Use of technological resources in the era of artificial intelligence and its effect on academic performance

Marín Y.R.; Caro O.C.; Tuesta J.N.A.; Huatangari L.Q.; Sánchez R.A.; Guevara J.L.M.; Santos R.C. · Frontiers in Computer Science (2026) · DOI: 10.3389/fcomp.2026.1876401

Resumen

The advancement of digital technologies and generative artificial intelligence (GenAI) has transformed learning environments in higher education; while tools such as ChatGPT offer personalized experiences and immediate feedback, their relationship with academic performance remains insufficiently understood in university contexts. This study analyzed the use of technological resources in the AI era, considering frequency of use, perceived usefulness, familiarity, and type of use, as well as their contribution to university students’ academic performance and identifying technological and contextual factors associated with it. A data-mining approach was employed, applying 10 supervised regression algorithms using 10-fold cross-validation, complemented by interpretability analysis using SHAP values to identify relevant predictors. Regularized regression models (Elastic Net; Lasso) showed the best relative predictive performance. The SHAP analysis revealed that the field of study (≈ 0.41) and study cycle (≈ 0.34) are the dominant predictors, while the use of AI for academic writing (≈ 0.10) emerged as the technological variable with the greatest positive contribution to academic GPA within the model. Academic performance was mainly associated with structural factors related to students’ educational trajectories, while AI tools showed a complementary role. The predictive contribution of technology appears to depend more on pedagogical integration than on frequency of use alone. These findings suggest the need to integrate digital strategies into higher education, implementing them in a differentiated manner by discipline and academic level, and promoting the guided and purposeful use of AI and technological resources in learning processes. Copyright

Instituciones: Pontificia Universidad Católica del Perú; Universidad Nacional Toribio Rodríguez de Mendoza de Amazonas