What shapes student perceptions of AI in university classrooms?

A major cross-institutional survey reveals how student attitudes toward generative artificial intelligence vary sharply depending on university, academic discipline, and demographic factors. Conducted at Ulster University in Northern Ireland and the Indian Institute of Technology Delhi, the research provides one of the most detailed comparative snapshots to date of AI’s reception in higher education classrooms.

Published in Education Sciences, the study titled "AI Across Borders: Exploring Perceptions and Interactions in Higher Education" analyses responses from more than 1,200 students across undergraduate and postgraduate programmes. It uses both statistical modelling and sentiment analysis to untangle how institutional culture, subject specialisation, and gender interact in shaping perceptions of AI’s role in teaching and learning.

Measuring awareness and perception across borders

The research team deployed a structured online survey adapted from earlier studies to assess two central dimensions: awareness of generative AI and perceptions of its value and risks in academic contexts. Students rated statements on a five-point scale covering personal and instructor use of AI, as well as attitudes toward its integration in coursework and assessment. Optional open-ended questions captured free-text comments that were later examined for sentiment and recurring themes.

The sample included 512 students from Ulster University and 699 from IIT Delhi. Participants were recruited through official institutional channels and represented a diverse spread of disciplines, from computer sciences and engineering to humanities, social sciences, and business. By combining two institutions in very different educational and cultural contexts, the study was able to probe whether differences in perception were driven by local academic environments or by broader trends in AI adoption.

An exploratory factor analysis revealed that the survey’s multiple items largely reflected a single underlying construct, indicating that student responses were generally consistent across different aspects of awareness and perception. This streamlined factor was then used in regression models to examine how institution, subject area, and gender, as well as their interactions, influenced attitudes toward AI in education.

Where you study matters more than who you are

Overall, students at IIT Delhi rated their awareness and perceptions of AI significantly higher than those at Ulster University. This gap held across most subject areas, suggesting that broader institutional and cultural contexts strongly shape how students view AI in learning environments.

However, the data also revealed an important nuance: in computer sciences, perceptions were virtually identical between the two institutions. This exception points to the role of subject-specific exposure in narrowing institutional gaps, with computer science students, regardless of location, having greater familiarity and perhaps more hands-on engagement with AI tools.

The study further explored whether gender differences played a role in shaping perceptions. At first glance, male students appeared to report more positive attitudes toward AI. But once the models accounted for institution and subject area, this apparent gender effect disappeared. The result indicates that gender differences were not independently driving attitudes but were instead a reflection of how men and women were distributed across disciplines and institutions.

These findings underscore that institutional culture and curriculum design may have greater influence over student perceptions than demographic factors alone. In effect, where and what you study may matter more than who you are when it comes to forming views on AI in education.

Balancing optimism with concerns

While the quantitative data highlighted measurable differences in perceptions, the qualitative analysis added depth to the story. Using sentiment analysis and aspect-based sentiment analysis, the authors found that student comments from IIT Delhi were generally more positive than those from Ulster University. Across both institutions, however, certain patterns emerged.

Students frequently noted AI’s usefulness for academic tasks, such as generating ideas, refining writing, summarising complex materials, and providing quick access to information. Many recognised that AI could enhance learning efficiency and provide personalised support that traditional resources often lack.

At the same time, caution was a recurring theme. Concerns clustered around three main issues: the potential for plagiarism and other breaches of academic integrity; environmental costs associated with large-scale AI computation; and the risk of factual inaccuracies or biased outputs from AI systems. These apprehensions reflect a growing awareness that while AI can offer significant benefits, its adoption in education requires careful management, ethical safeguards, and transparent guidance.

The research team’s multi-method approach highlights that these attitudes are not purely about technological capability - they are shaped by trust, policy frameworks, and the perceived alignment of AI tools with academic values. In particular, differences in institutional messaging and infrastructure for AI integration may explain why one institution’s students express higher confidence and acceptance than another’s.