Towards a Multidisciplinary Vision for Culturally Inclusive Generative AI (Dagstuhl Seminar 25022)

Abstract

This report documents the program and the outcomes of Dagstuhl Seminar 25022 "Towards a Multidisciplinary Vision for Culturally Inclusive Generative AI". The gathering focused on questions raised by the rapid deployment of Generative AI systems and their integration into global systems of cultural communication, consumption, and production. As these technologies shape our cultures, we urgently need conceptual foundations for investigating the cultural inclusivity of generative AI pipelines (from data collection, to model development and deployment, to evaluation), as well as methods to study the varying societal and cultural impacts of generative AI.
This Dagstuhl Seminar convened scholars and practitioners from computer science, social sciences, the tech industry, and creative industries to discuss the cultural implications of generative AI and find paths toward building generative AI that can be responsive to the diverse needs of individuals, groups, and societies around the world. Together, seminar participants began the challenging but necessary work of building shared language and frameworks for reshaping the technical and social architectures of generative AI.
The seminar was structured along three main dimensions for interdisciplinary discussions:  
- Examining the cultural values being currently centered in generative AI. 
- Studying the possibilities and risks of encoding cultural knowledge into generative AI technologies. 
- Understanding the cultural impact of these technologies. 
We succeeded in building an expert network committed to understanding and designing a culturally-attuned generative AI and to lay the foundation for an interdisciplinary research and practice agenda on global inclusion and generative AI.