Physical Learning Environments and AI-Powered Personalized Learning in Higher Education: A Systematic Literature Review

Abstract

Personalized learning systems with artificial intelligence (such as adaptive learning platforms, intelligent tutoring, and AI-based assessments) often show strong results in rigorous research. In fact, in these situations, they improve upon normal teaching by between 0.42 and 0.76 standard deviations. However, when universities actually use them, they do not perform nearly as well as the research suggests. Most of what has been written about this difference says it is due to a lack of the right technology or to teachers not being prepared for it. However, it misses something we know affects how people think: the physical space where learning happens. This paper asks whether the quality of a physical learning environment affects the effectiveness of AI-powered personalized learning in universities.

This study conducted a thorough review of 39 studies that had been properly assessed by other experts and analyzed those studies. These came from AI in education research from 2006 to 2026, and from studies of how environments and behavior link to learning spaces, from 1999 to 2024. They found these resources on IEEE Xplore, Springer, Elsevier, MDPI, and Frontiers. The studies were selected to answer three questions: what parts of a physical environment affect how focused students are in their thinking? How do these parts of the environment connect to what adaptive AI systems need people to do to make the personalization work? Moreover, could the environment possibly explain why adaptive learning performs so much better in a lab than in a real university? The main limitation is that almost no existing studies measure both physical environment quality and AI tool performance in the same setting, which means the relationship cannot yet be tested directly.

Thermal discomfort (being too hot or cold), insufficient natural light, and noise frequently reduce students' ability to pay attention and control their own learning. Adaptive AI systems rely on students being able to do those things for personalization to be useful. Barrett and colleagues (2015) found that physical classroom factors alone account for 16% of the variation in student academic outcomes, a figure directly comparable to the effect sizes reported for AI adaptive learning systems. What is important is that when AI learning tools are used in research, they are mostly in carefully designed and managed spaces, which most universities do not have.

These results mean that how good the physical learning environment is is not something minor. It is actually essential for AI to improve education. Designing learning spaces and introducing AI systems should be done together as a single decision by the university, not as two completely separate plans.