Qualidade de IA generativa em engenharia de software: estudo comparativo entre Claude e Codex
Revista ft · 2026-06-16 · 30 (159) · 01-24
DOI: 10.69849/efmbbt16 ↗At a glance
Journal article in Revista ft, published online on 2026-06-16. Written by Paulo Henrique da Silva Martins. The full text is freely available (open access, CC BY 4.0).
Abstract (original language)
Generative Artificial Intelligence (GenAI) tools have been increasingly adopted to support teaching and learning in higher education, including in computing. However, the technical quality of the answers these tools provide to Software Engineering questions is rarely assessed in a systematic and comparative way. This study aimed to comparatively evaluate the quality of the answers generated by two GenAI tools — Claude Sonnet 4.6 and Codex (GPT-5.5) — to ten technical Software Engineering questions, from an educational standpoint. A comparative case study with a mixed-methods approach was adopted, scoring three dimensions — accuracy, clarity, and depth — on a rubric and recording errors, complemented by objective verification of the generated code through execution in a controlled environment. The results showed full equivalence in accuracy and clarity, with no errors, and a statistically significant difference only in depth, favoring Claude and more pronounced in the practical questions. It is concluded that both tools are suitable as learning support for fundamental content, with complementary profiles, and that depth is a decisive criterion of educational quality. The study contributes methodologically by coupling rubric-based assessment with reproducible verification through code execution.
Cite
APA
Martins, P. H. D. S. (2026). Qualidade de IA generativa em engenharia de software: estudo comparativo entre Claude e Codex. Revista ft, 30(159), 01-24. https://doi.org/10.69849/efmbbt16
Vancouver
Martins PHDS. Qualidade de IA generativa em engenharia de software: estudo comparativo entre Claude e Codex. Revista ft. 2026;30(159):01-24. doi:10.69849/efmbbt16
BibTeX
@article{Martins2026,
author = {Martins, Paulo Henrique da Silva},
title = {Qualidade de IA generativa em engenharia de software: estudo comparativo entre Claude e Codex},
journal = {Revista ft},
year = {2026},
volume = {30},
number = {159},
pages = {01--24},
doi = {10.69849/efmbbt16}
}
EUVIMED