Using generative AI with a more reflective mindset

Wayne was interviewed by the SELFIE team as part of a discussion on how SELFIE tools can help schools and teachers think how to approach generative AI. Here's a summary of his insights.

Discussions on Generative AI (GenAI) in schools often jump straight to ‘how do we use it?’ However, Wayne Holmes, Professor of Critical Studies of Artificial Intelligence and Education at University College London and UNESCO Chair in the Ethics of Artificial Intelligence and Education, encourages us to approach GenAI with a more reflective mindset.

Holmes urges us to take a step back and think about whether, when, and on what terms it should be used.

GenAI introduces new challenges for teaching and learning

Holmes sees the key difference between GenAI and earlier waves of educational technology is its ‘human-like’ presence. Previous tools, from calculators to interactive whiteboards, were instruments. GenAI, by contrast, is designed to feel conversational, intuitive, confident, even human. This can mislead users, especially new users, into over-trusting the output.

The educational challenge is not only ‘spotting misinformation’ but understanding why GenAI output can look plausible even when it is wrong, incomplete, or inconsistent. In practice, he argues that students need support to question GenAI responses, verify claims and recognise when a tool is ‘simulating’ certainty rather than providing actual knowledge.

He also stresses that conversations about GenAI, rather than focusing on prompts and productivity, should prioritise the systems and trade-offs behind these tools, such as intellectual property, children’s rights, data protection and environmental impacts.

Holmes argues that we do not yet have enough evidence to confidently label many GenAI classroom uses as effective, not least because effectiveness depends on what education is trying to achieve, such as deep learning, learners’ wellbeing, or long-term independence. For Holmes, education’s role is not only helping learners acquire knowledge but also to enable them to become autonomous individuals who can think for themselves, relate to others and have their own interests and agency.

Any discussion of ‘appropriate use’ must also consider constraints such as risk classification and rules on minors’ data. He notes that safer technical approaches are being explored, such as using GenAI models trained on approved curricular materials inside ‘walled garden’ environments, but argues that, even then, dependence on foundational GenAI models still raises concerns.

Holmes is most wary of GenAI use that encourages over-reliance, particularly in core learning tasks, because of its long-term effects on habits and confidence.

The use of GenAI in teaching can reshape relationships and learner agency

Rather than treating GenAI as simply another classroom tool, Holmes places the discussion in a broader educational context. Quoting philosopher Gert Biesta, he points to three purposes of education: 

1. Qualification (knowledge and credentials)
2. Socialisation (learning to live with others)
3. Individuation (becoming oneself)

His concern is that GenAI discussions in education often emphasise efficiency, outputs, and grades, neglecting social development and personal growth. While GenAI may simplify some tasks, it may also disrupt classroom dynamics, affect student agency, and influence how learners value their own thinking and ideas.

Teachers and students need time and support to understand what these tools do, where they can mislead, and how they may change classroom norms. Educators also need stronger role models, and space for professional inquiry.

Equity in the educational use of GenAI involves more than access

Holmes suggests that the notions of ‘responsible’ and ‘equitable’ are broader than the notion of access. Equity is not only about who can use the tool, but also about who continues to benefit from human support and high-quality teaching. He warns that generative AI could widen existing inequalities if better-resourced communities maintain more human-intensive education, while others are increasingly directed towards automated substitutes.

Holmes recommends slowing down and strengthening research and guidance, and practical frameworks to help schools make informed decisions including when not to use GenAI.

Treat GenAI as a literacy challenge

Holmes argues that the most responsible move is to strengthen critical thinking and media literacy so teachers and learners can evaluate GenAI claims, verify information, recognise persuasive ‘human-like’ output, and understand the possible compromises behind adoption.

The priority should not be mastering prompts but building the judgment to decide when GenAI contributes to learning, and when it does not.

Our take: SELFIE tools for school-level reflection

The SELFIE tools can support school and teachers by offering structured reflection on GenAI readiness, goals, safeguards, and equity. This can help shift the conversation towards well-informed, context-based decisions. The SELFIE tools can be valuable in helping schools and teachers consider purpose, capacity building, resources, risks, and potential benefits, so that any adoption strengthens learning rather than diminishing it.

About Wayne Holmes

Wayne Holmes is a Professor of Critical Studies of Artificial Intelligence and Education at the UCL Knowledge Lab, which is part of the UCL Institute of Education at University College London.

He also holds a UNESCO Chair in the Ethics of Artificial Intelligence and Education at the International Research Centre on Artificial Intelligence at the Jožef Stefan Institute in Slovenia and is an adjunct associate professor at the University of Nova Gorica in Slovenia, as well as a foreign expert at Beijing Normal University Zhuhai in China.

Throughout his career, Wayne has been involved in education, providing valuable insights into its relationship with AI, as well as the ethical, human rights and social justice issues that arise from it.

Find Wayne on LinkedIn and read his university college of London profile