Nursing educators' perspectives on the integration of artificial intelligence into academic settings

Synthesis: A Technological Pedagogical Content Knowledge (TPACK)-framed phenomenological study of nursing educators in Bangladesh finds that the case for AI in Nursing Education is made almost entirely in the language of time and attention. Grading, progress tracking, and simulation free educators for mentorship, while Personalized Learning and early warning about struggling students are valued for what they let a teacher do next. Set against that, the barriers participants described are not about belief but about capacity and governance: no training, unreliable connectivity, and unanswered questions about who holds student data and whether an algorithm judges students fairly. The authors read their five themes as evidence that adoption depends less on tool quality than on Teacher AI Competency and institutional policy.

Key Findings

1. Thematic analysis produced five themes: (1) perceived benefits of AI, (2) barriers to AI integration, (3) ethical considerations in AI use, (4) educator readiness and adaptation, and (5) AI as a tool for personalized learning.
2. The benefits theme ran through two subthemes, enhanced teaching efficiency and improved student engagement. Educators described grading and progress tracking as work the technology absorbed, and virtual simulations as making clinical scenarios interactive rather than textual.
3. The barriers theme paired lack of training and support with infrastructure limitations. One participant reported abandoning a tool after failing to understand it, another described classrooms that were not equipped for AI tools, and a third attributed the stall to enthusiasm without resources.
4. Ethical concerns clustered on data privacy and bias in AI algorithms. Participants asked who has access to student data and what happens to it once entered into a system, and warned that a biased algorithm could misjudge students' abilities.
5. Readiness varied with prior exposure. Of the 30 participants, 43% had attended an AI-related conference or training, and the authors report that educators with such exposure showed greater confidence while those without it raised ethical, institutional, and technical concerns. Educators with under 10 years of experience were described as more open to AI tools and senior educators as skeptical.
6. The personalization theme covered tailored learning experiences and early identification of struggling students: adapting pace and content to individual progress, and flagging patterns a teacher might otherwise miss in time to intervene.

Method, and what the TPACK lens was asked to carry

The design is a qualitative phenomenological study using semistructured interviews and focus group discussions, reported against the Standards for Reporting Qualitative Research. Data collection ran between 10 July 2024 and 30 September 2024, with 14 semistructured interviews of roughly 50 to 60 minutes each, conducted in person or over Zoom, and three focus groups of four, five, and seven participants respectively. The abstract describes three focus group discussions involving 16 participants from three nursing colleges in Bangladesh; the results section counts 30 participants in total, the 14 interviewees plus the 16 focus group participants. Institutions were chosen in two steps, first for advanced classroom technology and then by random selection from that pool, two colleges in Dhaka and one in the Chittagong division, and the sample was closed on data saturation. Coding was deductive in orientation, mapped onto technological knowledge, pedagogical knowledge, and content knowledge, with two researchers coding independently and a third reviewer resolving disagreements.

That choice of framework shapes what the study can say. Technological Pedagogical Content Knowledge (TPACK) treats effective integration as the intersection of three knowledge domains, and the themes land unevenly across them. Technological knowledge is where participants spoke most concretely: which tools they had met, whether they had attended training, whether a simulation ran. Pedagogical knowledge appears mainly as a hope about time, that mentorship and course design would expand once grading contracted. Content knowledge, the nursing-specific judgment about which clinical competencies a tool actually develops, is the thinnest of the three in the reported quotes, surfacing as patient scenarios and clinical skills rather than as reasoning about what a curriculum should require, and the integrated knowledge at the center of the model is exactly what the readiness theme says is missing. The study therefore documents a pre-integration state rather than integration itself, which is consistent with its own finding that educators want training before they want tools.

Readiness as a distribution, not a deficit

The readiness theme is usually summarized as educators being unprepared, but the detail is more useful than that. What separates the confident from the hesitant in this sample is exposure, not seniority: attendance at a conference or training, prior experimentation, and colleagues who model use. The age pattern the authors report runs in the opposite direction from the usual assumption about digital natives and institutional memory, with senior educators skeptical and newer faculty more open.

The policy consequence is that a single training event addresses the wrong unit of analysis. The gap participants describe is between colleagues inside the same department, and the remedy they ask for is practice rather than instruction: workshops where they can use a tool, sessions that teach use, professional development as a priority rather than an afterthought. That is a request for Workplace Learning with conditions attached, and it is also a request for someone to decide who pays for it and when it happens. A published letter responding to this study argues the gap it reports is institutional rather than individual, and proposes an implementation-science readiness checklist in place of a competency frame (Bridging implementation gaps: the critical role of institutional ecosystems in AI integration for nursing education).

Ethics arrives as a governance question

The ethical themes in this study are not abstract. Participants asked about custody and access, meaning where student data is stored, who can reach it, and how long it is kept, and they treated those questions as unresolved rather than as risks to be managed by individual caution. The same is true of bias: the worry is not that an algorithm is biased in general but that it will sort real students into categories that stick, which is why Bias Mitigation and Privacy show up here as institutional obligations rather than as design preferences.

Read alongside the personalization theme, this is a sharper tension than the paper fully draws out. The features educators most valued, adapting content to individual progress and flagging students who are falling behind, are the features that require the most student data and the most consequential classification. The study's participants wanted both and did not describe any mechanism that would let them have both, which is where AI Governance and Equity enter: the choice between a system that personalizes and a system that protects is made in policy, not in the classroom.

What this means for practice

• Faculty developers. Build training around hands-on use of the specific tools a school has licensed, not around AI in general, and design it for mixed proficiency in the same room, since the reported gap sits between colleagues rather than between generations.
• Administrators. Treat connectivity and classroom hardware as the precondition for any AI adoption plan; participants described infrastructure failure, not resistance, as the thing that stopped them.
• Curriculum designers. State explicitly what students are expected to learn from a simulation or an adaptive platform, because the pedagogical and content knowledge that would justify a tool is the part of Technological Pedagogical Content Knowledge (TPACK) this sample reported least.
• Institutions. Publish data governance rules before pilots, covering storage, access, retention, and who can contest an automated judgment about a student, so that educators are not left improvising answers to the questions participants kept raising.
• Instructors. Where a tool flags a struggling student, keep the interpretation with the teacher: the value participants described was the chance to intervene early, not the classification itself.

Limitations

• The three colleges were selected for advanced classroom technology before random selection, so the sample represents relatively well-resourced institutions and the authors note this limits transfer to settings with fewer resources.
• The sample is reported inconsistently: the abstract gives 16 participants in three focus groups, while the results section reports 30 participants in total, 14 interviews plus 16 focus group participants. Figures in this page follow the results section, and the discrepancy itself should be read as a reporting weakness.
• Themes were developed deductively against the TPACK domains and derived from the most frequent codes, so the five themes reflect what was said most often rather than what was said most distinctively.
• All data are self-report from educators in one country, with no observation of AI use in teaching and no student outcome data, so the benefits described are anticipated rather than demonstrated.
• The paper reports no participant-level account of how the 14 interviewees and 16 focus group participants differed, and the group discussion data are presented through quotes without indicating how consensus or disagreement played out within groups.

Citation

Rony, M. K. K., Ahmad, S., Tanha, S. M., Das, D. C., Akter, M. R., Khatun, M. A., Begum, M. H., Khalil, M. I., Peu, U. R., Parvin, M. R., Alrazeeni, D. M., & Akter, F. (2025). Nursing educators' perspectives on the integration of artificial intelligence into academic settings. SAGE Open Nursing, 11, 1-17.