Critical Agency and Hybrid Cognition in digital art education: Human–AI co-creation with stable diffusion XL

The integration of text-to-image Generative AI (GenAI) in higher education challenges conventional notions of authorship and design pedagogy. This study examines human–AI co-creation with Stable Diffusion XL (SDXL), an open-source text-to-image model, in undergraduate digital art classrooms in Kuwait. While SDXL offers reproducibility and precise parameter control, it often reproduces Western-centric perspectives when generating images of non-Western cultures, a phenomenon termed “Algorithmic Orientalism.” We introduce the framework of Critical Hybrid Cognition, emphasizing the joint creative agency between students and AI. Through a practice-based methodology, we analyze student interactions with GenAI tools and document the “friction”—the disjunctions between students' creative intentions and AI outputs. Three recurring error types emerged when students used SDXL to represent Gulf cultural heritage: sociocultural dissonance (e.g., misrepresentation of gendered attire), temporal displacement (anachronistic or historically inaccurate details), and morphological hallucinations (structural or biological inaccuracies). The transparency of SDXL's open-source architecture fostered student engagement; students enacted “Algorithmic Resistance” by employing exclusionary syntax (negative prompting) and refining generative constraints to counter stereotyping. We argue that GenAI education should move beyond prompt construction to develop student skills in “curatorial verification”—the critical assessment and refinement of AI outputs. Ultimately, we propose that Hybrid Cognition in GenAI contexts relies on students acting as domain experts, leveraging technical agency to negotiate with, rather than passively accept, machine-generated results.