Navigating uncertainty: university teachers' experiences and perceptions of generative artificial intelligence

Synthesis: Navigating uncertainty — a qualitative study of 24 Swedish university teachers (higher education, disciplines centered on long-form writing: philosophy, law, sociology, education) who took part in assessment workshops with GAI outputs followed by focus-group interviews. Drawing on postphenomenology and technological mediation theory, the authors find that the emergence of GAI was experienced as alarming and overwhelming, inducing a state of vulnerability and ethical challenges (fairness, unequal access, bias), while prompting teachers to rethink assessment, re-evaluate teaching priorities (especially critical thinking), and worry that student learning is at risk (weakening of skills, desocialisation of learning). The study contributes a technology-mediation account of how teachers' roles and confidence are unsettled when GAI enters established educational practice.

Key Findings

• Experiencing the emergence of GAI as alarming and overwhelming. Teachers described a hype-driven public narrative ("all students are going to cheat") that demanded urgent institutional and personal response, alongside an opportunity framing: GAI forced reconsideration of what it means to know and to test knowledge in a system whose assessment mechanisms "don't really match" a new view of knowledge.
• A state of vulnerability. Teachers felt discomfort, insecurity, and low confidence driven by limited knowledge of GAI's capabilities/limitations, limited exposure, fear of "not being ahead of students," and worry about feeling incompetent when assessing student work potentially (co-)produced with AI. Junior teachers were seen as especially exposed.
• Emergence of ethical challenges. Concerns centered on (1) treating students fairly and avoiding falsely suspecting cheating (which could harm institution–teacher–student trust), (2) unequal access to GAI artifacts across paid/free tiers, and (3) GAI outputs perpetuating bias and inaccuracies that could misinform students.
• Rethinking assessment. GAI was seen as requiring re-evaluation of assessment purpose and design. Strategies discussed included "AI-proof" formats (banning home examinations, sit-in and oral exams, raising complexity), alternative/multiple and incremental submission types, transparency measures (honor statements, declarations of AI use), and more Feedback to help students see alignments/misalignments in GAI output.
• Re-evaluating teaching priorities. Teachers identified essential skills for learning with GAI — critical thinking, evaluating data sources, fact-checking output, "thinking on your feet," ownership of knowledge — and the responsibility to foster ethical use and academic integrity (authorship/ownership, avoiding plagiarism).
• Student learning at risk. Concerns about weakening skills through reliance on GAI (writing, reasoning, argumentation, research, on-the-spot responsiveness), the erosion of deep intuitive understanding, and the desocialisation of learning (students turning to chatbots instead of peers; GAI output missing the "social aspects" of disciplines like law).

Postphenomenological framing

The study is theoretically distinctive in grounding the analysis in postphenomenology and technological mediation theory (Ihde 1990; Verbeek 2006, 2011; Rosenberger & Verbeek 2015), treating GAI not as a neutral "tool" but as multistable technological artifacts that mediate teachers' perceptions and reconfigure their practices. On this account, GAI's "scripts" — rapid responsiveness, natural-sounding text that can achieve passing grades — shape teachers' experience of what competence means for learners and themselves, unsettle confidence, and push them beyond instrumental questions ("what use is acceptable?") toward re-evaluating their role, the meaning of teaching, and the future of the university in a landscape that feels "out of control." Theoretically, the paper applies an established philosophical framework (technological mediation) to the AI-in-education context rather than proposing a new one — an example of advancing an established theory into AIEd.

What this means for practice

• Instructors. Treat your own unease about GAI as a legitimate professional signal, not a gap in your competence. These experienced teachers (average 13 years) described a genuine "state of vulnerability," and the authors argue it calls for confidence- and Well-Being support rather than tool training alone.
• Instructors. Protect the pedagogical value of at-home long-form writing while adding transparency measures — honor statements and declarations of AI use — instead of defaulting to sit-in or oral exams, which the teachers warned could sacrifice self-paced, deep writing.
• Instructors. Make critical thinking, source evaluation, and fact-checking explicit course objectives — teachers identified evaluating data sources, checking output, and owning one's knowledge as newly central responsibilities for learners working with GAI, pointing to a reconfiguration of the teacher role around guiding critical engagement with GAI rather than transmitting content.
• Faculty developers. Create designated time and space for teachers to experiment with GAI, compare experiences, and draft guidelines collaboratively at institutional, departmental, and course levels, treating AI literacy and readiness as a supported collective endeavor — and give junior teachers targeted support, since respondents singled out colleagues with less course-specific teaching experience as especially exposed when assessing work potentially produced with GAI.
• Instructors and faculty developers. Treat unequal access as a design constraint, not a side issue: teachers raised paid versus free GAI tiers as a live concern, so avoid tasks that presume paid-tier features and treat equity as an access-and-infrastructure problem within the classroom, not only across countries or institutions.

Limitations

• The study is built on 24 teachers at one university in Sweden, purposively sampled from sociology, philosophy, law, and education courses in which long-form home essays dominate assessment; the authors state the findings are rooted in that setting and time frame and explicitly do not aim to generalize.
• The evidence is four 80–90 minute English-language focus groups of four to eight respondents, plus field notes — teachers' reported experiences and anticipations, with no classroom observation and no student outcome data.
• Coverage extends across three faculties (Social Sciences, Humanities, and Law) only; natural sciences, engineering, and medical education settings are not represented.
• Data were collected in spring 2023, in the first months after ChatGPT's public release, so the reactions are anchored to that early period rather than to teachers' settled practice.

Citation

Farazouli, A., Cerratto Pargman, T., Bolander Laksov, K., & McGrath, C. (2026). Navigating uncertainty: university teachers' experiences and perceptions of generative artificial intelligence in teaching and learning. Studies in Higher Education, 51(9), 1898–1913.