Co-created by Meriem Laifa, Samantha Ahern, and Asma Hadyaoui
(As part of the Knowledge Equity & Open Scholarship Week co-creation experiment.)
Introduction
The rapid spread of generative AI has changed how students, educators, and institutions access information, produce content, and approach learning. When GPT-based tools became widely available through simple chat interfaces, they appeared to open new possibilities for ideation, summarisation, feedback, and knowledge access.
However, this apparent openness has also exposed deeper inequalities. Access to AI tools is uneven across institutions, regions, languages, and socio-economic contexts. Some learners and educators can use advanced paid tools, stable infrastructure, and institutional support. Others rely on limited free versions, unstable connectivity, or little guidance on how to use these technologies critically and responsibly.
These differences matter because AI access is not neutral. The quality of the tools available, the ability to evaluate outputs, the language in which users work, and the level of institutional support all shape who benefits from AI and who is left behind. At the same time, debates around copyright, assessment integrity, data use, reliability, and the environmental cost of computing continue to complicate the role of AI in education.
This blog post brings together perspectives from Algeria, Tunisia, and the United Kingdom to explore a shared question: how can openness in the age of AI support equity rather than reproduce existing inequalities?
Case study: Meriem
Working as an associate professor for computer science at Bordj Bou Arreridj University, and being the manager of its AI House, has placed me at the intersection of two questions this blog post is mainly about: ‘What does open access to AI literacy look like in practice, and who does it actually reach?‘
Algeria’s AI House initiative, rolled out across universities nationwide, was conceived as an answer to a real problem: how do we ensure that AI literacy reaches everyone, not just those who seek it out individually? Each AI House is meant to serve as a free campus hub for workshops, training, and experimentation, a deliberate attempt to democratise access to knowledge about these tools. As someone who helps run one, I believe in the intent behind this initiative. But intent and impact are not the same thing.
In practice, the initiative runs into a familiar wall: availability does not automatically produce meaningful engagement. Many students and staff remain unaware of what the AI House offers. Workshops take place, but what is learned there rarely travels into the classroom without deliberate effort. The tools introduced are predominantly in English, which adds a quiet but real layer of friction for students and teachers alike who are mostly thinking in Arabic or French. And across all of this, the deeper challenge persists: access to a workshop, like access to an open resource, does not on its own build the critical capacity to engage with AI thoughtfully. It is worth noting that our institution provides no dedicated AI tools for staff or students. What it does offer is Google Workspace, and while some colleagues and students have told me they use its AI features, this is far from universal, and it is a long way from structured provision. Meanwhile, students who can afford premium tools privately are working with significantly more capable systems. The institution may feel it has addressed the access question; the reality on the ground is more uneven than that.
This is the gap that open education, at its best, should be designed to close. It is also the gap I have been trying to address in my own teaching. In Algerian universities, classes of 60 to 300 students with limited support staff are the norm. Individualised, in-room guidance is largely unavailable to us. But rather than treating this as a deficit, I have come to see it as a reason to think more carefully about what open, responsible AI integration in the classroom could look like. If students are going to use these tools regardless, and if there are not enough of us in the room to guide every interaction, then the pedagogical design itself must carry some of that weight.
The framework I am currently experimenting with in my courses requires students to use AI openly, but not passively. In each class, they must explain the outputs the tool produces and reflect critically on its limitations. The aim is to make AI a starting point for thinking, not a substitute for it. It is early, and I cannot yet report outcomes with confidence. But the premise is rooted in an open education principle: that equitable participation requires not just access to tools, but structured opportunities to develop the judgment to use them well.
What Algeria’s AI House initiative is still working to become, and what I am working towards in my own classroom, is a form of openness that goes beyond provision. Not just a space where AI tools are introduced, but one where the habits of critical, ethical engagement with those tools are genuinely cultivated. In a context of large classes, limited support, and uneven infrastructure, it is both harder and more urgent than anywhere else.
Case study: Asma
Working in higher education in Tunisia has made one point increasingly clear to me: openness and equity are not the same thing. Making digital tools, AI platforms, or educational resources available does not mean that everyone can benefit from them in the same way.
This became especially visible during the development of the E-Mallette Pédagogique initiative at the University of Sousse, a SPOC-based train-the-trainer programme designed to support university educators in digital pedagogy, open education, and AI literacy. The initiative showed that the real challenge is not only access to open educational resources or AI tools, but the ability to use them critically, ethically, and meaningfully in teaching practice.
During training activities, differences quickly appeared. Some educators were already confident experimenting with AI-generated content, refining prompts, adapting resources, and questioning the reliability of outputs. Others were discovering these tools for the first time and expressed uncertainty about how to use them, how far to trust them, or how to integrate them into their courses without weakening pedagogical quality.
In multilingual contexts such as Tunisia, these challenges become even more complex. Translation quality, limited Arabic educational content, cultural adaptation, and unequal access to advanced AI tools all shape who benefits from openness and who remains at the margins of it.
This experience challenged a common assumption in open education: that availability naturally leads to inclusion. It does not. Open access may increase opportunities, but without guidance, support, and capacity building, it can also reinforce existing inequalities. Those who already have stronger digital skills, better infrastructure, and more confidence with AI move further ahead, while others struggle to turn access into meaningful learning or teaching practice.
For me, the key lesson is simple: openness creates possibilities, but equity requires design. It requires structured educator training, ethical guidance, contextual adaptation, and learning environments that help both teachers and students engage critically with AI rather than passively consume its outputs.
In the age of AI, openness should therefore mean more than access to tools and resources. It should also mean the capacity to use them responsibly, critically, and meaningfully across diverse educational realities.
Case study: Samantha
My department delivers a combination of formal credited modules that are part of postgraduate taught programmes and workshops for postgraduate research students and researchers across our institution. UCL has a three-level system regarding the use of generative AI tools in assessment; no use, use in an assistive role, and integral. What is permitted varies assessment-by-assessment. These were introduced as part of the initial response to the tools and to alleviate concerns about assessment security.
There has been some centrally organised guidance and training for academic staff around these tools, however, levels of understanding and capability vary greatly within and across teams.
Initially, Microsoft CoPilot chat was only available to staff, so we felt comfortable experimenting with regards to whether or not it could be used for coursework completion. As at that time the responses were mediocre, we were happy to allow students to use these tools in an assistive way. Partway through that academic year, Microsoft CoPilot chat became available to students as well as staff. Currently, we do not have any other generative tools available at an institutional level. We have continued to permit use as these capabilities are increasingly appearing in development environments that our students are and will be interacting with. Currently, we have not yet seen any notable change in student grades with regards to the coursework component.
As students on our courses come from a range of geographies and socio-economic backgrounds, the nature and versions of tools they are able to access varies greatly. Some have access to the latest cutting-edge models, others only the free-to-access versions. We specify that only free-to-access versions can be used on the assessment, however, we have no way of monitoring this. Often in timetabled sessions, we see students using premium model versions, these can potentially outperform other versions. How you interact with the tools can greatly impact their performance. Within our modules we do not teach students how to interact with these tools, instead focusing on traditional approaches to code development. An increasing area of concern is the drop off in interaction between staff and students. A growing concern is the declining interaction between students and research software engineering professionals, as students increasingly seek guidance from AI tools rather than from the staff present to support their learning process.
Converging themes
Despite the differences between our institutional and geographical contexts, several common themes emerged across our experiences. In each case, openness created possibilities, but not necessarily equitable participation or outcomes.
Access to AI tools and open educational resources remains uneven across institutions, regions, languages, and socio-economic contexts. Yet the challenge extends beyond infrastructure alone. The ability to critically evaluate AI-generated content, formulate effective interactions with AI systems, adapt resources to local realities, and integrate these technologies into meaningful learning practices also varies considerably.
Another recurring theme concerns the growing tension between accessibility and guidance. As AI tools become increasingly available, educators and students often rely on them without sufficient pedagogical, ethical, or critical support. This creates new forms of dependency, uncertainty, and inequality, particularly when institutional strategies and training opportunities remain fragmented.
Across our different contexts, one shared conclusion became clear: openness in the age of AI cannot be reduced to access alone. Equity requires structured support, AI literacy, contextual adaptation, and opportunities for critical engagement. Without these conditions, openness risks reinforcing existing disparities rather than reducing them.
Call to Action
As AI technologies continue to reshape education, the conversation around openness must evolve beyond access alone. Providing tools and resources is no longer sufficient if educators and learners are not equally prepared to use them critically, ethically, and effectively.
Higher education institutions, policymakers, and open education communities must therefore invest not only in technological access, but also in AI literacy, educator capacity building, multilingual inclusion, ethical guidance, and sustainable support structures. Open educational practices should help learners develop critical engagement with AI rather than passive dependence on automated systems.
The future of open education will not be determined solely by how widely technologies are distributed, but by how equitably people are able to participate in, question, adapt, and benefit from them.
In the age of AI, openness without support risks reinforcing inequality. Openness designed with equity in mind can become a powerful mechanism for more inclusive and meaningful educational futures.