Preparing the eLuna Visual Language for Human-AI Co-Specification and Code Automation

The eLuna Visual Language, a visual specification framework, is examined for its readiness to support human-AI co-specification and automated code generation. The central finding is that while eLuna performs well as a co-design method for human stakeholders and programmers, its visual language is not yet expressively complete or unambiguous enough to serve as a reliable basis for AI-assisted specification or direct code generation from blueprints. The work is grounded in research on proactive AI programming support, agile development methodology, and the growing use of generative AI in games.

The study is a participatory action research effort centered on the Super Climate Model Game, a mobile narrative game-based learning tool for climate science. A usability inspection involving six experts produced 22 unique issues drawn from 53 reports. Addressing these issues and the broader shortcomings in ambiguity and expressiveness led the researchers to propose seven modifications to the visual language, including distinguishing quest qualifiers, enabling event looping, and supporting conditional CRUD procedures.

Budget and milestone constraints prompted the team to adopt a hybrid eLuna and SCRUM development process, a practical adjustment that the authors present as evidence of eLuna's flexibility. They conclude that eLuna can complement established agile methods rather than replace them. The Super Climate Model Game was delivered feature-complete and is being prepared for educational distribution.

The scope of the work is focused on a single game project and a small expert evaluation, situated within educational game development for climate science. Its contribution lies chiefly in identifying concrete language-level gaps that must be closed before AI-driven code automation from visual specifications becomes practical.