Abstract:

Generative artificial intelligence (AI) is making it easier to produce educational content at scale, including narrative and interactive learning experiences. For game-based learning, however, technical generation alone is not sufficient. If automatically generated scenarios are confusing, narratively inconsistent, or simply unengaging, they are unlikely to be useful in practice. Interactive fiction offers a manageable test case because it combines educational content with branching narrative and learner choice.

This paper presents a pilot evaluation of AI-generated interactive fiction for educational use in higher education. It builds on previous work that generated Ink-based scenarios from structured educational content in a domain-agnostic way. For the present evaluation, the same pipeline is used to generate a comparable set of scenarios from a shared educational content base, allowing narrative quality to be examined across outputs grounded in the same subject matter.

Participants engage with selected generated episodes in a brief online setting and then complete a short questionnaire. The evaluation focuses on three dimensions: whether the scenario and its educational content are understandable, whether story and learning content fit together coherently, and whether the experience is engaging. Open-ended responses are used to capture recurring strengths, points of confusion, and weaknesses that structured ratings may miss.

The contribution is a first user-centred step beyond automated generation metrics. In addition to asking whether generated scenarios compile and run, the paper examines how they are experienced by players. The result is a cautious pilot assessment of perceived quality that can inform larger follow-up studies, including later work on learning effectiveness.

Keywords:

Artificial intelligence in education, interactive fiction, game-based learning, serious games, technology-enhanced learning, higher education.