Colin Lowry — Generative AI Policy Framework for Higher Education in Ireland Colin Lowry comes from Ireland, where he is Senior Manager for Teaching and Learning Enhancement and Digital Innovation at Ireland’s Higher Education Authority. The Authority is the development body for higher education and research in Ireland. They have recently developed a generative AI policy framework, which he is going to share with us. Thank you. Colin Lowry: Thank you very much for that kind introduction. It is a real honour to be here, and thank you to EDEN for the invitation as well. I also want to thank all of you for your presentations. I’ve learned a great deal from you over the last couple of days, which I’ll certainly take back to Ireland and into my work. I was going to start with a perception that we had, but Anne stole my thunder. When we first set out on our work around generative AI readiness for the higher education system in Ireland, there was a perception that students were looking for access to tools, shortcuts, and so on. Instead, what we found was that they were looking for clarity and fairness. That came through very strongly in some of the work I’m going to share with you. For some context, the Higher Education Authority in Ireland is the statutory funding, planning, and regulatory body for higher education and research. We sit, in a sense, as middleware between the publicly funded institutions and our government department responsible for further and higher education. Our sister agencies include, for example, the quality assurance agency. We have about 279,000 students in public higher education in Ireland, across 18 publicly funded institutions that receive core public funding. We are quite a small system, and that gives us a unique advantage in terms of a certain degree of agility, particularly when it comes to this type of work. My unit is called the National Forum for the Enhancement of Teaching and Learning in Higher Education. Originally, we sat outside the Higher Education Authority, but since 2022 we have been established on a permanent and sustainable basis as a key function of the Authority. Our mandate is to lead the enhancement of teaching and learning in partnership — and partnership is the key word — with students, staff, and institutional leaders. We do not work by directive. The National Forum is very much about convening and collaboration. We do not instruct institutions on how to teach or assess. Academic freedom and institutional autonomy are, of course, significant parts of our system. Funding and System Development If we start with funding, money shapes a lot of behaviour — “follow the money,” as they say. Our mechanism for funding teaching and learning enhancement is the Strategic Alignment of Teaching and Learning Enhancement fund. We have been fortunate in Ireland to invest €28.8 million since 2022, providing multi-annual stability so institutions can experiment, scale what works, and share their work. We have a number of thematic priorities in this area, including digital transformation, academic integrity, and education for sustainable development. As you can imagine, AI cuts across all of these. One of the challenges with enhancement or innovation funding is the development of pilots. The graveyard of innovation is often full of well-funded pilots that died when their funding ended. A key mechanism we have put in place to address this is a condition attached to our funding: outputs must be openly licensed and shared so that others can learn from them and build upon them. To support this, we developed a central platform within the Higher Education Authority called the National Resource Hub. The person responsible for that work is here in the audience and led that piece of work. Where appropriate, funded outputs are required to be uploaded to the National Resource Hub so they can be shared. They are made available under open licences, meaning anyone can access them. Another area in which we have been supporting readiness across the sector is through centrally supported short courses, developed by the sector, for the sector. These are centrally hosted through a Moodle platform that we operate. We have offerings ranging from approximately 25-hour short courses down to two- or three-hour courses. Again, we use a commons-based, open-licensing approach. Institutions can not only participate in the courses centrally through the platform; they can also take the courseware away, embed it within their own virtual learning environments, adapt it, and remix it for their own contexts. That approach has been quite successful. One suite of short courses that we funded has seen approximately 950,000 enrolments globally. Developing a Generative AI Policy Framework Funding and courses, of course, are the easy part. Policy is where systems can sometimes overreach. We seconded a person from the higher education sector to lead this work over a period of 18 months and develop a generative AI policy framework that institutions could use. The reason we chose somebody from within the sector rather than a consultant was that we wanted somebody who would have to live with the results — or the repercussions — of what they developed. That person was Professor James O’Sullivan, who is also here in the audience and was part of yesterday’s panel. To give you a sense of how the policy development work took place, it began with sectoral engagement. We reached out to individuals and groups to understand the shape of the landscape during 2023 and 2024. We also conducted a desk review of Ireland’s national AI strategy, the EU AI Act, UNESCO competency frameworks, and the work already taking place within institutions. We did not want to duplicate what already existed. We wanted to bring it together. That engagement with stakeholder groups was subsequently broadened through advisory groups and similar structures. Ten Initial Considerations The first output, produced within roughly the first six months, was an attempt to establish a shared language through ten considerations. Some of the issues that came through strongly in the initial sectoral engagement included academic integrity, allowable uses of AI, sustainability, and related questions. At a stage when institutions were developing their own policies and guidelines, this provided a shared language. What we found after publishing this initial, relatively light-touch piece was that institutions began publishing their own policies within two, three, or four weeks. No single institution wanted to go first. But once one went, they all went. That demonstrated the power of our role in convening and supporting the system. That initial work has now been superseded by the Generative AI Policy Framework. National Case Studies Database As part of the 18-month development cycle, we also established a national case studies database. We asked educators, people supporting teaching and learning, and those working in senior management to send us case studies showing how they were dealing with generative AI in their own contexts. We then shared those nationally. The result is a public database containing a range of case studies, and we continue to collect submissions. This helps us understand the work taking place across the sector. It also enables people within institutions to identify useful points of connection with work happening elsewhere. It has proven very useful. Consultation and Sectoral Perspectives For the consultation phase of the policy framework itself, we convened a series of focus groups involving 76 stakeholders across a wide range of roles. These included not only academics, but academic developers, technologists, librarians, representatives from our National AI Advisory Council, our quality agency, infrastructure providers, industry, student representative groups, and others. The groups explored approximately ten key themes that had emerged through the earlier work. We published what we heard openly and candidly in our Sectoral Perspectives Report last September. At that stage, the report was particularly useful to the sector. It captured much of what we had heard, as well as the implications for policy. We were then able to take those findings forward into the development of the framework. I want to give you a sense of some of the voices that came through. One participant said: “I’m not sure if what I’m doing is encouraged, ignored, or discouraged. No one’s talking about it.” That represented what we called the clarity gap. It came from both students and staff. There were not enough conversations taking place. There was not enough openness about how people were using AI. There was a certain shyness around it. There are, of course, highly critical voices around AI, as well as uncritical and highly enthusiastic voices. In that environment, people were sometimes hesitant to discuss their actual use of AI. Another comment was: “They’re submitting perfect-looking assignments, but when you talk to them, the thinking just isn’t there.” That spoke to concerns around skills degradation. Another participant said: “We don’t want students who can prompt well. We want students who can think about what the prompt is doing.” That reflected anxiety about the possibility that AI could complicate people’s ability to think critically and creatively. Emerging Policy Themes We synthesised the Sectoral Perspectives Report and the recommendations for policy development. Among the strongest emerging themes were: the need for national and system-level coordination; a re-examination of educational purpose and authorship; inclusion that is designed rather than assumed; and assessment reform that moves away from detection and towards authentic, process-based approaches. For me, having worked in teaching and learning for the past ten years or so, that was very validating. Much of this is work we were already doing before generative AI. Generative AI now gives us an opportunity to leverage that work and move it forward. The phrase “away from detection” became something of a defining position. It received some pushback, and it is not something we have entirely figured out. But the central idea is to govern according to educational purpose rather than making everything about individual tools. If we do that, the framework becomes transferable across different contexts and helps us build a stronger system more broadly. The Generative AI Policy Framework Fast-forward to the framework itself. In December, we brought all of this work together and produced the Generative AI Policy Framework for Higher Education Teaching and Learning. It consists of two main documents. The first is the policy framework itself, containing five principles, which I’ll touch on in a moment. The second concerns the principles for ethical adoption and provides greater detail. We also developed a series of annexes alongside the framework to address the operational work and supporting evidence. Throughout the development process, James repeatedly made the point that the framework had to be useful. It could not simply be a policy document that sat on a shelf. We therefore created a series of annexes covering areas such as: EU AI Act compliance; AI literacy and training; the evidence underpinning the framework; assessment practices; vendor and procurement governance; and role-by-role responsibilities to support institutional structures. Separating these different layers also makes the framework easier to revise. It is intended to be a living framework. We want to be able to update it, and separating some of the individual components makes version control much easier. Five Principles The framework is based around five principles: Academic integrity and accountability Equity and inclusion Critical engagement and human oversight Privacy and data governance Sustainable pedagogy These are five principles that do not necessarily change every time a new model is released by Claude, OpenAI, or anybody else. We therefore do not need to revise the highest level of the framework every time the technology changes. Importantly, we are approaching the issue from a values perspective. Beneath each principle, the second document breaks the principle into more detailed components. For example, under critical engagement and AI literacy, there are eight components, ranging from embedding AI literacy as a core competency through to the governance and evaluation of that literacy. Each principle also comes with a set of recommendations that institutions can consider when developing their own policies. For critical engagement, for example, the recommendations include treating AI literacy as a core graduate attribute, scaffolding it across programmes, and resourcing educators so that they can teach it credibly. And this, in part, is our answer to the detection question: assess authentically. That might involve portfolios, case-based tasks, collaborative projects, and other forms of assessment that foreground the learning process. Governance and Ongoing Review Finally, our role comes from an enhancement perspective. So how do we govern the framework? We have not created any new audit apparatus. We have not introduced new reporting requirements for institutions. Instead, we monitor implementation through our existing dialogue processes. We already have a number of mechanisms through which we engage with institutions. Through those, we learn how institutions are adopting the principles and recommendations. Based on what we hear back, we will revise the framework. We will also consider whether other mechanisms are needed if we identify gaps in the system, whether through conditions attached to funding or other measures. That is a quick tour of how we have worked towards a level of AI readiness through a system-level approach in Ireland. Thank you very much.