AI-deation: When the Teacher is a Transformer in Role-Playing to Create Privacy Decision Serious Games

Generative artificial intelligence, specifically GPT-4, serves as a viable and effective substitute for human stakeholders in the collaborative design of serious games focused on privacy decision-making. By acting as a structured teacher within a co-design framework, the model successfully assists human teams by evaluating game concepts, identifying conflicting design affordances, and generating balanced, actionable feedback. This integration demonstrates that transformer models can maintain high levels of participant satisfaction and utility, performing at a level largely comparable to human collaborators in ideation and conceptual balancing.

The research, conducted within an academic setting focused on game design pedagogy, highlights that while the AI excels at maintaining logical consistency and identifying design tensions, it faces limitations in creative novelty and temporal awareness. The model tends to favor existing design propositions over the generation of entirely new concepts, and its reliance on scheduled prompts often results in interventions that lack alignment with the fluid, evolving discourse of a human design team. These findings suggest that the efficacy of human-AI co-design is heavily dependent on the synchronization of AI feedback with the immediate needs of the creative process.

Ultimately, the study confirms that generative AI is a functional tool for supporting complex design workflows, provided it is constrained by clear, iterative instructions and structured prompts. Future advancements in this field must prioritize the development of context-aware AI that can sense ongoing team dynamics to provide more strategic, flow-conscious interventions. By refining the timing and creative capacity of these models, developers can better leverage AI as a sophisticated partner in the iterative design of educational and serious games.