Futures in Context
Rose Luckin
Professor, AI and Education Thought Leader, Author and Speaker
Rose Luckin is Professor of Learner Centred Design at University College London. She is founder of EDUCATE Ventures Research Ltd and has worked in AI and education for more than thirty years.
AI & EDUCATION — CHAPTER 1
General overview
I have been working in AI and education for over thirty years. For most of that time, AI in education was a niche academic pursuit. What changed was ChatGPT: freely available, easy to use, and generating humanlike behavior that anyone could experience. AI was no longer an abstraction but a conversation partner that could help with homework, draft emails, and explain concepts.
This accessibility has revealed something important about our education systems: We have undervalued human intelligence for too long. We have built assessment systems that reward precisely what machines do well (pattern matching, information recall, standardized responses) while undervaluing uniquely human capabilities. If AI can pass exams and write essays, what have we actually been measuring?
The global response has been fragmented. Europe emphasizes ethical frameworks, the U.S. focuses on data privacy, while China has mandated nationwide AI education with detailed guidelines. Teachers find themselves navigating distinctions between traditional AI systems, generative AI tools, and proprietary educational solutions, often without clear frameworks for quality or safety.
This accessibility has revealed something important about our education systems: We have undervalued human intelligence for too long.
The essential conditions for preparing the next generation are threefold. First, AI literacy: the ability to work effectively with AI systems, recognise their limitations, and understand their economics. Second, learning mastery: knowing how to learn, including the metacognitive skills to plan, monitor, and regulate our own thinking. Third, knowledge mastery: understanding what knowledge is, that it is constructed and contextual, that it evolves and is always tentative. This epistemic cognition is beyond the capacity of AI.
New forms of intelligence, learning, and schooling
The arrival of AI forces us to reconsider what intelligence actually is. I propose that human intelligence is interwoven, consisting of seven elements: knowledge about the world (academic intelligence), knowledge about what knowledge is (meta-knowing intelligence), social interaction capabilities (social intelligence), and knowledge about our cognition, context, emotions, and self-efficacy.
AI excels in knowledge about the world. It can learn faster and recall more accurately than humans within this domain. But AI cannot achieve human-level social interaction, has no awareness of subjective experience, and cannot develop metacognitive awareness. These meta-level capabilities are what set us apart, and they are precisely what our education systems have undervalued.
Here lies a profound risk. Human cognition is not static. Our memory is already changing due to the Google effect, our navigation abilities have atrophied because of satnavs, and our cognitive-reward mechanisms are shifting because of gamification. This process is accelerated by technology and will be magnified by AI. If we offload cognitive tasks to AI without careful thought, we risk diminishing the very capabilities we need to develop. We must adapt to change thoughtfully, offloading intelligent activity to AI carefully so that we maintain the integrity of our human intelligence.
Here is the paradox: AI tutors support an extraordinarily narrow repertoire of learning acts, primarily exposition, rehearsal, and tutorial dialogue. Research identifies at least nineteen distinct ways humans learn. AI tutors primarily support three. Reflection cultivates metacognitive skills. Collaborative work builds social intelligence. Problem-focused learning encounters knowledge in authentic contexts. These are not optional extras but fundamental to understanding that lasts and transfers. If we narrow learning to what AI can support, we are not preparing students for a world with AI. We are preparing them to be cognitively dependent on AI.
The capabilities education should strengthen most urgently include metacognitive skills (interpreting and regulating our own mental activity), social intelligence (the basis of individual thought and communal intelligence), epistemic cognition (understanding what good evidence looks like and how to make judgments), and accurate perceived self-efficacy. The goal is not for education to merely adapt to AI, but to become strategically more powerful because of it.
In a world where technology changes weekly, AI literacy becomes a lasting skill when we teach patterns rather than products. Understanding how machine learning works, why AI systems hallucinate, and how training data shapes outputs provides durable knowledge that helps people evaluate any AI system, including those that do not exist yet.
Limits and the future role of educators
The threats of AI in education include data-privacy risks from shadow AI (tools adopted informally outside IT oversight), algorithmic bias that can affect life trajectories when used for assessment or placement, and the risk of inappropriate cognitive offloading that diminishes rather than develops human capability.
This last risk deserves particular attention. We are at a critical juncture where AI is becoming increasingly capable at precisely the moment when humans need to become more sophisticated in their intelligence. If we allow AI to do our thinking, our remembering, our evaluating, and our creating, we risk a downward spiral: The less we exercise these capabilities, the weaker they become, and the more we depend on AI to compensate. We must resist the temptation to languish in outdated definitions of intelligence. Instead, we must learn to enjoy developing our intelligence, accepting that we will never be intelligent enough and that we must always keep learning.
If we narrow learning to what AI can support, we are not preparing students for a world with AI. We are preparing them to be cognitively dependent on AI.
The deepest threat is a dystopian future in which the disadvantaged receive AI tutoring in curriculum basics while the privileged receive richer, human-led education that develops the full breadth of their intelligence. We must guard against increased social immobility.
Educators must lead the global AI discussion because education is where we decide what kind of future we are building. If we ignore the need for education about AI, we fail to empower people to make key decisions about what AI should and should not do for society. The development of AI teaching assistants provides an opportunity for deepening teacher expertise rather than replacing it.
This is a call to action for the human race to become smarter. We have these technologies that are quite smart, and now we need to ensure that we continue to evolve our human intelligence to be ever smarter. Human intelligence is not a finished work: We are still evolving. I have two hopes: that we raise awareness about undervaluing human intelligence, and that we engage in finding better ways to develop this intelligence beyond the power and potential of AI.
