Design as Language Co-Construction: Toward Responsible, Transparent, and Adaptive AI Support for Educational Game Design

Abstract

Despite proven benefits, educational games are rarely adopted in higher education due to the complexity of designing games that fit diverse teaching needs. Most instructors lack game design expertise or access to professional support. Advances in AI, particularly large language models (LLMs), now make it possible for instructors to design games with AI assistance. However, effective AI support must go beyond generating content; it should build trust, encourage instructor agency, and ensure the final game aligns with teaching goals. My dissertation explores an interactive framework where instructors and AI co-construct a shared design language, enabling collaborative, transparent, and context-aware game design. The in-development tool uses LLM-driven dialogue to transform course materials and pedagogical aims into game design artifacts such as mechanics outlines and design documents. Its text-based interface, supported by lightweight visual scaffolds, allows iterative refinement and explanation of AI suggestions, empowering instructors to create educational games suited to their classrooms.