Generative AI Meets Adventure: Elevating Text-Based Games for Engaging Language Learning Experiences

Abstract

This work explores generative AI-enhanced text-based games for language learning. Utilizing ChatGPT, the study modifies an early version of Colossal Cave Adventure to generate contextually appropriate responses, making the game more interactive and engaging. The method involved creating a tailored model to emulate the original game, followed by iterative refinements. Playthrough-based comparisons revealed that the model successfully reproduced the game’s core mechanics while introducing dynamic content and contextual hints, enhancing the overall gaming experience. At the same time, the model’s nondeterministic nature showed benefits in the context of language generation while introducing challenges, such as inconsistencies in the progression of language complexity. Altogether, generative AI holds promise for improving educational, text-based games by increasing engagement and adaptability; thus, educators are encouraged to consider AI-enhanced gaming as an innovative tool for language instruction and interactive learning. However, further research is also needed to address AI limitations to ensure learning consistency, alongside the value of exploring broader applications beyond language acquisition.