Virtual Laboratories Assisted by AI with Intelligent Feedback for Sustainable Engineering: A Quasiexperimental Evaluation in Agricultural Irrigation
Academic Article in Scopus
Limited access to physical laboratories in engineering hinders equity, experimental repeatability, and the connection with real-world challenges. This study evaluates the integration of virtual laboratories assisted by Generative AI with intelligent feedback to enhance practical teaching in the course Physical Experimentation and Statistical Thinking, through an authentic challenge: designing a sustainable agricultural irrigation system using a Mariotte Bottle. A quasi-experimental design with a mixed-methods approach was implemented with 93 students (control group n=47 with physical practice; experimental group n=46 with physical practice plus simulations). Instruments included pre/post-tests, validated Likert surveys (¿>0.85), standardized rubrics, and qualitative analysis of reflections. Statistical analysis (t-test, ANOVA, correlations, and effect sizes; power >0.80) showed significant advantages for the experimental group in conceptual retention, motivation, and consistency in problem-solving. The normalized gain was g=0.67 compared to g=0.34(p<0.05). Reflections explicitly linked learning outcomes with the Sustainable Development Goals (SDGs) 2, 4, 6, 9, 12, and 13, emphasizing ethical and social dimensions. Overall, the findings indicate that AI-assisted virtual laboratories complement rather than replace physical practice, providing a replicable, scalable, and ethically grounded model for STEM innovation in higher education.
