AI in higher education is often treated as a training problem: show faculty how the tools work, offer a few prompt-writing workshops, and expect meaningful classroom use to follow. I have never found that assumption convincing. Knowing how to operate an AI tool says very little about whether a teacher can use it well.
A new study by Suresh C. Joshi examines this gap through Intelligent-TPACK, an extension of the familiar TPACK framework. Its main contribution is the placement of ethical judgment inside teacher knowledge, where it shapes decisions from the start. This question belongs naturally within our wider discussion of AI and teaching practice.
Summary of the Study
Joshi used a two-phase mixed-methods design with university teachers at a private university in northern India. The quantitative phase surveyed 478 educators who had at least one year of experience using AI tools.
It examined how AI-specific technological knowledge, pedagogical knowledge, content knowledge, and ethical assessment relate within Intelligent-TPACK. The qualitative phase drew on interviews with 17 faculty members from several disciplines to explore how these relationships appear in actual teaching decisions.
The central finding is straightforward: technical familiarity with AI supported pedagogical and subject-specific uses, yet it did not lead directly to fully integrated AI teaching. Ethical assessment and pedagogical knowledge mattered greatly. Interviewed teachers described checking automated grades, overriding questionable recommendations, worrying about bias against non-native speakers, and balancing efficiency with instructional autonomy.
This develops the argument in the earlier post on Intelligent-TPACK and ethical AI integration: ethical judgment is part of the professional knowledge teachers use while deciding whether an AI output belongs in the classroom at all.
The interviews make the framework more persuasive because they reveal what the survey categories cannot. Faculty did not experience technology, pedagogy, content, and ethics as separate boxes. A single decision about AI-generated feedback could involve accuracy, disciplinary expectations, fairness, workload, and the teacher’s responsibility to the student.
Disciplinary differences also mattered. Some humanities educators questioned whether available tools fitted their ways of knowing and teaching, while faculty in data-intensive fields reported a wider range of applications. That point connects with TPACK in the age of generative AI, where context is treated as a force shaping every part of technology integration.
The study still asks its evidence to carry too much weight in places. The quantitative data are cross-sectional and self-reported, so they can identify relationships among perceived competencies but cannot demonstrate that one kind of knowledge caused another or that teaching improved.
Participants came from one institution, were recruited through convenience sampling, and already had AI experience. The statistical model was also revised after the initial analysis, which makes independent testing especially important. Interviews added depth, though the Intelligent-TPACK framework inevitably shaped the later interpretation of themes. The paper gives us a useful model of how teachers understand AI integration. Evidence of what they actually do in classrooms will require observation, instructional artefacts, and student learning measures.
Practical Insights
Faculty development built around this study would begin with instructional decisions. Tool demonstrations alone leave the hardest questions untouched. A workshop could give teachers an AI-generated grade, feedback comment, learning recommendation, or student-risk prediction and ask them to examine four questions:
- Is the output accurate enough for this instructional purpose?
- Could it disadvantage particular students or privilege one linguistic or cultural norm?
- Can the teacher explain how the recommendation was produced and defend its use?
- Who retains the authority to revise, reject, or override the output?
These questions turn ethical assessment into a teaching practice. They also make professional learning more concrete. Faculty can compare decisions across disciplines, identify where the same tool creates different risks, and document why human judgment changed an AI recommendation.
Universities could support this work through small interdisciplinary review groups. Faculty might bring one proposed AI-supported activity, examine its pedagogical purpose, test the tool’s output, and revise the activity with colleagues. Technical staff can explain system behaviour, while educators judge whether the application makes sense for their students and subject. This is more useful than a generic session built around a rotating collection of tools.
The study also has a clear implication for research. Future work should pair self-report measures with lesson plans, classroom observations, samples of AI-supported feedback, and performance tasks. Longitudinal and cross-institutional designs could show how Intelligent-TPACK develops over time and across disciplines.
Researchers should also test the revised model with new samples before treating its pathways as settled. This concern echoes broader reviews of AI literacy frameworks, which have found that proposed competencies often advance faster than their empirical validation.
The paper leaves us with a useful correction. AI proficiency is one part of teacher readiness. Responsible classroom use also depends on pedagogical purpose, disciplinary judgment, awareness of bias, and the willingness to question the system’s recommendations. A teacher who knows where to click may still be unprepared to decide when the tool should be trusted.
Reference
Joshi, S. C. (2026). Exploring Intelligent-TPACK among university teachers: A two-phase mixed-methods study. Journal of Computers in Education. Advance online publication. https://doi.org/10.1007/s40692-026-00395-0
