AI Academic Integrity: 83% of Irish Academics Still Have Concerns
Almost 90% of Irish academics surveyed by two unions say they meet AI issues in their work, and 83% still have concerns about academic integrity as institutions test policies that run from prohibition to full engagement.
Key Takeways
- The Irish Times reports that 54% of academics surveyed by IFUT and TUI meet AI issues frequently and another 33% meet them occasionally.
- 83% of the academics surveyed still have concerns about academic integrity and plagiarism.
- The Technical University of the Shannon has built a tiered program of AI integration that runs from prohibition to full engagement, with the level of AI use set by the outcomes academics want for students.
- Checking student understanding face to face takes staff time and money, which makes the cost of verification a budget question for every university leader.
Bring Alex to your organization: keynotes, workshops and advisory, in person or virtual. Check Alex's availability
More than half of the Irish academics surveyed by their unions, 54%, now deal with AI issues frequently, and 83% still have concerns about plagiarism and AI academic integrity. The Irish Times reports the figures from a conference on AI in third level, organized by the Irish Federation of University Teachers (IFUT) and the Teachers' Union of Ireland (TUI) and held on Friday at Technological University Dublin.
Add the 33% who meet AI issues occasionally and the total comes to almost 90%. For most of the academics in this survey, AI is now an ordinary part of the working week, and the paper describes policies on handling it as still developing.
One limitation belongs up front. The report does not give the survey's sample size, method, dates or response rate, and it does not say whether the two unions ran it jointly. The percentages are a clear indication from the people doing the teaching, and they work best as a prompt for your own questions.
Irish institutions are already testing answers
The more useful part of the report is what institutions are trying. Denise MacGiolla Rí of the Technical University of the Shannon described a tiered program of AI integration that runs from prohibition to full engagement. In MacGiolla Rí's account, academics need to be clear about the outcomes they want for students when deciding the level of AI use in a course.
Some academics said their institutions had considered a more open model, where students may use AI in any way they like as long as they can show they understand the results and how they were reached. Speakers also said face-to-face engagement with individual students must be maintained to establish human judgment in the work students submit.
Both ideas point in the same direction. When a machine can produce the written work, the evidence of learning moves toward the conversation about it.
The four-year bet now rests on showing understanding
I call a degree the four-year bet. Students and their families wager years of time and money that the credential at the end will still mean something to the employers and communities who read it. For generations the bet rested on a simple chain of trust, in which the essay and the exam script were the proof that a person could do the work.
Read the 83% figure in that light. A large majority of the academics surveyed still worry about whether that proof can be trusted, and the idea of letting students use AI freely while asking them to explain their results is an attempt to rebuild the chain on firmer footing. A degree is a promise about a person. The submitted work was always the evidence for that promise, and the evidence is changing.
If you teach, here is a plain question: how many submissions did you read twice last term, wondering how much of the thinking was the student's own?
Assessment carries the shortest curriculum half-life
Curriculum half-life is the time it takes for half of what a program teaches, or the way it tests, to stop doing its job. In my experience the content of a course holds up longer than its assessment design. A take-home essay can still cover the right material and tell you far less than it once did about who did the thinking.
A tiered approach like the one at the Technical University of the Shannon fits this problem, because it asks each module to set its AI level against its outcomes. Those settings will need revisiting more often than a syllabus does. For a wider view, I wrote recently about a University of Minnesota consensus report on AI and learning.
Verification costs money, and someone has to carry it
Checking understanding face to face takes staff time, and the report notes that monitoring is time-consuming and expensive. James O'Sullivan of University College Cork put the question plainly.
“Who is going to pay for all of that?” (James O'Sullivan, University College Cork, as reported by The Irish Times)
O'Sullivan also said: “Higher education was already underfunded before all of this came along.” Academics at the conference said funding could be used to better control the use of the technology among students. The cost of a credible degree now includes the cost of checking understanding, student by student, and that line belongs in the budget conversation from the start.
As Innovator-in-Residence at Tulane University, and as an advisor to a university system's AI work, I keep returning to a short set of questions. I'm taking them into my own meetings, and you are welcome to borrow them.
The first is about outcomes. What do we want a student to demonstrate at the end of this module, in a form we can still trust a year from now? MacGiolla Rí's point applies directly here, since the right level of AI use follows from the outcome.
The second is about cost. If the answer involves a conversation with each student, what does that cost in staff hours across a full cohort, and where does the money come from? O'Sullivan's question deserves a number attached to it before any policy is signed off.
The third is about ownership. Who in the department owns the assessment design, and who decides when it has expired? A policy without a named owner sits unchanged long after the tools have moved on.
Students get a question too. Could you explain how you reached your last answer, to a person, without the tool open?
Other campuses face the same bet
Employers hire on the strength of a degree, and families take on real costs to earn one. I wrote recently about the ASU agentic AI conference, and more of this thinking sits in my AI in education collection. It is also a thread through my forthcoming book, The Great Relearning. A degree that can prove understanding keeps the four-year bet worth making.
Nobody has this solved yet. The two unions behind this survey, and the institutions experimenting with tiers, are doing the useful work of putting the problem on the table in public.
So take one question into your next department meeting: which assessment in your program would you still trust next September, and who owns the answer?
What does the Irish Times report leave out about the survey?
The report gives the headline percentages but not the sample size, method, dates or response rate, and it does not say whether IFUT and TUI ran the survey jointly. In my view, anyone planning to cite the 83% figure in a policy paper should ask the unions for the methodology first, since that detail would show how far the figures can be generalized.
What could a first practical step look like for a single department?
My suggestion is to start small, with one assessment. Decide what students should be able to demonstrate and the level of AI use that fits that outcome, in the spirit of the tiered approach described at the Technical University of the Shannon. Then estimate the staff hours a short conversation with each student would take, so the department has a real cost and a named owner before anything grows.
Here is what makes Alex a credible voice on this topic: Alex is Innovator-in-Residence at Tulane University, advises a university system's AI work, and shaped Cisco's $1.1B innovation portfolio, so he writes and speaks on assessment and credentials for the university leaders who hold decisions like this one.
.svg)
