin American Contexts