CULTURAL CHALLENGES IN AI INTERPRETING
DOI:
https://doi.org/10.32782/2410-0927-2025-23-17Keywords:
AI Interpreting, S2ST (Speech-to-Speech Translation), cultural dimensions, Hofstede’s Model, pragmatic correction, pragmatic equivalence, power distance index (PDIScore), social acceptability, prosodic adaptation, adaptive cascadeAbstract
The article addresses the limitations of current AI Speech-To-Speech Interpreting (S2S) systems in capturing cultural and pragmatic nuances It highlights that AI interpreters often fail to recognize subtle emotional cues, social dynamics, and cultural context, leading to miscommunication, alienation, or a perception of insincerity. The methodology involves analyzing existing architectures and proposing an advanced Adaptive Cascade architecture that integrates modules like Pragmatic Correction (PCM) and Tone & Style Control (TCS). These modules are trained via reinforcement learning with human feedback to enable dynamic sociolinguistic adaptation, considering cultural dimensions such as Hofstede’s model and social parameters like Power Distance Index (PDI) and Individualism/Collectivism (IDV). The scientific novelty lies in the systematic incorporation of sociolinguistic metrics – such as pragmatic adequacy and social acceptability – beyond traditional lexical accuracy. The approach emphasizes modeling social parameters and developing datasets annotated with cultural and emotional metadata to improve the interpretive quality. Overall, achieving truly effective cross-cultural AI interpreting requires moving beyond static models and lexical metrics toward dynamic, context-aware, sociolinguistically informed systems. The paper also discusses the potential for these systems to enhance diplomatic communication and international collaboration by reducing misunderstandings. It advocates for interdisciplinary research combining linguistics, AI, and cultural studies to create more nuanced and ethically responsible interpreting tools. The authors suggest that future work should include real-world testing in diverse cultural settings to validate these models’ effectiveness. Such advancements could significantly improve global communication, fostering greater mutual understanding and respect among different cultures. Integrating cultural dimensions into AI Interpreting can help prevent cultural insensitivity and promote respectful intercultural exchanges.
References
Al-Rubaie A., Ahmad S. The Impact of Hofstede’s Cultural Dimensions on UX/UI Design for Global AI Applications: A Review of Emerging Trends. International Journal of Human-Computer Studies. 2024. Vol. 181. Article 103045.
Austin J. L. How to Do Things with Words. 2nd ed. Harvard University Press, 1975. 168 p.
Chen M., Chen Z., Chen J. Scaling Speech-to-Speech Translation to Large Language Models: A Unified Framework for Multimodal Pragmatic Adaptation. Transactions on Audio, Speech, and Language Processing. 2024. Vol. 33. P. 150–165.
Cui L., Hu J. Reinforcement Learning for Dialogue Management: A Reward Model Based on User Satisfaction and Trust. IEEE Transactions on Audio, Speech, and Language Processing. 2024. Vol. 32. P. 1200–1215.
Gao S., Li Y., Liu Z. Addressing Social Bias and Stereotypes in Neural Machine Translation via Reinforcement Learning with Human Feedback (RLHF) for Politeness. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (ACL). Toronto: ACL, 2023. P. 7821–7835.
Graham Y. A Cursory Analysis of the Curation of the BLEU Metric. Proceedings of the 10th Workshop on Statistical Machine Translation (WMT). Lisbon: ACL, 2015. P. 280–288.
Hall E. T. Beyond Culture. Anchor Books, 1976. 298 p.
Hofstede G. Culture’s Consequences: Comparing Values, Behaviors, Institutions and Organizations Across Nations. 2nd ed. SAGE Publications, 2001. 616 p.
Kripalani A., Zhang W. Mitigating Sociocultural Bias in Large Language Models through Cross-Cultural Prompting. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (ACL). Toronto: ACL, 2023. P. 4421–4435.
Lee H. G., Kim D. Y. A model for the cultural adaptation of interface design: The case of the Korean Web. Interacting with Computers. 2000. Vol. 12, no. 5. P. 459–475.
Lee K., Nagao M. Modeling Politeness and Social Status in Neural Machine Translation: A Korean-Japanese Case Study. Computational Linguistics. 2018. Vol. 44, no. 4. P. 857–883.
Ponti E. M., Liska R., Waseem Z. On the Inadequacy of Current Evaluation Metrics for Cross-Lingual Pragmatic Tasks. Computational Linguistics. 2023. Vol. 49, no. 2. P. 301–320.
Rubin P., Cappe S. Prosody and Perception: Cultural Variation in the Expression and Interpretation of Emotion in Speech Synthesis. Speech Communication. 2020. Vol. 125. P. 1–11.
Seleskovitch D. Interpréter pour traduire. Didier Érudition, 1984. 308 p.
Searle J. R. Expression and Meaning: Studies in the Theory of Speech Acts. Cambridge University Press, 1979. 187 p.
Watanabe H., Shindo H. Cross-Cultural Communication in S2ST: Bridging the Gap between Prosody and Illocutionary Act. Interspeech 2022, Proceedings of the 23rd Annual Conference of the International Speech Communication Association. Interspeech, 2022. P. 1980–1984.
