A Qualitative Analysis of Creative Agency and Offloading in an AI-assisted Educational Game Design Assignment

The growing adoption of generative AI has raised concerns about its influence on creativity and learning in education. This qualitative case study examines how creative agency manifests in students' cognitive offloading when using generative AI as a design partner in a creative educational game design assignment. Twenty-six students used Microsoft Copilot to design educational game concepts in a university course. We analysed students' reflection essays using thematic analysis to explore how creative agency, conceptualized as the higher-order construct guiding what students delegate to AI and what they retain for themselves, shaped their offloading practices during the design process.We identified a spectrum of patterns ranging from strong to limited expressions of creative agency. Strong creative agency was characterised by students positioning AI as a complementary support tool, engaging in strategic offloading, evaluating AI outputs by recognizing their limitations and appraising its value, and actively maintaining ownership of the design work. Limited creative agency was reflected in tendencies to overestimate AI's capabilities, expect ready-made answers, cede ownership of the creative process, and lack clear strategy for directing the interaction. Notably, these patterns were not mutually exclusive within individual students. The findings suggest that creative agency is not inherently diminished through offloading to AI but rather is shaped by offloading practices, students' AI literacy, and the design of the task. Drawing on these findings, we introduce the concept of creative offloading, describing delegation of elements of a creative process to AI in ways that redistribute creative authorship between human and AI.