BETWEEN CODE AND THE CLASSROOM: DIGITAL PEDAGOGICAL MEDIATION, TECHNICAL REVIEW, AND COMPUTING TEACHER TRAINING

Abstract

This study investigates the coherence between the pedagogical commitments declared by a digital mediation tool for academic writing, the Orientador Digital de TCC, and its actual technical implementation. It is an applied, qualitative study grounded in the concepts of academic writing as a situated social practice, the zone of proximal development, and the technological pedagogical content knowledge framework, which guide the analysis of a client-side system structured around retrieval-augmented generation. The methodology combined source-code inspection, an initial round of automated tests that reimplemented the system's logic in isolation, and, subsequently, real execution of the code in a simulated browser environment, allowing a comparison of the fault-detection power of each approach. Results revealed five silent failures related to data storage and processing, including content overwriting between two distinct document structures, along with two access barriers linked to technical configuration previously required from the user. All failures were corrected and revalidated, and the shift in verification method revealed two additional defects that the first approach had failed to detect. Based on these findings, a ten-point verification checklist is proposed, reusable for auditing similar educational technologies. The study concludes that pedagogical commitments declared by AI-based educational systems require continuous technical verification, and that this practice constitutes a relevant formative exercise for computer science teacher education.