Faculty must embrace the ‘messy middle’ to guide AI proficiency

Those shaping how future educators are learning to use artificial intelligence are often not teachers or scholars but the technology companies building the tools. Major platforms offer training sessions, lesson ideas and professional development for educators curious about tools such as ChatGPT or Copilot. These resources can be useful, but they also raise important concerns about the built-in biases and commercial agendas of the companies developing AI.

If educators don’t learn how to harness these tools and train teachers accordingly, the task will fall to people who know less than we do about what quality education looks like. As educators of teachers and history educators ourselves, we know that for universities to retain intellectual leadership in this space, we must more actively shape how AI enters teaching and learning. That means moving beyond abstract debates about whether AI is good or bad and engaging directly with what these systems do.

Policy as a starting point 

Many universities understandably begin with policy: which tools are allowed, how academic integrity should be protected, and guidelines for students. But policy often struggles to keep pace with the technology it seeks to regulate. Guidance, in fact, can have a shelf life not unlike a carton of milk left out on the faculty lounge table.

• Students are asking for AI guidance, not just policy
• How to create a higher education AI policy
• Why AI literacy must come before policy

This does not mean policy is unimportant. But meaningful guidance depends on educators engaging with the technology well enough to evaluate it critically. Without that, universities risk simply reacting to AI rather than shaping its role in education.

In practice, engagement often happens in what we might call the “messy middle” – the space between enthusiastic adoption and outright rejection, where educators experiment, test limits and discover how AI behaves in real academic contexts. We use the phrase “techno-curiosity” to describe this stage. It names a stance of disciplined exploration: a willingness to investigate what GenAI can do from the inside – in media res – matched by an insistence on asking what it cannot. Let us explain what we mean.

Layers of AI engagement

In our research and teaching, we think about the process of engaging with AI as moving through layers: configuration, systems, direct interaction and speculative futures.

Configuration is the environment in which AI use takes place. Institutional policies, disciplinary norms, personal beliefs about technology and even regional regulations all influence how educators and students approach AI. A history classroom, a computer science lab and a teacher-education programme will likely encounter the technology in different ways.

The second layer involves the systems themselves. Different platforms – whether ChatGPT,  Copilot or Gemini – have distinct capabilities, limitations and training contexts. Educators cannot meaningfully guide students without understanding how these tools behave, collectively and individually.

Our regular direct interaction with the technology is the third layer. This “liveness” is where the messy middle becomes unavoidable. It’s through use that AI systems reveal their quirks: their stylistic patterns, the confident inaccuracies they produce and the situations in which they can genuinely support learning. Educators who experiment with these systems often develop a practical intuition that cannot be gained through policy documents alone. They recognise when AI can provide useful feedback, scaffold ideas or support self-regulated learning – and when it is more likely to produce polished nonsense.

Finally comes the layer of speculative futures. Public discussion about AI often swings between extremes: utopian visions of limitless productivity and dystopian warnings of technological takeover. In reality, the decisions institutions make today will shape the future of AI in education. Educators remain central to those decisions. AI systems may suggest options or propose next steps, but humans determine which paths to follow. Students and teachers can accept those suggestions, ignore them or pursue entirely different directions. In other words, the algorithm may make recommendations, but it still does not get the final vote; we do.

Recognising agency in the final layer of AI engagement is essential. Without it, universities risk surrendering the development of educational practice to commercial interests whose priorities may not align with disciplinary learning or intellectual enquiry. The goal is not to reject collaboration with technology companies out of hand; their tools will undoubtedly remain part of the educational landscape. But universities should not approach these technologies as passive consumers; they should be critical partners. That requires curiosity, experimentation and a willingness to tolerate a bit of uncertainty.

The messy middle, after all, is where most meaningful learning happens.

Amy Allen is assistant professor of social studies in the Elementary Education programme, and David Hicks is professor of history and social science education, both at Virginia Tech. Their open access book, Teaching History with Chatty Geeps: A Technocurious Approach to Generative AI in the Classroom (VT Press, 2026), will be published in July.

If you would like advice and insight from academics and university staff delivered direct to your inbox each week, sign up for the Campus newsletter.