Renewing the Purpose of Education in an AI World: From Purpose to Practice
Part 2 of 3
In the first part of this series, I argued that artificial intelligence is forcing us back to an old question: what is education for? I proposed that its purpose is human flourishing, agency and the common good, pursued through four pillars – knowledge, agency, character and service. This second piece turns from purpose to practice. What would a school built on those pillars actually look like? It considers curriculum and co-curriculum, pedagogy, the relationship between teachers, learners and AI, digital citizenship, wellbeing and, finally, assessment – the fulcrum on which every reform turns.
Curriculum
A curriculum is never simply a body of knowledge to be covered or a sequence of qualifications to be completed. It shapes not only what learners know but how they think, who they become and how they engage with the world. That is why artificial intelligence matters so much. When information, explanation and fluent prose can be produced instantly, retrieval and reproduction are no longer the scarce capacities. What becomes scarce is the ability to ask good questions, to judge the quality of an answer, to recognize when something is wrong, one-sided, or manipulative, and to act wisely on what we know.
Knowledge remains important. We cannot evaluate what a machine tells us without secure understanding of our own. The historian who knows the period spots the fabricated source; the scientist who grasps the underlying principles notices when a confident answer is wrong. A curriculum for an AI world must therefore begin with strong disciplinary foundations, including an understanding of how AI works and where it fails. But it cannot end there. Knowledge becomes judgement only when learners use it: justifying an argument, or applying it to authentic problems, working with others, creating something new, and reflecting on the ethical implications of what they know.
A renewed curriculum should give greater prominence to areas of learning that have long been undervalued. Design and technology develop practical intelligence and the ability to turn ideas into action. Vocational education develops expertise that lives in the hands as well as the head: the diagnostic judgement of the engineer, the craft knowledge of the builder, the care of the nurse. The creative and performing arts cultivate imagination, interpretation and aesthetic judgement. Philosophy, ethics and religious education invite learners to explore questions of truth, meaning and responsibility. Oracy develops the capacity to listen carefully, explain clearly and engage in respectful dialogue. Physical education and sport build health, teamwork and resilience, and remind us that learning is embodied, not only intellectual.
The case for these areas is not that machines cannot create, write, compose or argue. Increasingly they can. It is that doing these things forms the person who does them. Learning to make a case aloud, to perform before an audience or to build something that works develops judgement, confidence and character in ways that no tool can supply on our behalf. Vocational education has a further claim. The skilled, physical and relational work of the engineer, the builder and the nurse remains far harder to automate than much white-collar work and is likely to grow in value as a result. It deserves genuine parity of esteem, not a place on the margins.
Time in school is finite, and every addition asks what gives. Not every student can study everything, nor should they, but every student should be able to combine practical, creative and academic learning rather than being sorted early into one route or another.
In practice, those choices are largely determined by what we assess and what we hold schools accountable for, a point I return to below. Nor are their effects neutral. Too often it is disadvantaged students, in schools under the greatest pressure, whose curriculum narrows first. A curriculum for flourishing must be an entitlement for all, not an enrichment for some.
Such a curriculum does not sacrifice academic rigour for breadth. It treats them as mutually reinforcing. Deep knowledge is what makes good judgement possible, and broad experience is what turns that knowledge into wisdom.
The co-curriculum
Many of the qualities that shape a flourishing life cannot be taught directly; they are formed through practice. Aristotle made the point more than two thousand years ago. We become just by doing just acts, and brave by doing brave acts (Aristotle, 2009, Book II). This is the deepest rationale for the co-curriculum. It is where young people practise being the people they hope to become.
John Dewey added an important qualification: not every experience is educational. Experience educates when it builds on what came before, opens possibilities for what comes next and is reflected upon (Dewey, 1938). A chess club, an orchestra or an expedition is valuable not simply because it happens, but because of the conditions under which it happens. Several features make co-curricular experience distinctively formative:
- It is chosen. Learners opt in and commit to something they care about. This is agency in its purest form.
- It has real stakes. Matches can be lost, performances can go awry and campaigns can fail. Resilience is built by meeting setbacks, not by avoiding them.
- It is collective. Young people work towards shared goals, learning to depend on others and to be depended upon.
- It creates different relationships, across year groups, with adults who coach and mentor rather than assess, and with communities beyond the school gates.
- It is embodied and present, in rehearsal rooms, on pitches and on hillsides, with other people in the same place at the same time.
The distinction between curriculum and co-curriculum is therefore best understood not as one of content but of character. Debating, drama, music and sport belong in the curriculum too. What sets the co-curriculum apart is that it is voluntary and self-directed, and that it carries consequences young people feel. Kurt Hahn, whose ideas shaped Gordonstoun, Outward Bound and the Duke of Edinburgh’s Award, built an educational philosophy on the belief that young people discover more in themselves when they are given genuine challenge and responsibility.
These experiences matter all the more in an AI world. As young people’s lives become increasingly mediated by screens and algorithms, the co-curriculum offers something scarce: sustained attention, physical presence, unscripted encounters with other people, and the productive friction of doing difficult things together. AI may help learners research, plan events or develop campaigns, but it should serve shared human endeavor rather than substitute for it.
The case for the co-curriculum should not rest chiefly on measurable outcomes. Evidence of impact on attainment varies considerably between activities (Education Endowment Foundation, n.d.), and justifying these experiences mainly as routes to better grades or careers concedes too much. They are not merely a means to a flourishing life; they are part of it.
Participation matters as much as provision. Take-up of extra-curricular activities is strongly patterned by family income (Social Mobility Commission, 2019). Cost, transport, caring responsibilities, part-time jobs and activities timetabled after school all exclude some young people. Pierre Bourdieu (1986) showed how the confidence, familiarity and tastes cultivated in privileged homes are converted into educational advantage. Schools cannot simply bridge this gap, but they can widen access to the experiences through which confidence and belonging grow. They can protect time within the school day, subsidize places, encourage participation and pay careful attention to who actually takes part and how their contributions are valued.
The UK Department for Education’s enrichment framework for schools and colleges, published in June 2026, is therefore a welcome signal (Department for Education, 2026). Its five areas – civic engagement; arts and culture; nature, outdoor and adventure; life and future skills; and sport and physical activity – align closely with the vision set out here. Yet now that Office for Standards in Education, Children’s Services and Skills, or Ofsted, considers enrichment as part of personal development, schools may feel pressure to evidence provision rather than nurture the growth of young people. Growth of that kind cannot be audited by logging activities and counting participants. The framework will succeed only if it is matched by the time, funding and staffing that rich provision demands.
A strong co-curriculum reminds us that education is measured not only by examination results but by the lives young people learn to lead and the people they become.
Pedagogy
Pedagogy should be consistent with the purposes it serves. If education aims to cultivate knowledge, agency, character and service, then learners must practice those capacities. We learn judgement by judging and collaboration by collaborating. At the same time, the activity that matters most is cognitive. Learners remember what they think about (Willingham, 2009). Busy classrooms are not always thinking classrooms, and learning outcomes must be intentional, scaffolded and developmentally coherent.
Foundations: guidance first
The approaches described below broaden rather than replace traditional teaching. Good teaching can begin in different ways. A teacher may introduce a new idea directly, or students may first explore it for themselves, as in flipped learning, before their findings are clarified and consolidated in class. What matters is not who presents information first, but that students are guided, and knowledge is made explicit and secure. Minimally guided discovery is ineffective for novices (Kirschner, Sweller and Clark, 2006), but structured exploration followed by clear consolidation can deepen understanding (Kapur, 2016). As expertise grows, the balance shifts. Guidance that helps beginners can constrain more expert learners (Kalyuga et al., 2003). This makes possible a progression from structured, scaffolded teaching towards more open inquiry, in which learners take increasing responsibility. The ultimate goal is for all learners to become self-directing and autonomous.
Structure matters not only for understanding and academic rigor but for equity too. Loosely guided inquiry, especially in the early stages, tends to favor learners who already have background knowledge and support at home.
Throughout, formative assessment and feedback remain among the most powerful tools teachers have (Black and Wiliam, 1998).
The challenge of cognitive offloading
Much learning depends on productive struggle – the effortful retrieval, reasoning and revision that Robert Bjork (1994) called “desirable difficulties”, and which the approaches described below are intended to foster. AI makes it easy to bypass that struggle, enabling students to produce the answer without the learning, and challenging teachers to adapt their methodology while still cultivating the skills and attributes already outlined.
The approaches that follow are not alternatives to good teaching but ways of extending it. Each engages young people with authentic problems and contemporary issues, and each develops, in different ways, the knowledge, agency, character and service on which this series rests.
Project-based learning
Project-based learning is already well established through the Higher Project Qualification (HPQ), usually taken at 14–16, and the Extended Project Qualification (EPQ), taken alongside A levels. It widens knowledge and deepens understanding through inquiry, research, analysis and reflection, enabling learners to apply their knowledge in meaningful contexts and to maintain ownership of their thinking. I have seen its benefits firsthand, having helped pilot and develop the EPQ when it was a very new qualification and championed its adoption in all the schools I have led.
The various models of project-based learning offered through the exam boards provide scope not only for individual work but for performances, practical outcomes, design solutions or exhibitions. The opportunity to offer a sequence of projects across all the Key Stages, from tightly scaffolded to genuinely independent, is one which schools already have at their disposal. The evidence offers a note of caution. Well-designed, well-supported programmes have shown positive effects (Krajcik et al., 2023), while poorly implemented ones have not (Menzies et al., 2016). Quality of design and teaching is decisive.
There is an obvious objection. AI can now produce a convincing project report, so how can we trust the validity of this kind of learning? There are a number of safeguards. The EPQ already insists on production logs, sustained supervision and a presentation that requires learners to explain and defend their thinking. Learners should also critique their use of AI, demonstrating intellectual honesty and clarity about where human judgement remains essential. When the focus shifts from written output to process and intellectual agency, projects become powerful learning tools, foregrounding the importance of ethical and transparent inquiry and preparing students for ways of working that more closely mirror those of universities and workplaces.
Design thinking and problem solving
Design thinking asks learners to investigate real problems, understand the needs of others, generate creative possibilities, prototype solutions, and evaluate their impact. The Nueva School in California, working with Stanford’s d.school and IDEO, has integrated it across its curriculum, aiming to help children “believe in themselves” as agents of positive change (Nueva School, n.d.). The evidence base remains largely descriptive, and Nueva is a selective school, so claims should be modest. But the approach is worth considering for problem-solving activities, particularly for its emphasis on imagination, empathy, collaboration, resilience and reflection. Its value lies in connecting knowledge to action, through a process widely used in industry and public services. Schools can adapt it to their own purposes, provided it remains grounded in disciplinary understanding and reflective practice rather than becoming a formula.
Service learning
Service learning connects academic study with meaningful contribution to others. In the curriculum it might include a science class investigating local air quality and presenting its findings to the council, or history students recording the oral histories of older residents. It gives the Service pillar a home in the classroom and teaches that knowledge carries responsibilities as well as rewards.
Socratic and dialogic learning
In a complex and divided world, the capacity to engage thoughtfully with disagreement grows ever more important. Socratic questioning, typically teacher-led, probes assumptions and cultivates intellectual humility. Dialogic teaching, as developed by Robin Alexander, is more collective. Teachers and learners build understanding together through talk that is reciprocal, supportive and cumulative (Alexander, 2020). It has supportive trial evidence (Jay et al., 2017). AI tutors can now pose Socratic questions, but a machine cannot provide a community of inquiry in the same way as a classroom. The experience of reasoning with others whose views differ, and of learning that respectful disagreement is part of democratic life, should remain central to education’s mission.
Reflective practice
Reflection and metacognition – the capacity to plan, monitor and evaluate one’s own learning – have among the strongest evidence bases of any approach (Education Endowment Foundation, 2018). They need careful inter-weaving with classroom practice, however. Reflection is most effective when it is specific and linked to action, not simply a matter of filling in journals for their own sake.
In an AI-rich world, learners should also consider when and how they used intelligent systems, what those systems contributed, where they were wrong and where human judgement proved essential. That prepares the ground for assessment that values the process of learning as well as its products.
Teachers, learners and AI
Artificial intelligence is unlike previous educational technologies. Interactive whiteboards and virtual learning environments changed how teaching was delivered; AI performs tasks that resemble teaching itself. It can explain, question, give feedback and adapt to individual learners, instantly and at scale. Yet the early evidence registers the need for caution. Carefully designed tutors, especially those working alongside human teachers, have shown promising results, but large-scale classroom trials have been less convincing, and unrestricted use can actively harm learning. Students who relied on an unguarded chatbot during practice later performed worse than those who never had access (Bastani et al., 2025). What matters, in other words, is not the technology itself but how it is designed and directed. That is precisely why the relationship between teachers and AI must be thought through carefully rather than allowed simply to evolve.
It is tempting to describe that relationship as a partnership. But partners share responsibility, and AI cannot. It is better understood as a powerful collaborator working under human direction, and answerable to people who remain accountable for what it does. The central question is therefore not what AI can do, which changes almost monthly, but which responsibilities must remain human, and why. Four stand out:
- Accountability. Responsibility requires an agent who can explain a decision, answer for it, be challenged and learn from mistakes. AI can generate a recommendation, but it cannot be answerable for it.
- Relationship. Young people work harder for teachers who know them, believe in them and notice when something is wrong. Being known by another person is not incidental to education but part of its substance.
- Example. Character is formed partly by imitation. Young people learn what judgement, curiosity, integrity and kindness look like by watching adults exercise them.
- Legitimacy. Decisions about what young people should learn and who they might become are moral and democratic questions, to be settled by people accountable to communities, not by the design choices embedded in a system.
AI can enrich all of this. It can generate practice questions, draft feedback for a teacher to refine, identify patterns in learners’ work and make learning more accessible for students with special educational needs. In a well-designed workflow, AI might produce initial feedback on an essay, the teacher reviews and personalizes it, and the student responds. Each contributes what it does best, and the teacher remains in charge of the judgement.
We should, however, be realistic. Technology has repeatedly been promised to reduce teacher workload, and the time saved has often been absorbed by new demands. Some apparently routine tasks, such as marking, are also how teachers come to understand their students. And there are wider risks, from emotional dependency on AI companions to the influence of large technology companies in education and the temptation to replace teachers rather than support them, which I consider in Part 3.
Governments and international bodies have already offered guidance on AI in education, much of it focused rightly on safety, data protection and risk (UNESCO, 2023; Department for Education, 2025; 2026). The principles below complement that guidance by grounding decisions in educational purpose. Each comes with a simple test that schools can apply.
- Purposeful. Educational goals should determine how AI is used, not the other way round. Can we name the learning this use of AI serves, and explain why it serves it better than the alternatives?
- Accountable. AI may inform decisions, but responsibility for educational judgement and the wellbeing of learners must always rest with identifiable people. If this goes wrong, who is answerable, and do they understand what the system is doing?
- Transparent. Learners and teachers should know when AI has been used, what it contributed and its limitations, and should treat its outputs as material for inquiry rather than authority. Would we be comfortable explaining this use to the students and their parents?
- Fair. AI should widen opportunity rather than entrench advantage and should protect privacy and dignity. Who benefits, who might be disadvantaged, and whose data is being used?
AI may prove one of the most powerful resources education has ever seen. Precisely because of that power, it must be directed by human values, exercised through professional judgement and held to account by people.
Digital citizenship
Today’s young people will spend their lives in environments shaped by algorithms they did not design and cannot see. These systems influence what they read, whom they meet, what they believe and how they spend their attention. Equipping young people to understand and shape these systems is a precondition of autonomy and democratic participation.
Three related but distinct capacities are involved. Digital literacy is the understanding of how digital systems work – how data is collected, how algorithms rank and recommend, and how AI systems are trained and why they make mistakes. Digital wellbeing is the capacity to manage attention, relationships and identity online in ways that support flourishing. Digital citizenship goes further. It asks who governs digital platforms, what rights citizens should have over their data, how algorithmic decisions can be held to account, and how young people can take part in shaping the rules. Schools need to cultivate all three, but the third is the most often neglected.
Much is already changing. Following the Curriculum and Assessment Review, the government has committed to rebalancing the computing curriculum, introducing a new Computing GCSE, making citizenship compulsory with media literacy at its heart, and embedding digital, media and AI literacy across subjects (Department for Education, 2025a; 2025b). These are welcome steps. But content scattered across many subjects risks fragmentation, and a skills-and-safety framing does not by itself address the deeper questions of autonomy, ethics and power.
The government has also announced that under-16s will be barred from social media from 2027, and that AI companion chatbots offering romantic or intimate interactions will be restricted for under-18s (HM Government, 2026). Whatever one’s view of these measures, they recognize that education alone cannot protect young people. It is unreasonable to ask a 13-year-old to resist, unaided, systems designed by thousands of engineers to capture attention. Regulation limits exposure to harm; education builds the understanding and judgement young people need when they do engage. And if young people are to enter social media at sixteen, digital education cannot end there.
Digital citizenship should range across the technologies that shape young people’s lives, not AI alone. These include social platforms, virtual and augmented reality, content creation and robotics. Each raises its own questions about identity and presence in immersive spaces, about authorship, consent and truth in what young people make and share, and each offers opportunities to create as well as consume.
I therefore propose a coherent digital citizenship programme for all learners from five to eighteen, whose parts build on one another rather than being scattered across subjects:
- In primary school, it would begin with age-appropriate habits: paying attention, kindness online and questioning what we see, working closely with parents, since digital habits form largely at home.
- In secondary school, learners would develop a foundational understanding of how digital systems and AI work, including the basics of programming, alongside privacy, misinformation, data use and algorithmic influence, and the ethical questions these raise.
- In the sixth form, all learners would continue with a core strand preparing them for full participation online: entering social media wisely, understanding their rights over their data and engaging critically with AI. For those who wish to go further, an Advanced Digital Skills Pathway would offer deeper study of AI, software development, robotics, entrepreneurship and human-centered design, actively designed to counter the persistent under-representation of girls in computing.
Throughout, the different elements of the programme should be taught through applied experience: investigating real digital systems, debating ethical dilemmas, undertaking design projects and reflecting on one’s own digital life.
The programme also has a civic purpose. Young people should understand that the rules governing digital life are human choices that can be debated and changed, and that they have a voice in changing them. This will not happen without time on the timetable and teachers equipped to teach it, and specialist computing teachers are already in short supply.
Understood in this way, digital citizenship is not an optional extra. It is how education prepares young people to live freely, wisely and responsibly in a world shaped by intelligent technologies – and to help decide what kind of world that will be.
Wellbeing, belonging and human flourishing
Human flourishing rests on wellbeing and belonging. But what we mean by wellbeing matters. Psychologists distinguish between hedonic wellbeing, which is feeling good, and eudaimonic wellbeing, which is living well: having purpose, growing, forming meaningful relationships and exercising one’s capacities (Ryan and Deci, 2001). The vision set out in this series rests firmly on the second. Flourishing is not the absence of difficulty. It includes challenge, effort, setback and the satisfaction of achieving something hard.
This distinction answers a serious critique. Some argue that school wellbeing agendas risk becoming “therapeutic,” treating young people as fragile and lowering expectations (Ecclestone and Hayes, 2008). That danger is real. But the answer is not to neglect wellbeing but to understand it properly. Young people flourish when they are both supported and stretched. The desirable difficulties of good teaching and the real stakes of the co-curriculum are not in tension with wellbeing. They are part of its substance.
Belonging
Learners who feel known, valued and connected are more likely to engage, persevere and develop confidence. Belonging is therefore a condition of learning, not a soft extra. It is also unequally distributed. Students with special educational needs, those from disadvantaged backgrounds, those at risk of exclusion and those experiencing bullying are less likely to feel that school is a place where they belong. This matters urgently. Persistent absence in England remains well above pre-pandemic levels, and a lack of belonging is among its drivers. Relationships lie at the heart of belonging, yet they take time, and stretched teachers struggle to sustain them. Teacher wellbeing, workload and retention are therefore preconditions for student wellbeing. Schools should also be realistic about their role. They can promote wellbeing and notice early when something is wrong, but they are not mental health services, and they need timely access to specialist support.
Schools as communities
The causal evidence on technology and youth mental health is genuinely contested. Some argue that smartphones and social media are a major cause of declining youth mental health (Haidt, 2024); others contend that the evidence for causation is far weaker than often claimed (Orben and Przybylski, 2019; Odgers, 2024). For some young people, particularly those who are isolated, disabled or LGBTQ+, online communities offer real support. Whatever the precise causal story, schools may increasingly be among the few places where young people regularly experience sustained face-to-face relationships and participation in a genuine community. As restrictions on social media take effect, that role will only grow, and schools should embrace it deliberately.
Contact with the natural world matters too. Time outdoors can reduce stress and replenish attention (Kaplan, 1995), and Hilary Cremin (2025) argues that education should reconnect young people with the living world. So do the basics too often overlooked – sleep, physical activity and nutrition. Schools must also recognize the pressures they themselves create. High-stakes examinations are a significant source of anxiety for many young people, which is one reason why assessment reform is central to this series.
Attention as an educational priority
Simone Weil argued that the real purpose of school studies is to develop the capacity for attention. Whether or not a student solves a geometry problem, she suggested, the effort of sustained attention is never wasted, because attention is the foundation of both learning and love for others (Weil, 1951). Iris Murdoch later treated attention as a moral capacity – the patient, just and loving regard for what is real (Murdoch, 1970).
These ideas have never been more relevant. Many digital technologies are designed to capture attention through continuous stimulation and algorithmic recommendation, fragmenting concentration and crowding out reflection. Schools can help young people learn to direct their own attention through deep reading, sustained concentration, dialogue, time outdoors and clear policies on phones in school. The goal is not merely to limit screen time, but to cultivate the ability to give one’s attention, freely and wholeheartedly, to what matters: a text, a problem, a piece of music, another person. In an age designed to distract, the capacity for attention is a form of freedom.
Assessment
Assessment is the fulcrum on which education turns. However ambitious curriculum reform may be, schools inevitably give greatest attention to what is assessed. Assessment does more than measure learning; it shapes it. It influences what teachers prioritize, how learners study and, ultimately, what society comes to value. If assessment remains unchanged, much of what this series proposes will remain difficult to realize. If education exists to enable human flourishing through knowledge, agency, character and service, assessment must recognize those same capacities: not only what learners know but how they think, apply, create, communicate, reflect and contribute.
Reliability, validity and purpose
At the heart of the debate lies a trade-off. Reliability is whether an assessment produces consistent results; validity is whether it measures what actually matters. Written examinations are designed to maximize reliability, and they do that well. But reliability can come at the expense of validity. Capacities such as oral reasoning, sustained inquiry, collaboration and practical judgement are hard to capture in a timed paper, and what is not assessed tends to be squeezed out of teaching.
It is also helpful to distinguish the purposes assessment serves. It supports learning, certifies achievement, selects for the next stage and holds schools to account. Much of the distortion in our system arises because a single set of examinations at 16 is asked to do all four at once. Now that young people in England must remain in education or training until 18, it is fair to ask whether a high-stakes national examination at 16 still serves them well. Most assessment, of course, never appears in a results table. It happens daily in classrooms as formative assessment, among the most powerful influences on learning (Black and William, 1998), and it depends entirely on professional judgement.
The case for exams, and its limits
The Curriculum and Assessment Review concluded that externally set and marked examinations remain “the fairest and most reliable method of assessment” (Department for Education, 2025a, p. 136). It recommended no expansion of written coursework, citing the risks posed by generative AI, while rightly recommending a reduction in examination time of at least 10%. Its wider conclusions rest on three arguments, each of which deserves a serious answer.
The first is objectivity. Examinations are often presented as objective and teacher judgement as unreliable. This is a false opposition. All assessment depends on human judgement: in deciding what is worth assessing, designing tasks, interpreting responses and setting grade boundaries. England’s exams regulator, Ofqual’s, own research shows that in essay-based subjects there is a substantial probability that a candidate’s grade would differ if their script were marked by a senior examiner (Ofqual, 2018). The question is not whether judgement is used, but how it can be exercised rigorously, consistently and fairly.
The second is fairness, and this is the most serious. Research has found teacher assessments systematically lower, relative to test scores, for some ethnic minority students, disadvantaged students and students with special educational needs (Burgess and Greaves, 2013; Campbell, 2015). Coursework can favor those with educated parents and quiet rooms at home. These concerns are real. But the response is to design for fairness, not to abandon broader assessment. Such design would include clear shared standards, anonymized marking where possible, robust moderation, external quality assurance and evidence triangulated across several tasks. Examinations carry inequities of their own, in the advantages conferred by private tuition and the toll of exam anxiety. The teacher-assessed grades of 2020 and 2021 are often cited against professional judgement, but they were produced in crisis conditions, without the moderation and quality assurance that trustworthy assessment requires.
The third is artificial intelligence. AI does make the authenticity of unsupervised written work harder to establish. But this is an argument against unsupervised written products, not against richer assessment. The answer is to assess learning as a process as well as a product: planning, drafting, supervised work, practical demonstrations, oral dialogue and reflective commentary, with evidence triangulated rather than resting on any single task. A learner who must explain and defend their thinking in conversation cannot outsource it. Assessment should also begin to ask not whether learners used AI, but how well they used it: whether they questioned its outputs, verified its claims and exercised their own judgement.
It is reasonable, therefore, to question whether examinations are necessarily the fairest and most reliable method of assessment, and to look more closely at the conditions needed to support teacher judgement in ways that command confidence across the system.
A broader system
A better system does not need to abolish examinations. A slimmer core of external examinations could continue to certify secure knowledge reliably. Alongside it, a broader profile would recognize what examinations cannot capture: an extended project, oral assessment, practical or creative performance, and evidence of contribution and service. Mastery-based assessment, allowing learners to demonstrate understanding when they are ready, could reduce reliance on high-pressure, fixed-time testing. On-screen assessment will make this more feasible over time, although current plans in England are cautious (Ofqual, 2025).
Models that combine several forms of assessment already exist. The International Baccalaureate combines examinations with an extended essay, oral assessment and service. In Scotland, the Hayward review proposed a Scottish Diploma of Achievement, recording a broader profile of learning alongside qualifications (Hayward, 2023). Rethinking Assessment has developed a learner profile designed to do the same in England.
At St Paul’s Girls’ School, we replaced some GCSEs with our own courses in art, music, drama, computer science and creative technologies. They were assessed through a variety of methods, including presentations, written examinations, open-book assessment, problem-solving tasks and projects. All were externally moderated by teachers from other schools. The sample was small and drawn from a highly selective school, so the findings should be treated with caution. But our own analysis of university applications found that students who had taken one or more of these courses were at least as successful in applying to highly selective universities, and if anything, more so.
These examples suggest that rigorous alternatives to the standard examination model can be designed, assessed and trusted. A national system would need shared standards, trained assessors and a moderation infrastructure. Building that capacity, and the confidence in professional judgement that goes with it, is the real work of reform.
AI has made clearer than ever that if education is to change, assessment must change with it. When assessment values only what is easily measured, education narrows; when it reflects the full purposes of education, teaching and learning are free to pursue them. Educational values will succeed or fail according to what schools are asked to recognize and reward.
Achieving this requires honesty about what could go wrong. The final part of this series turns to the risks AI poses, and the safeguards they demand.


