‘Children are wired to trust authority. If AI speaks with confidence, they will believe it’ (Schleicher, 2026)
For many years, technology has been encouraged as a tool for learning and acquiring knowledge, both in classrooms and beyond them. It has supported a gradual shift away from teaching based primarily on the transmission of knowledge toward constructivist approaches, where teachers take on a more facilitative role and students are encouraged to question, collaborate, solve problems, and learn how to learn. Most of these tools provide learners with new sources of information and new ways of communicating and creating, while leaving the processes of thinking, questioning, and constructing knowledge largely to the learner.
However, generative AI is different. It can not only provide access to information, but it can also produce explanations, formulate arguments, solve problems and generate finished responses on the learner’s behalf. This creates a tension with the constructivist principles that have shaped education over recent decades. If students increasingly turn to AI for answers before trying to think through a problem themselves, their critical thinking skills are at risk, emphasising the need to understand AI’s true impact on learning.
Views on AI in education are mixed. International organisations such as UNESCO advocate a human-centred, rights-based approach to AI in education, recognising its potential while highlighting concerns around equity, privacy, safety and governance. The OECD’s Digital Education Outlook 2026 argues that generative AI can support learning when used with a strong pedagogical purpose. In this case, it can support collaborative learning and strengthen students’ argumentation skills. However, the OECD also claim that using general-purpose AI can improve students’ performance on a task without necessarily producing genuine learning gains. This is because when students use AI for cognitive tasks, they risk disengagement and what Noorbehbahani and Oyibo (2025) describe as metacognitive laziness.
Problem 1: AI systems present information fluently and confidently, even when that information is inaccurate or biased. Young people, who are still developing their identities and their ability to evaluate information, may be particularly vulnerable to accepting apparently authoritative answers without questioning them. If students routinely turn to AI before attempting to solve a problem themselves, they may lose their ability to question, analyse, evaluate, and form independent judgements. Moreover, AI models are increasingly being used not only for information and schoolwork but also for advice and emotional support. This raises questions about dependency and, perhaps more importantly, about authority. AI systems are not neutral sources of knowledge. Their responses are shaped by their training data, design and underlying models, and they can reproduce biases or reinforce existing assumptions.
Problem 2: Teachers are not necessarily prepared for these risks. At present, much of the discussion surrounding education focuses on students using AI to complete homework or cheat in assessments. These are important concerns, but they are only part of the picture. The deeper question is what happens when students increasingly use AI to think, when education is supposed to help them learn to do it themselves. The danger, therefore, is not simply that students might copy their homework. It is that they may gradually become less inclined to think for themselves.
Solution: Teacher training is fundamental, but it should not focus only on using generative AI to make teaching more efficient. Instead, it should highlight pedagogical strategies that turn AI into a tool for critical thinking, encouraging discussion and evaluation. To do this, however, we need more research on how young people actually use generative AI, both inside and outside the classroom.
Question: What is your experience? Most of us have some contact with generative AI in school education, whether through our own children, friends’ children, nieces, or nephews. How are young people using AI? What patterns or concerns have you observed?
Post by Dr Lesley Fearn
