“Opening the Black Box: Configuring the Algorithm of a GPT-Based Professional Writing Coach”

This article investigates the pedagogical configuration of the Professional Writing Coach (PWC), a GPT-based chatbot developed to support university students in learning professional writing within an English-Medium Instruction (EMI) context. Unlike many studies that broadly assess the impact of generative AI, this paper focuses on its structured use within the classroom as an integrated component of instructional design. It opens the “black box” of chatbot creation, detailing the epistemic foundations, modular architecture, interaction rules, and embedded evaluation rubrics that inform the PWC’s development. Drawing on Berlyne’s theory of epistemic curiosity and Floridi’s concept of semantic capital, the chatbot is framed as a cognitive and dialogic tutor that fosters rhetorical awareness, learner agency, and iterative self-revision. Crucially, the PWC is not designed to simply generate professional texts, but to teach students how to write them — through guided prompting, reflective questioning, and dialogic engagement. The article positions this approach as a rethinking of authorship, responsibility, and the instructor’s role in AI-augmented writing education.