TPACK and AI: What Teachers Need to Know Beyond the Technical Skills

By /

I enjoy reading original research that challenges our long-held concepts and notions about education. Now that AI is reshaping how knowledge is created, represented, and assessed, it makes sense that our frameworks for understanding teacher knowledge need updating too. In this paper, Mishra, Warr, and Islam (2023) take one of the most widely used models in educational technology, the TPACK framework, and ask what happens to it when generative AI enters the picture.

Their answer: almost everything changes.

The Conversation Has Been Too Narrow

Mishra et al. open with an interesting observation. Most of the debate around AI in education has focused on plagiarism and cheating. Legitimate concerns, sure, but they occupy a narrow slice of a much larger conversation. The authors argue that “larger questions about the shifts in the very nature of teaching and learning have received less attention” (p. 239).

I’ve seen this play out in my own conversations with teachers and school leaders. The first question is almost always about academic integrity. How do we detect AI-generated work? How do we stop students from cheating? Important questions, but they keep the conversation at the surface.

Meanwhile, the deeper questions about how AI changes what it means to know something, how it alters the relationship between teacher and student, how it reshapes entire professional fields, those rarely get the attention they deserve.

TPACK and AI

TPACK and Generative AI: What Makes GenAI Different

The paper revisits three established properties of digital technologies: they’re protean (flexible across contexts), opaque (difficult to understand internally), and unstable (constantly changing). GenAI amplifies all three. But Mishra et al. argue it also introduces two characteristics that set it apart from every previous educational technology.

First, generativity. Every response to a prompt is produced in real time and can vary even when the same prompt is used again. The authors emphasize that “every response to a prompt is unique, created in real time” (p. 241). You’re no longer operating a tool. You’re in an ongoing exchange.

Then there’s the social quality. GenAI systems interact through natural language and sustain conversational continuity. We inevitably attribute agency and intention to them, even when we know the system has no understanding or consciousness. Mishra et al. describe these systems as psychologically “real” in the sense that users respond to them as if they were social actors. And that has consequences for how students engage, how they build trust, and how they process what AI tells them.

I covered the trust dimension in a previous post on cognitive surrender, where Shaw and Nave (2026) showed how students accept AI-generated outputs without critically evaluating them. Mishra et al. help explain why: the conversational nature of GenAI makes it feel like a credible interlocutor. The system speaks with fluency and confidence, and that can easily be mistaken for accuracy.

TPACK in the Age of Generative AI: Rethinking Pedagogical Knowledge

The paper’s treatment of Technological Pedagogical Knowledge (TPK) is where things get most practical. Mishra et al. suggest that teachers need to rethink assessment, feedback, and classroom interaction in light of what GenAI can do.

One example they offer: have students generate drafts with AI and then critique or annotate them. The learning shifts from production to evaluation. Students aren’t assessed on whether they can write a first draft anymore. They’re assessed on whether they can identify weaknesses in an argument, spot logical gaps, recognize bias, and improve coherence. These are higher-order thinking skills, and GenAI opens up new ways to practice them.

Teachers also need to understand hallucination, bias, and prompt design well enough to help students navigate these limitations. Mishra has described GenAI elsewhere as a “smart, drunk intern” (love that metaphor, Mishra, 2023), someone who produces impressive-sounding work but can’t be fully trusted without careful oversight.

Celik (2023) arrived at a similar conclusion through a different methodology. His Intelligent-TPACK study found that pedagogical knowledge, the ability to interpret AI outputs and connect them to instructional goals, was the strongest predictor of effective AI integration. Technical knowledge alone didn’t get teachers there. Mishra et al. provide the conceptual argument for why that’s the case.

Content Knowledge and the Future of Professions

The paper pushes the analysis beyond pedagogy into content domains. Law, journalism, coding, data analysis, and other knowledge-intensive professions are already changing as AI automates cognitive tasks that were previously considered high-skill. Human expertise is shifting toward problem-framing, output evaluation, and professional judgment.

Mishra et al. raise a structural concern that educators need to take seriously: “AI is likely to reduce employment for college-educated workers by automating tasks previously considered high-skill” (p. 243). If that trajectory holds, educational institutions need to reconsider what they’re actually preparing students for. The competencies that matter in an AI-saturated economy may look very different from what current curricula prioritize.

Mishra et al. expand the TPACK framework’s treatment of contextual knowledge (XK) to include policy, institutional constraints, and cultural forces. A teacher might design a brilliant AI-integrated lesson only to run into a district-wide ban on ChatGPT. Context shapes everything.

Here’s a point I think deserves more airtime: technologies rarely change schooling directly. They change the world in which schooling operates. Social media is the cautionary example. Educators focused on classroom applications of Twitter and YouTube. Few anticipated the long-term erosion of trust, polarization, and mental health consequences that followed. GenAI could follow the same pattern, and the most significant educational impacts may come from outside the classroom entirely.

The paper’s most ambitious contribution is philosophical. Mishra et al. argue that GenAI collapses older categories like tool versus machine. These technologies invite a relational stance. We interact with them through dialogue. We shape them and they shape us right back.

They write: “we’re not just users or operators, we’re co-creators, shaping and being shaped by these technologies in a continuous and dynamic process of co-constitution” (Mishra et al., 2023, p. 245).

I find that compelling and a little unsettling at the same time. It means the questions we ask about AI in education can’t stop at “how do we use it?” We also need to ask how it’s using us, how it’s changing our habits, our expectations, and our relationship with knowledge itself.

Reference

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top