Same AI, Different Eyes: Educators and Students Read University Change in Opposite Ways

Artificial intelligence has swept through universities faster than almost any educational technology before it, but a new study suggests that the people living through this transformation are not telling the same story about it. Research conducted in Singapore with 223 university participants, including 71 educators and 152 students, has found that educators and students recognise largely the same AI-driven developments in higher education, yet consistently interpret their significance through the lens of their institutional roles. The study, published in Discover Education, adds a temporal dimension to a debate that has often been frozen in the present, asking participants not only what AI is doing now, but what it has already done and what they expect it to do next.

The research, led by Chin-Siang Ang of the Lee Kong Chian School of Medicine at Nanyang Technological University and TMC Academy, adopted an exploratory qualitative design built around a deceptively simple instrument: a sentence-completion survey. Participants were asked to finish three prompts, one addressing changes AI had already brought to education, one describing how AI is currently changing it, and one anticipating future transformation. Sentence completion, a semi-structured qualitative technique, encourages people to express their views in their own words while keeping responses focused on a shared topic. Responses were collected anonymously online during the first half of 2025, following institutional review board approval, and analysed in NVivo 15 using a two-stage process combining inductive thematic analysis with content analysis.

The analytical pipeline was designed to let patterns emerge from the data rather than imposing predefined categories. Every response was read repeatedly, coded inductively for its principal idea, and grouped into broader thematic categories through constant comparison. Where participants expressed more than one distinct idea, each idea was coded separately to preserve the breadth of interpretation. To check consistency, roughly 20 percent of responses were independently re-coded two weeks after the initial coding, yielding an agreement rate of 89 percent. The resulting thematic framework was then applied as a consistent structure for comparing patterns across the two participant groups and across the three temporal prompts, with frequency counts used descriptively rather than as measures of statistical significance.

When asked what AI had already changed, educators and students named many of the same broad developments but located them in different corners of university life. Educators most frequently cited the need to redesign assessment, with 58 responses on that theme, followed by the reduced effectiveness of plagiarism detection and authorship verification, the normalisation of technology dependence, the elimination of assessment tasks that AI can readily complete, and the continuing shift away from memorisation-based learning. Students, by contrast, centred their accounts on learning support and the practical organisation of study: 82 responses highlighted increased access to information and academic support, followed by a higher baseline quality of submitted work, reduced time to complete assignments, changing expectations about which academic skills matter, and more personalised learning support.

The pattern repeated for ongoing change, though with a subtle shift in tone. Educators described the present in terms of uncertainty and adaptation, most prominently uncertainty about the authenticity of student work, cited in 64 responses, alongside the need for continual adaptation of teaching practice, changing classroom dynamics, increasing workload despite promised efficiency gains, and a growing emphasis on learning processes rather than content acquisition. Students again emphasised the texture of daily study: immediate access to learning support, changing approaches to completing assignments, reduced anxiety when facing academic difficulties, opportunities to explore topics beyond course requirements, and, notably, pressure to keep pace with peers who were also using AI. Both groups agreed that academic practice was still evolving; they simply disagreed, in effect, about where the change was happening.

It was the future where the two narratives pulled furthest apart. Educators anticipated declining validity of traditional assessment, reduced value placed on human expertise and teaching, diminished deep learning and critical thinking, widening educational inequalities, and a changing meaning attached to academic credentials. Some responses reached further still, questioning the purpose of higher education itself and the commodification of learning. Students projected almost the opposite trajectory: expanded access to high-quality learning, a shift from knowledge acquisition towards skill application, more personalised learning pathways, faster degree completion, and the normalisation of lifelong learning. Compared with the past and present prompts, relatively few themes were shared across the two groups, suggesting that expectations about AI become increasingly shaped by the priorities attached to institutional roles.

The study frames these divergences through the concept of sensemaking, the idea that people construct meaning when situations challenge established routines and expectations. From this perspective, the findings do not reveal two camps holding opposing opinions about AI. Instead, they reveal two vantage points onto the same process. Educators are responsible for designing learning experiences, evaluating achievement, and maintaining academic standards, so it is unsurprising that they interpret AI primarily through assessment, authenticity, and the credibility of academic work. Students encounter AI most directly while studying, completing assignments, and managing the demands of university life, so their accounts naturally centre on support, efficiency, and access. The perspectives are complementary rather than contradictory, two readings of a single transformation experienced from different positions within the institution.

The temporal design also offers an explanation for why the wider literature on AI in higher education can appear conflicted. Studies emphasising learning support tend to report practical benefits for students, while those focused on assessment frequently highlight concerns about integrity and validity. The new findings suggest these accounts may simply capture different moments and different roles within a continuing process of educational change. Participants described coherent narratives rather than isolated observations: educators moved from completed changes in assessment, through continuing adaptation in teaching, to long-term concerns about standards and institutional purpose, while students moved from increased access to support, through the integration of AI into everyday study, to expectations of flexible and personalised education. Expectations about the future, the study notes, were closely connected to present experience, meaning that future-oriented responses should be read as reflections of current educational realities rather than predictions.

The implications reach into policy and assessment design. Because educators and students do not start from a shared understanding of AI, institutional strategies that assume consensus may fail to address the concerns of either group, and the study argues that developing AI policy through consultation with multiple stakeholder groups could produce clearer expectations and stronger confidence in institutional responses. On assessment specifically, the findings point away from a primary focus on detecting AI use and towards creating opportunities for students to demonstrate reasoning, judgement, and disciplinary understanding, with approaches that make learning processes more visible likely to reflect contemporary realities better than restriction or surveillance alone.

The study acknowledges its limits. Sentence completion captures concise interpretations across a large sample but lacks the depth of interviews or classroom observation; participants were not given a formal definition of impact, so responses reflect their own reference points; the sample was drawn through snowball sampling from universities in a single national context; and the data represent a snapshot during a period of rapid change rather than a longitudinal record. Future work, the author suggests, could include academic leaders, learning designers, librarians, employers, and policy makers, and could follow educators and students over time. What the study establishes in the meantime is a point that universities cannot afford to ignore: as AI continues to evolve, institutional responses will depend as much on how educators and students interpret what the technology means as on what the technology can actually do.

Subject of Research: Educators' and students' interpretations of artificial intelligence-driven educational change in higher education across past, present, and future perspectives

Article Title: Educators’ and students’ views on the impact of artificial intelligence in higher education

Article References: Ang, C.-S. (2026). Educators’ and students’ views on the impact of artificial intelligence in higher education. Discover Education, 5(1), Article 967. https://doi.org/10.1007/s44217-026-02174-6

Image Credits: AI Generated

DOI: 10.1007/s44217-026-02174-6

Keywords: artificial intelligence, higher education, generative AI, assessment, academic integrity, sensemaking, qualitative research, students, educators, educational change, Singapore, sentence completion