Martin Debattista, lecturer at ITS, lecturing on Digital Media, Digital Heritage Interpretation, Digital Tourism and ICT.

This forms part of a multipart blog series focusing on how academics are perceiving and dealing with the impact of generative AI on the education system.

There’s no beating about the bush. The temptation to get AI to write this blog post is enormous, and I do not doubt that with a few decent prompts, the machine would have made a decent job of it. Then put my name to it, and that’s it. Easy peasy.

Invariably, this is about AI. And yes, I have overcome temptation. I could not ignore the AI earthquake of the highest magnitude rocking education. I have been feeling the ground shaking under my academic feet and watching cracks across higher education getting bigger.

While I don’t claim to have definitive answers, I believe our path forward requires three things: honest institutional reckoning with what we’re actually assessing, educators modelling the responsible AI use we expect from students, and a fundamental redesign of assessment that prioritises what humans can do that AI cannot. This isn’t about banning or embracing AI wholesale: it’s about being intentional rather than reactive.

Another revolution? This time it’s different

Educators are tired of being told every few years that a new digital technology is about to “revolutionise” education, and watching technologists, with the help of slick marketers, try to shove it down our throats. Despite the hype fuelled by mainstream media, smartphones and tablets are increasingly being banned from classes, virtual reality is not present in every classroom, the “metaverse” was killed by its own creator, and MOOCs have not replaced universities.

So why should we believe AI is any different? Because, unlike those previous technologies, which offered new delivery mechanisms for education, generative AI directly challenges the core activities we ask students to perform: researching, writing, analysing, and creating. The printed book, the computer, the internet – each fundamentally transformed how knowledge was accessed and shared. AI is doing the same, but this time it’s automating the very tasks we use to measure learning itself. When the assessment becomes trivially easy to outsource, we cannot simply adjust our teaching methods. We have to rethink what we’re teaching and why.

This is genuinely different. The question is whether we’ll adapt thoughtfully or desperately.

A distress call—and what I told my colleague

Just a few days ago, I received an email from one of my younger academic colleagues:

“I’m sending this email to ask for guidance. One of my international students has submitted an assignment with a level of English which is definitely not hers. Turnitin has indicated 3% plagiarism and 71% possible AI-generated. I tried a couple of online AI detectors, and they have given mixed results. What way forward would you suggest?”

This had been brewing for some time, and despite our formal discussions and casual conversations over the past couple of years, we are nowhere near an effective solution.

Here’s what I told her: First, that AI detection scores are unreliable and should never be the sole basis for academic misconduct cases. False positives are too common, especially for non-native English speakers whose writing patterns can trigger detectors. Second, we need to examine our assignment design itself. If a student can complete it entirely through AI without demonstrating their actual learning process, then we’ve created an assessment problem, not just a student integrity problem.

What I suggested was a conversation with the student – not an interrogation, but a genuine dialogue. Ask her to walk through her research process, to explain her key arguments in her own words, to discuss what she found most challenging. If she used AI to polish her English, that’s a teaching moment about citation and transparency, not necessarily a violation. If she outsourced the entire thinking process, that reveals itself quite quickly in conversation.

But here’s the uncomfortable truth: this approach doesn’t scale. It requires time that most educators don’t have. It requires institutional support for redesigning assessments that most institutions aren’t yet providing. And it requires us to acknowledge that our old methods of catching cheating – the plagiarism detectors, proctored exams – are becoming obsolete faster than we can replace them.

Institutional responses and why we’re stuck

Many higher education institutions are developing and adopting policies on the use of AI by both faculty and students. The common approach is a visible emphasis on the “ethical” and “responsible” application of this “magical” piece of technology: it was Arthur C. Clarke who famously observed that any sufficiently advanced technology is indistinguishable from magic. The typical guidance suggests using ChatGPT, Claude, Gemini, Perplexity and others as aids: a sophisticated search engine, a grammar checker on steroids, acceptable as long as the original idea is yours, as long as you can demonstrate that you are doing the critical thinking. There is agreement that we are also in a race against time, where AI develops at such breakneck speeds that attempts to manage emerging risks are left far behind. AGI is just around the corner. We’re either on a road to nowhere or some brave new world where most white-collar workers will need to find a new job. That includes educators and those in the business of ‘education’.

But here’s where it gets murky. What does “your original idea” mean when AI can generate a dozen sophisticated starting points? How do we measure “critical thinking” when the student’s entire visible thought process might have been mediated by AI? We’re applying pre-AI concepts to a post-AI reality, and the contradictions are showing.

We often fall into the trap of focusing more on the technology itself than on its impact on the people using it, in a multi-generational workspace where different approaches to technology pose their own challenges. Gen Z students seem to want more guidance from their institutions on how to use AI responsibly. While they readily adopt the technology, many are worried about their future jobs, skill atrophy, and other societal risks, leaving them ambivalent about using AI in their studies. Gen X academics, meanwhile, tend to frame AI as one more digital tool in an already long trajectory of edtech adoption, emphasising critical thinking, creativity, and active learning when they integrate generative AI. They are generally more cautious: while they acknowledge the gains, they are more likely to stress the risks and call for regulation, training, and clear institutional guidelines before full-scale adoption.

Neither approach is wrong, but the disconnect creates institutional paralysis. Students want clarity; educators want autonomy. Administrators want policies that protect the institution; faculty want flexibility to experiment. And while we debate, AI capabilities advance monthly, leaving any consensus we reach outdated before it’s implemented.

Rethinking what we teach and how we assess

Research on generative AI is telling us that its use can both support and hinder critical thinking, and more work is needed to understand its negative impacts. Students surrendering their most fundamental skills and taking the easy route with GenAI is clear to see. But we can’t simply blame students for doing what we’ve inadvertently incentivised: finding the most efficient path to completing assignments that may not have been well-designed in the first place.

The real question isn’t “How do we stop students from using AI?” It’s “What do we need our students to actually learn, and how do we assess that learning in ways that can’t be easily outsourced?”

Do we really want to retreat to pen-and-paper exams in halls or resort entirely to oral examinations? These might be stopgap measures, but they’re not sustainable solutions for developing the skills students will need in AI-augmented workplaces.

There are ways of remaining relevant in an era where the death of traditional universities has been heralded time and again. One concrete approach is revisiting Bloom’s Taxonomy, not to replace it, but to recognise how AI has fundamentally shifted where human effort needs to be concentrated. The framework itself remains sound, but generative AI has essentially automated the lower levels. Remembering, understanding, summarising, and applying: these can now be performed with ease by AI systems.

What this means is that we need to shift our assessment design upward, toward higher-order thinking that requires human judgment, contextual understanding, and authentic engagement. As institutions such as UCD have suggested in their guidance, this means designing for:

• Contextual analysis: Not just “summarise this article,” but “analyse how this argument would need to be adapted for three different audiences, and justify your reasoning.”
• Defensible judgment: Not just “what’s the solution?” but “evaluate three possible approaches, explain the trade-offs of each, and defend your recommendation given specific constraints.”
• Reflective evaluation: Not just submitting a final product, but documenting the decision-making process, dead ends explored, and why certain paths were abandoned.
• Evidenced creation over time: Requiring portfolios that show iterative development, not just polished final submissions. If students are using AI, the portfolio should show how they critically evaluated and built upon AI-generated content.

Here’s what this looks like in practice. Instead of assigning “Write a 2000-word essay on digital heritage preservation,” I might assign: “Document your three-week process of developing a digital heritage preservation proposal. Include your initial research questions, how they evolved, three approaches you considered and why you rejected two of them, screenshots or notes from tools you used (including AI), and a reflective analysis of what you learned about your own thinking process.”

Can AI complete this? Partially. But it can’t fake authentic process documentation, it can’t demonstrate learning over time, and it can’t provide the kind of context-specific, personally reflective insights that emerge from genuine engagement with material.

Is this more work for educators to design and assess? Absolutely. Does it require institutional support, training, and perhaps adjusted teaching loads? Yes. But the alternative – continuing with easily-gamed assignments and then being surprised when students game them – is not sustainable.

There is deep uncertainty about how to rebalance learning goals, student autonomy, effective assessment design, and human-AI collaboration in higher education. Conflicting opinions and contradicting solutions are leading to paralysis.

Sorting ourselves out first

Here’s the uncomfortable truth: I don’t have all the answers, not only because I didn’t ask AI for them, but because educators and education leaders themselves need to sort out our own relationship with AI first. It would be hypocritical to come down hard on students without practising self-regulation.

AI-based fake papers have already been submitted to academic publications, complete with hallucinated authors and fabricated citations, creating enough AI “slop” to trigger warnings about the “destruction of knowledge.”

If we, as educators and researchers, can’t maintain our own standards, how can we credibly hold students accountable?

This means being transparent about our own AI use. When I use AI to help structure my thoughts or draft routine communications, I’m clear about it with colleagues. When preparing teaching materials, I document what AI-generated content I’ve adapted and why. This isn’t about confession: it’s about modelling the kind of thoughtful, transparent engagement with AI that we should expect from students.

It also means acknowledging our blind spots. Many of us are still learning these tools ourselves. We’re experimenting, making mistakes, and sometimes feeling overwhelmed by how quickly capabilities evolve. That’s fine, even necessary. But we need to create institutional cultures where this learning process is supported, not stigmatised.

Moving forward: Three starting points

If you’re an educator reading this, here’s how I believe we might begin to thrive, rather than merely survive, in the age of AI:

First, audit your current assessments honestly. For each assignment, ask: Could this be completed entirely by AI without the student learning what I intend them to learn? If yes, redesign it. This doesn’t mean every assessment needs to be AI-proof – that’s likely to be impossible. But it should require enough human engagement, contextual understanding, and process documentation that using AI becomes a tool for learning rather than a shortcut around it.

Second, experiment publicly and share honestly. Try incorporating AI explicitly into some assignments. Ask students to use AI to generate initial ideas, then critique and improve them. Have students compare AI-generated analysis with their own and reflect on the differences. Document what works and what doesn’t and share it with colleagues. We need collective intelligence on this, not isolated experiments.

Third, advocate for institutional support. This transition requires resources: time to redesign courses, professional development on effective AI integration, potentially reduced course loads during redesign periods, and technological infrastructure to support new assessment methods. If your institution is creating AI policies without providing these resources, speak up. Policies without support are just performative gestures.

None of this is easy. All of it requires sustained effort, institutional commitment, and tolerance for uncertainty. The alternative – pretending we can maintain pre-AI educational practices in a post-AI world – is a path to irrelevance.

The ground is shifting beneath our feet. We can choose to be intentional about how we respond, or we can wait until the cracks become chasms. I know which future I’m working toward.

How might we collectively build it together?