Enhancing English language learning through moral dilemmas: A comparative study of GPT and human-written stories

This study investigates the effects of generative pre-trained transformer (GPT) generated vs. human-written moral dilemma stories on English as a foreign language (EFL) learners’ speaking skills, storytelling ability, and behavioral regulation, framed within the theoretical context of embodied cognition. A total of 120 university students were divided into two groups: the experimental group, which engaged with GPT-generated moral dilemma stories, and the control group, which worked with human-authored moral dilemma stories. Over an 11-week experimental period (including pre-test, 10-week intervention, and post-test), each group was exposed to weekly dilemma stories, followed by speaking assessments using IELTS Part 2 tasks. Analysis of covariance results demonstrated that the GPT group significantly outperformed the control group in fluency, lexical use, and grammar. Furthermore, participants in the GPT group exhibited higher levels of intrinsic motivation and reduced amotivation compared to their counterparts. The GPT-generated stories also enhanced storytelling ability, particularly in the dimensions of content, organization, and innovation. These findings suggest that AI-generated content, specifically using large language models like GPT, can be an effective tool for improving both language proficiency and behavioral regulation in EFL learners, providing valuable insights into the integration of artificial intelligence (AI) technologies in language education.