Responsible AI in Education: Multilingual, Multimodal, and Relational Futures

Artificial intelligence (AI) is rapidly reshaping education, yet most AI in education systems are grounded in Western, monolingual, and individualistic models of learning. These systems can perpetuate inequities by overlooking the multilingual, multimodal, and culturally situated ways in which students engage in learning. This paper argues that responsibility must be reimagined beyond technical and ethical compliance toward cultural, relational, and pedagogical commitments. Drawing from insights across sociocultural theories of learning, critical and decolonial education research, translanguaging theory, and participatory design research, it proposes four interrelated orientations for rethinking responsible AI in education (RAIED): culturally situated learning as a foundation for AI design, collaborative and participatory design with stakeholders, multilingual and multimodal lenses for recognizing diverse ways of knowing, and equity-centered technical design of AI systems. Together, these orientations contribute generative starting points for building RAIED systems designed in collaboration with communities and grounded in the diversity of human learning, ways of knowing, and contexts.