Although generative artificial intelligence is a global phenomenon, it is still too often discussed in universal terms, without sufficient consideration for cultural, linguistic, and social diversity. The authors of “Global AI Cultures” – including Prof. Aleksandra Przegalińska from Kozminski University – emphasize that the future of this technology should be shaped by cultural pluralism, broadly understood as the collection of ideas, customs, and social behaviors that define different communities and groups.
The publication points out that every interaction with AI takes place within a specific cultural context, and that concepts such as creativity, trust, or the role of work are interpreted differently across various parts of the world. Ignoring this perspective in the model design process can lead to bias, the marginalization of alternative ways of engaging with technology, and a limitation of its innovative potential. The article draws attention to the dominance of English in large language models, which limits their effectiveness in other languages and cultures, as well as to the prevalence of visual and narrative patterns derived primarily from Western culture, which universalize narrow conceptions of artificial intelligence.
The authors note that regulatory conditions – such as copyright and privacy laws – shape both access to data and the performance of the models themselves. In many cases, these regulations are created in the U.S., Europe, or China, while regions in the Global South remain merely their recipients, with limited ability to influence the global ruleset. The data used to train AI models is always embedded in specific social and cultural realities, which is why it is essential to expand training datasets to include languages, experiences, and perspectives that have so far been marginalized, while also respecting the rights of communities that do not wish their data to be used.
In conclusion, the authors stress that an inclusive approach to generative AI not only strengthens the fairness and ethical dimensions of the technology but also enhances its functionality and resilience. Technologies designed with cultural diversity in mind are better prepared for a wide range of applications and for operation in different contexts, enabling them to respond to the needs of a truly global community more effectively.
The full article “Global AI Cultures” is available on the Communications of the ACM website: https://cacm.acm.org/opinion/global-ai-cultures/
