Toward a Sustainable Innovation in the Faculty Evaluation Process: A Systematic Review of AI, Data Science, and NLP Applications in Higher Education

Cristian Ulises Barenca-Sotelo — 2026-02-23

Context. Teaching evaluation in public multicampus universities still relies on averages and closed-ended surveys, which are limited in capturing the complexity of academic performance. In response, this study presents a systematic review (2019–2024) on the use of Artificial Intelligence (AI) and Natural Language Processing (NLP) in university teaching evaluation processes, conceptualized as process innovations from Oslo Manual and aligned with Sustainable Development Goals (SDGs) 4 and 9.

Problem. Traditional systems in high-enrollment institutions exhibit low analytical precision, limited use of qualitative data, and delayed feedback. Accordingly, the following research question is posed: which techniques and algorithms does the literature report for the integrated analysis of quantitative and qualitative data in university teaching evaluation?

Purpose. To critically analyze the use of data science and AI—particularly NLP—to enhance teaching feedback, overcoming the limitations of traditional approaches and promoting timely, in-depth, and scalable evaluation processes.

Methodology. The PRISMA protocol was followed through searches using Boolean operators in Scopus, Web of Science, IEEE Xplore, and ERIC. After applying inclusion criteria and a two-phase peer-review process, 17 studies published between 2019 and 2024 were analyzed.

Findings. Techniques such as sentiment analysis, topic modeling (LDA), and large language models (LLMs)—notably DistilBERT, with accuracy levels close to 93%—consistently outperform traditional methods in managing large volumes of information.

Originality. The study’s originality lies in integrating dispersed literature on AI and NLP in higher education within a coherent process-innovation framework, combining technical, empirical, and ethical rigor.

Conclusions and Limitations. Advanced AI and NLP models show high potential to transform evaluation in Latin American university networks by enabling personalized feedback. However, challenges remain regarding model interpretability, bias mitigation, the scarcity of labeled data in Spanish, and institutional resistance to change, opening relevant avenues for future research and applied developments

Scientia et PRAXIS

Toward a Sustainable Innovation in the Faculty Evaluation Process: A Systematic Review of AI, Data Science, and NLP Applications in Higher Education