From Vibe Coding to Playable Educational Escape Rooms: Extending Room2Educ8 for AI-Assisted Prototyping

Abstract

Educational escape rooms (EERs) can support active, collaborative and problem-based learning, but developing digital and hybrid versions can create technical and interaction-design challenges for educators. Generative AI has been used to support EER development through persona creation, puzzle ideation, hint generation and debriefing, yet its role in implementing established designs as playable artefacts has received less attention. This study investigates AI-assisted prototyping within Room2Educ8, a Design Thinking framework for learner-centred EER development. An exploratory qualitative multiple-case study involved 23 postgraduate User Experience Design students working in four groups. All groups designed an EER as assessed coursework and adopted vibe coding during prototyping, although its use was neither required nor demonstrated in class. They used natural-language prompts with AI-assisted coding tools to create and revise software. Data sources included assessment portfolios, playable artefacts, project demonstrations, group viva responses, retained prompts, screenshots, individual reflections and testing records. Across the cases, AI-assisted coding was introduced after target audiences, learning objectives, narratives and puzzle structures had been established. It was used to implement interfaces, answer validation, feedback, hints, timers, navigation and transitions between digital and physical tasks. Playtesting identified problems involving ambiguous instructions, alternative valid answers, feedback, hint scaffolding, progression and accessibility. Broader challenges concerned behavioural specification, control over generated outputs, pedagogical alignment, documentation and consistency across generated media. The findings show how AI-assisted coding supported the implementation and revision of interactive EERs, although development time, effort and learning outcomes were not measured. The paper proposes Room2Educ8-AIP, a preliminary extension comprising seven iterative actions within the Prototype stage: specify, generate, inspect, revise, playtest, align and document. The extension offers educators and learning designers a structured way to integrate AI-assisted coding into EER development while maintaining design control, pedagogical alignment and clear documentation of implementation decisions.