Beyond Bias: Evaluating Cultural AI Through Participation and Interpretation

Generative AI systems are increasingly understood as cultural technologies that produce and circulate meaning, yet their evaluation remains largely focused on harm mitigation through bias, fairness, and safety frameworks (Bender et al., 2021; Buolamwini & Gebru, 2018). This paper presents Beyond Bias, a collaboration between Gooey.AI and Goethe-Institut India, as a participatory framework for evaluating cultural AI combining collaborative dataset creation, artist-led model fine-tuning, accessible tooling, and co-authored governance practices. Drawing from participatory design and interpretive approaches, we propose three ideas for evaluating cultural AI systems: cultural integrity, participation and agency, and interpretive capacity, while also examining tensions around authorship, ownership, and cultural extraction in generative systems.