Can Everyone Have a World Class Tutor?

Introduction

A world-class tutor for everyone is one of the most concrete versions of the AI bloom idea. If advanced AI can make high-quality guidance cheap, patient, multilingual and available at any hour, then “abundant intelligence” would not only belong to laboratories, firms or wealthy families. It could reach children learning fractions, adults changing careers, disabled learners needing accessible materials, and communities where expert teachers are scarce.

The promise is real but conditional. Tutoring has long been one of education’s strongest interventions, yet it is expensive and hard to scale. AI tutors could widen access to personalised learning by giving instant feedback, adapting explanations, asking guiding questions and helping teachers support more learners. But the same tools can also produce shallow shortcuts, biased advice, false confidence and new forms of cheating. The central question is therefore not whether AI can answer homework questions. It is whether AI can help more people genuinely understand, practise, remember, create and think for themselves. EEF[2pnas.org]pnas.orgGenerative AI without guardrails can harm learningby H Bastani · 2025 · Cited by 198 — Without guardrails, students attempt to use GPT-4…

Why tutoring matters so much

The appeal of AI tutoring begins with an old education problem: individual help works, but most systems cannot provide enough of it. The Education Endowment Foundation’s evidence summary finds that one-to-one tuition can produce about five additional months of progress on average, while small-group tuition can also have sizeable effects when it is well targeted and delivered. This does not mean tutoring is magic; quality, frequency, subject fit and pupil engagement all matter. But it explains why “a tutor for every learner” has such strong intuitive force.[EEF]educationendowmentfoundation.org.ukrogress on average. Short, regular sessions (about…Read more…

The famous version of this ambition is Benjamin Bloom’s “2 sigma problem”. In 1984, Bloom asked whether group instruction could be made as effective as one-to-one tutoring combined with mastery learning. The paper became shorthand for a powerful aspiration: not merely giving learners more content, but helping them master each step before moving on. Modern AI tutoring inherits that aspiration, though the original result should not be treated as a simple promise that any digital tutor will double learning.[Massachusetts Institute of Technology]web.mit.eduMassachusetts Institute of Technology Benjamin Bloom, The 2-sigma problem</span><span class="citation-popover-snippet">Massachusetts Institute of Technology Benjamin Bloom, The 2-sigma problem

What makes AI different from earlier educational technology is not just digitisation. Schools have had screens, learning platforms and adaptive quizzes for years, with mixed results. The new possibility is conversational help at scale: a system that can diagnose a misconception, generate a fresh example, translate an explanation, simulate a patient examiner, or coach a novice tutor in real time. That could matter most where human attention is scarce.

This is also why education belongs near the centre of the AI bloom thesis. Scientific acceleration, better institutions, healthier lives and post-scarcity production all require people who can learn faster and reason better. If AI only automates tasks for a minority, it narrows human agency. If it helps billions of people learn, retrain and participate more fully in knowledge work and civic life, it expands the human base from which future discovery and judgement can grow.

What personalised AI tutoring could change

A strong AI tutor does more than provide answers. It notices what the learner is trying to do, identifies the next obstacle, and keeps the learner mentally active. In practice, the best designs tend to use familiar teaching principles: worked examples, retrieval practice, spaced repetition, hints before solutions, Socratic questioning, feedback on mistakes, and adaptation to a learner’s pace.

Recent evidence is encouraging, especially when AI is deliberately designed around pedagogy rather than bolted onto learning as a general chatbot. A Harvard-led randomised study published in Scientific Reports found that undergraduate physics students using a custom AI tutor learned significantly more in less time than students in an active-learning classroom, while also reporting higher engagement and motivation. The important detail is that the tutor was built around the same research-based teaching principles as the classroom lessons; it was not simply an unrestricted answer machine.[Nature]nature.comSource details in endnotes.

Another useful case is Tutor CoPilot, a Stanford-led human-AI tutoring study involving 900 tutors and 1,800 K-12 students from historically underserved communities. Rather than replacing tutors, the system gave tutors real-time suggestions based on expert tutoring strategies. Students whose tutors had access to Tutor CoPilot were four percentage points more likely to master lesson topics, with gains of up to nine percentage points for students assigned to lower-rated or less-experienced tutors. The researchers also found that tutors using the system were more likely to ask guiding questions and less likely to give away answers.[arXiv]arxiv.orgarXiv Tutor Co Pilot: A Human-AI Approach for Scaling Real-Time ExpertisearXiv Tutor Co Pilot: A Human-AI Approach for Scaling Real-Time Expertise

A 2025 exploratory randomised trial in UK secondary schools tested a pedagogically tuned AI model, LearnLM, inside chat-based maths tutoring on the Eedi platform. Expert tutors supervised and could edit the AI-drafted messages. Supervisors approved 76.4% of the model’s drafted messages with no or minimal edits, and students guided by the AI-supported system performed at least as well as those chatting with human tutors on measured outcomes, with a 5.5 percentage-point advantage on novel problems in later topics. That finding is not final proof of general effectiveness, but it is a concrete sign of what carefully supervised AI tutoring may be able to do.[arXiv]arxiv.orgarXiv Tutor Co Pilot: A Human-AI Approach for Scaling Real-Time ExpertisearXiv Tutor Co Pilot: A Human-AI Approach for Scaling Real-Time Expertise

The mechanism is not mysterious. Good tutoring compresses the feedback loop. A pupil no longer waits days to discover that they misunderstood equivalent fractions. An adult learning coding can ask why an error happened, not just copy the fix. A language learner can practise speaking without embarrassment. A teacher can see which misconceptions are spreading across a class. The bloom-relevant shift is from scarce expertise to always-available cognitive scaffolding.

Several changes would matter if this became reliable and widely available:

The optimistic case is strongest when AI is understood as an amplifier of teaching and learning, not a substitute for the human purposes of education. That distinction matters because education is not merely information transfer. It is also confidence, discipline, curiosity, social belonging, ethical formation and the slow development of judgement.

The danger is fluent shortcuts, not just wrong answers

The central risk is not that AI will sometimes be inaccurate, although that matters. The deeper risk is that learners may produce better-looking work while learning less. A system that writes the essay, solves the equation, produces the code or summarises the book can make performance appear to improve while reducing the mental effort that builds durable skill.

That risk is now supported by direct evidence. A 2025 paper in Proceedings of the National Academy of Sciences found that access to GPT-4 without guardrails could harm learning: students used it as a “crutch” during practice and then performed worse when they later had to work without it. The same research found that tutor-like safeguards could mitigate the negative effects, which points to a practical lesson: the design of the AI system matters enormously.[pnas.org]pnas.orgGenerative AI without guardrails can harm learningby H Bastani · 2025 · Cited by 198 — Without guardrails, students attempt to use GPT-4…

A separate study of ChatGPT use in coding education reached a similar conclusion. Students benefited when they used large language models as tutors — asking for explanations and discussing the topic — but learning suffered when they relied on the model to complete exercises for them. The authors also found that students’ perceived gains could exceed their actual gains, creating a risk of overconfidence.[arXiv]arxiv.orgarXiv Tutor Co Pilot: A Human-AI Approach for Scaling Real-Time ExpertisearXiv Tutor Co Pilot: A Human-AI Approach for Scaling Real-Time Expertise

This is the “false mastery” problem. A learner may feel fluent because the AI produces fluent output. But fluency in the tool is not the same as fluency in the subject. A student who can prompt an essay on photosynthesis may still be unable to explain the process aloud. A novice programmer who can paste errors into a chatbot may still be unable to trace a loop. A manager who uses AI to summarise a report may miss the assumptions hidden inside the summary.

Good AI tutoring therefore needs friction. It should sometimes refuse to give the answer immediately. It should ask the learner to predict, explain, compare, retrieve from memory, show working, or critique an AI-generated response. In education, speed is not always the goal. Some difficulty is productive because it forces the mind to organise knowledge.

The OECD’s 2026 Digital Education Outlook makes this distinction sharply: generative AI can support learning when guided by teaching principles, but outsourcing tasks to AI may improve immediate performance without creating real learning gains. That is the dividing line between cognitive empowerment and cognitive dependence.[OECD]oecd.orgoecd digital education outlook 2026 062a7394 enOECD Digital Education Outlook 202619 Jan 2026 — OECD research findings suggest that Generative AI can support learning when guided b…

Access for children, adults and underserved learners

The most morally important use of AI tutoring may be where conventional education is failing to provide enough individual support. Globally, UNESCO reports that 251 million children and young people remain out of school, while the World Bank has estimated that about seven in ten children in low- and middle-income countries cannot read and understand a simple text by age ten. UNESCO has also warned of a global need for 44 million additional primary and secondary teachers by 2030. These figures describe exactly the kind of scarcity that AI tutoring is often claimed to relieve.[UNESCO]unesco.orgOpen source on unesco.org.[World Bank]worldbank.orgOpen source on worldbank.org.

But access is not just a matter of releasing an app. The learners who could benefit most may have the least reliable internet, the least private study space, the least institutional support, and the least ability to pay. UNESCO’s 2023 Global Education Monitoring Report on technology in education warns that digital tools should be adopted on evidence of being appropriate, equitable, scalable and sustainable, and that technology should support rather than replace the human connection at the centre of teaching. UNESCO[Digital Library]digitallibrary.un.orgSource details in endnotes.

For children, the best model is likely to be school-integrated rather than child-alone. AI tutors can help pupils practise, but children still need adults who know them, protect them, interpret their progress and notice when a tool is failing. This is especially important for younger learners, who may not recognise when an explanation is misleading or when they are becoming dependent on hints.

For adults, the opportunity is different. AI tutors could make lifelong learning less intimidating and more flexible. Adults often need to retrain around work, caring responsibilities, disability, migration, redundancy or technological change. The OECD has emphasised the importance of high-quality lifelong learning for personal development, economic adaptation and social cohesion, and adult-learning policy increasingly centres on upskilling and reskilling for green and digital transitions. AI could lower the barrier to starting, practising and persisting.[OECD]oecd.orgOpen source on oecd.org.

For disabled learners, AI could become a major accessibility layer. Current and emerging tools can help with transcription, captioning, text-to-speech, speech-to-text, translation, simplified explanations, alternative formats and personalised pacing. Reviews of AI-driven assistive technologies point to benefits for learners with visual, physical, auditory, cognitive and language-related needs, while also stressing the need for educator training, inclusive infrastructure, privacy protection and technical standards. ScienceDirect[CoSN]cosn.orgSource details in endnotes.

The equity test is simple: do AI tutors reduce the gap between learners with abundant support and learners with little support, or do they widen it? A wealthy child with excellent teachers, private tutoring and premium AI may benefit first. A rural learner using a low-quality free chatbot on an unreliable phone may receive weaker help and more misinformation. Without policy, procurement standards and public investment, “AI tutor for everyone” could become another way that advantage compounds.

What teachers should hand over — and what they should keep

The strongest implementation model is not “AI replaces teachers”. It is “AI handles some forms of routine cognitive support so teachers can focus on the human and professional work that machines do poorly”. Tutor CoPilot is important precisely because it improved human tutoring rather than removing the tutor. It gave less-experienced tutors access to expert-like suggestions, helping them ask better questions and avoid simply giving away answers.[arXiv]arxiv.orgarXiv Tutor Co Pilot: A Human-AI Approach for Scaling Real-Time ExpertisearXiv Tutor Co Pilot: A Human-AI Approach for Scaling Real-Time Expertise

Teachers may reasonably hand over some tasks to AI: generating practice questions, producing alternate explanations, translating material, adapting reading levels, drafting formative quizzes, summarising patterns in student errors, or giving low-stakes practice feedback. These tasks are time-consuming and often benefit from speed and variation.

But teachers should keep responsibility for aims, trust, relationships, assessment validity, safeguarding, classroom culture and professional judgement. A teacher can notice that a pupil is anxious, bored, hungry, isolated or pretending to understand. A teacher can decide when a class needs discussion rather than more individual screen time. A teacher can connect learning to local context, moral questions, collaboration and shared attention.

Teacher training is therefore not an optional add-on. UNESCO’s guidance on generative AI in education and research calls for human-centred use, policy planning and capacity-building so that tools protect human agency and genuinely benefit learners. UNESCO has also developed AI competency frameworks for students and teachers, aiming to help education systems prepare people to use AI safely, ethically and critically.[UNESCO]unesco.orgglobal report teachers what you need knowglobal report teachers what you need know[UNESCO]unesco.orgOpen source on unesco.org.

A practical school or college policy should ask:

If those questions are ignored, AI tutoring may become a glossy layer over the same old problems: overloaded teachers, unequal access, weak assessment and insufficient support for struggling learners.

Assessment has to change

AI tutoring blurs the boundary between help and authorship. A calculator changes arithmetic assessment. A spellchecker changes writing assessment. A fluent AI assistant changes almost every take-home task. Schools and universities therefore need to distinguish between learning with AI, demonstrating learning without AI, and demonstrating the ability to use AI well.

This is already a live issue. UK universities were warned in 2025 to “stress-test” assessment after research by the Higher Education Policy Institute and Kortext found that 92% of students surveyed were using generative AI, up from 66% the previous year. Students reported using AI to explain concepts, summarise articles and suggest research ideas, while only about a third had received formal AI training from their institution.[The Guardian]theguardian.comuk universities warned to stress test assessments as 92 of students use aiuk universities warned to stress test assessments as 92 of students use ai

The answer cannot be simple prohibition. AI tools are becoming normal parts of work and study, and detectors are unreliable enough to create serious fairness problems, especially for non-native English speakers and students with different writing patterns. Reporting on UK higher education has already highlighted both admitted misuse and cases where students say they were wrongly accused of AI cheating.[The Guardian]theguardian.comuk universities warned to stress test assessments as 92 of students use aiuk universities warned to stress test assessments as 92 of students use ai

A better approach is assessment pluralism. Some tasks should be AI-free and supervised, because society still needs people who can reason, calculate, write and explain without a machine doing the work. Some tasks should be AI-permitted, because using tools critically is now part of competence. Some tasks should involve oral defence, process logs, drafts, in-class problem solving, practical performance or reflective explanation so that teachers can see the learner’s thinking rather than only the final product.

This matters for cognitive empowerment. If assessment rewards polished AI-generated outputs, students will rationally optimise for output. If assessment rewards understanding, judgement, revision, explanation and responsible tool use, AI can become a learning partner rather than a substitute mind.

AI literacy is part of the curriculum now

If AI becomes a common learning companion, students need more than access. They need AI literacy: the ability to understand what these systems are good at, where they fail, how they may encode bias, how to verify claims, how to protect privacy, and how to use assistance without surrendering judgement.

The OECD and European Commission’s AI Literacy Framework for primary and secondary education describes AI literacy as the capacity to understand AI and make meaningful, ethical decisions about its use. UNESCO’s student competency framework similarly goes beyond basic tool use, aiming to help students become responsible users and contributors to more inclusive and sustainable AI systems.[AILit Framework]oecd.orgSource details in endnotes.[OECD]oecd.orgOpen source on oecd.org.

This is not just a computing topic. A history student needs to know how AI can fabricate citations or flatten historical disagreement. A science student needs to know the difference between a plausible explanation and experimental evidence. A literature student needs to understand voice, authorship and interpretation. A vocational learner needs to know when AI-generated instructions are unsafe. A citizen needs to understand recommendation systems, synthetic media and automated decision-making.

AI literacy should also include habits of independence. Learners should practise asking: What do I know before I ask the AI? What answer do I expect? Can I explain the result in my own words? What source would verify this? What would change my mind? These habits turn AI from an oracle into a sparring partner.

For adults and older learners, AI literacy has a second purpose: avoiding exclusion. Research on conversational AI use suggests that lower education levels may be associated with greater avoidance of AI tools, and studies of older adults find motivation to learn about AI but uncertainty about where to begin. If AI becomes a gateway to services, jobs and civic information, then AI literacy becomes part of democratic inclusion.[arXiv]arxiv.orgarXiv Tutor Co Pilot: A Human-AI Approach for Scaling Real-Time ExpertisearXiv Tutor Co Pilot: A Human-AI Approach for Scaling Real-Time Expertise

The long-term bloom case: more capable humans, not passive users

The deepest educational question is what kind of people AI tutoring helps create. A narrow version of AI education aims to raise test scores, finish assignments faster or produce employable skills. Those goals matter. But the AI bloom frame asks for more: can abundant intelligence help people become more capable thinkers, creators and citizens over a lifetime?

The optimistic answer is plausible if AI tutors strengthen the habits that make humans more powerful: curiosity, persistence, memory, abstraction, explanation, collaboration, taste, moral judgement and self-correction. A good tutor does not merely transfer knowledge; it helps the learner build an internal model of the subject. It gradually withdraws support so the learner can stand alone.

This is why AI tutoring should be judged by delayed and transferable learning, not only immediate task success. Can students solve new problems later? Can they explain without the tool? Can they detect when the tool is wrong? Can they use AI to explore a question more deeply rather than to avoid thinking? Can struggling learners catch up without becoming dependent? Can advanced learners go further without losing humility?

In the long run, the most important educational benefit of AI may be cognitive democratisation. Today, access to expert feedback is heavily rationed by geography, wealth, language, disability, confidence and institutional gatekeeping. If AI makes patient, high-quality explanation widely available, more people could enter domains that once felt closed: mathematics, programming, science, law, design, medicine, engineering, philosophy, music, languages and public reasoning.

That would not abolish inequality by itself. Education is shaped by housing, nutrition, safety, culture, funding, language, disability support, teacher quality, family time and political choices. But AI tutoring could become one of the major tools for loosening the constraint of scarce expert attention.

What would make AI tutoring genuinely empowering?

The difference between empowerment and dependency will be decided by design, institutions and incentives. A commercial system optimised for engagement may keep learners clicking. A school system optimised for league tables may use AI to drill test performance. A platform optimised for data extraction may turn children’s learning into a surveillance market. None of those futures is the same as human flourishing.

A better path would treat AI tutoring as public learning infrastructure. That means independent evaluation, child safety standards, accessibility requirements, teacher involvement, transparent procurement, strong privacy protections and evidence of learning gains across different groups. UNESCO’s technology-in-education work repeatedly cautions that technology should be used because it improves learning and inclusion, not because it is fashionable or commercially available. UNESCO[Digital Library]digitallibrary.un.orgSource details in endnotes.

The most promising systems will probably share several features:

• They teach by questioning, not just answering. They prompt learners to think, predict, retrieve and explain.
• They adapt without isolating. Personalisation should not trap students in lonely screen-based pathways or deny them shared classroom experiences.
• They are teacher-visible. Educators should be able to inspect, correct and learn from AI interactions.
• They are evidence-tested. Claims should be judged by learning outcomes, transfer, retention and equity, not by demos.
• They protect children’s data. Voice, writing, disability information, behavioural patterns and learning struggles are highly sensitive.
• They support many languages and needs. Otherwise, AI tutoring will reinforce the dominance of already advantaged learners.
• They build AI literacy. Learners should understand both the power and limits of the tool they are using.
• They fade support over time. The end goal is a stronger learner, not a permanent dependency.

The practical aim is not to create a world where everyone asks machines for answers. It is to create a world where more people have the support to ask better questions, master difficult subjects and contribute to the shared intelligence of civilisation.

The balanced verdict

AI tutoring is one of the clearest near-term pathways from advanced AI to broad human flourishing. The evidence is no longer just speculative: carefully designed AI tutors and human-AI tutoring systems have shown promising learning gains in real educational settings. At the same time, unrestricted generative AI can damage learning by letting students outsource the very effort that builds understanding.[Nature]nature.comSource details in endnotes.

The right question is therefore not “Will AI replace teachers?” or “Can a chatbot do homework?” The right question is whether societies can build and govern AI tutoring systems that make real understanding more available than shortcuts. If they can, education may become one of the main channels through which AI bloom reaches ordinary people: not as passive abundance delivered from above, but as expanded human capability from below.

A future of abundant intelligence should not mean fewer people learning because machines can perform for them. It should mean more people able to learn, reason, create, retrain and participate in shaping the long future. That is the standard AI tutoring has to meet.

Amazon book picks

Further Reading

Books and field guides related to Can Everyone Have a World Class Tutor?. Use these as the next step if you want deeper reading beyond the article.

Book

Make It Stick

By Peter C. Brown, Henry L. Roediger III et al.

Explains what tutoring must achieve to create durable learning.

See on Amazon

Book

Co-Intelligence

By Ethan Mollick

Covers how people can use AI as a collaborator and coach.

See on Amazon

Book

The AI Classroom

By Dan Fitzpatrick, Amanda Fox et al.

Practical match for AI tutoring and classroom integration.

See on Amazon

Book

Why Don't Students Like School?

By Daniel T. Willingham

Explains cognitive principles that AI tutors need to respect.

See on Amazon

Browse more on Amazon:Make It StickCo IntelligenceThe AI Classroom

As an Amazon Associate I earn from qualifying purchases.

eBay marketplace picks

Marketplace Samples

Example marketplace items related to this page. Use the search link to explore similar finds on eBay.

Shop location

UsingUSA USA USAUKCanadaAustraliaIreland

Example eBay listing

Kids Drawing Robot Educational Drawing Machine Voice Interaction Learning Toy

SearcheBay.co.uk: educational robot

Browse similar oneBay.co.uk

Example eBay listing

10-in-1 Electric STEM Building Toy, Educational Activity Robot Toy Gift for Kids

SearcheBay.co.uk: educational robot

Browse similar oneBay.co.uk

Example eBay listing

Drawing Robot For Kids With 100 Cards Automatic Painting Robot Education Toy UK

SearcheBay.co.uk: educational robot

Browse similar oneBay.co.uk

Example eBay listing

Kids Drawing Robot with Voice Interaction Educational Montessori Toy Birthday

SearcheBay.co.uk: educational robot

Browse similar oneBay.co.uk

Browse more oneBay.co.uk

Example items shown for inspiration; availability and pricing can change. Branchoria may earn a commission if you purchase through outbound eBay links.

Endnotes

1. Source: educationendowmentfoundation.org.uk
Link:https://educationendowmentfoundation.org.uk/education-evidence/teaching-learning-toolkit/one-to-one-tuition

2. Source: pnas.org
Link:https://www.pnas.org/doi/10.1073/pnas.2422633122

3. Source: oecd.org
Title: oecd digital education outlook 2026 062a7394 en
Link:https://www.oecd.org/en/publications/oecd-digital-education-outlook-2026_062a7394-en.html

4. Source: educationendowmentfoundation.org.uk
Link:https://educationendowmentfoundation.org.uk/education-evidence/teaching-learning-toolkit/small-group-tuition/technical-appendix

5. Source: nature.com
Link:https://www.nature.com/articles/s41598-025-97652-6

6. Source: arxiv.org
Title: arXiv Tutor Co Pilot: A Human-AI Approach for Scaling Real-Time Expertise
Link:https://arxiv.org/abs/2410.03017

7. Source: nssa.stanford.edu
Title: tutor copilot human ai approach scaling real time expertise
Link:https://nssa.stanford.edu/studies/tutor-copilot-human-ai-approach-scaling-real-time-expertise

8. Source: arxiv.org
Link:https://arxiv.org/abs/2512.23633

9. Source: arxiv.org
Title: arXiv AI Meets the Classroom: When Does Chat GPT Harm Learning?
Link:https://arxiv.org/abs/2409.09047

10. Source: unesco.org
Link:https://www.unesco.org/reports/gem-report/en/2024-monitoringsdg4

11. Source: unesco.org
Title: global report teachers what you need know
Link:https://www.unesco.org/en/articles/global-report-teachers-what-you-need-know

12. Source: unesco.org
Link:https://www.unesco.org/gem-report/en/publication/technology

13. Source: oecd.org
Link:https://www.oecd.org/en/topics/sub-issues/artificial-intelligence-and-education-and-skills.html

14. Source: oecd-ilibrary.org
Title: 85748b7b en
Link:https://www.oecd-ilibrary.org/content/dam/oecd/en/publications/reports/2024/07/readying-adult-learners-for-innovation_77f4315f/85748b7b-en.pdf

15. Source: sciencedirect.com
Link:https://www.sciencedirect.com/science/article/pii/S2666188825006069

16. Source: cosn.org
Link:https://www.cosn.org/wp-content/uploads/2024/09/Blaschke_Report_2024_lfp.pdf

17. Source: unesco.org
Title: guidance generative ai education and research
Link:https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research

18. Source: unesco.org
Link:https://www.unesco.org/en/articles/what-you-need-know-about-unescos-new-ai-competency-frameworks-students-and-teachers

19. Source: unesco.org
Title: ai competency framework students
Link:https://www.unesco.org/en/articles/ai-competency-framework-students

20. Source: oecd.org
Link:https://www.oecd.org/en/events/public-consultations/2025/09/ailit-framework.html

21. Source: arxiv.org
Link:https://arxiv.org/abs/2507.07881

22. Source: arxiv.org
Link:https://arxiv.org/abs/2504.14649

23. Source: unesco.org
Title: Artificial intelligence in education
Link:https://www.unesco.org/en/digital-education/artificial-intelligence

24. Source: unesco.org
Title: what you need know about ai and right education
Link:https://www.unesco.org/en/articles/what-you-need-know-about-ai-and-right-education

25. Source: unesco.org
Link:https://www.unesco.org/en/digital-education

26. Source: unesco.org
Link:https://www.unesco.org/en/articles/unesco-survey-two-thirds-higher-education-institutions-have-or-are-developing-guidance-ai-use

27. Source: iicba.unesco.org
Title: artificial intelligence and education systems
Link:https://www.iicba.unesco.org/en/artificial-intelligence-and-education-systems

28. Source: unesco.org
Title: ai and education guidance policy makers
Link:https://www.unesco.org/en/articles/ai-and-education-guidance-policy-makers

29. Source: unesdoc.unesco.org
Link:https://unesdoc.unesco.org/ark%3A/48223/pf0000388832

30. Source: unesdoc.unesco.org
Link:https://unesdoc.unesco.org/ark%3A/48223/pf0000385723.NATO

31. Source: unesco.org
Link:https://www.unesco.org/gem-report/sites/default/files/medias/fichiers/2023/07/Summary_v5.pdf

32. Source: unesdoc.unesco.org
Link:https://unesdoc.unesco.org/ark%3A/48223/pf0000390204

33. Source: gem-report-2023.unesco.org
Title: technology in education
Link:https://gem-report-2023.unesco.org/technology-in-education/

34. Source: unesdoc.unesco.org
Link:https://unesdoc.unesco.org/ark%3A/48223/pf0000391600

35. Source: unesdoc.unesco.org
Link:https://unesdoc.unesco.org/ark%3A/48223/pf0000395373.locale%3Den

36. Source: oecd.org
Title: 062a7394 en
Link:https://www.oecd.org/content/dam/oecd/en/publications/reports/2026/01/oecd-digital-education-outlook-2026_940e0dd8/062a7394-en.pdf

37. Source: oecd.org
Link:https://www.oecd.org/en/publications/oecd-digital-education-outlook_7fbfff45-en.html

38. Source: oecd.org
Link:https://www.oecd.org/content/dam/oecd/en/publications/reports/2024/12/quality-matters_920e920e/f44a185b-en.pdf

39. Source: arxiv.org
Link:https://arxiv.org/pdf/2410.03017

40. Source: arxiv.org
Link:https://arxiv.org/pdf/2605.16296

41. Source: unesco.at
Link:https://www.unesco.at/fileadmin/Redaktion/Bildung/2023GEM-Report_Tech_in_Ed_long.pdf

42. Source: sciencedirect.com
Link:https://www.sciencedirect.com/science/article/pii/S2666920X24001516

43. Source: web.mit.edu
Title: Massachusetts Institute of Technology Benjamin Bloom, `The 2-sigma problem
Link:https://web.mit.edu/5.95/readings/bloom-two-sigma.pdf

44. Source: worldbank.org
Link:https://www.worldbank.org/en/news/press-release/2022/06/23/70-of-10-year-olds-now-in-learning-poverty-unable-to-read-and-understand-a-simple-text

45. Source: digitallibrary.un.org
Link:https://digitallibrary.un.org/record/4020460?ln=en

46. Source: theguardian.com
Title: uk universities warned to stress test assessments as 92 of students use ai
Link:https://www.theguardian.com/education/2025/feb/26/uk-universities-warned-to-stress-test-assessments-as-92-of-students-use-ai

47. Source: theguardian.com
Link:https://www.theguardian.com/technology/2024/dec/15/i-received-a-first-but-it-felt-tainted-and-undeserved-inside-the-university-ai-cheating-crisis

48. Source: ailiteracyframework.org
Link:https://ailiteracyframework.org/

49. Source: openknowledge.worldbank.org
Link:https://openknowledge.worldbank.org/entities/publication/81b862e6-fdda-470a-a142-4a7c43e7b049

50. Source: blogs.worldbank.org
Title: is ai making us smarter or just making us look smart
Link:https://blogs.worldbank.org/en/education/is-ai-making-us-smarter-or-just-making-us-look-smart-

51. Source: documents1.worldbank.org
Link:https://documents1.worldbank.org/curated/en/099757104152527995/pdf/IDU-b1e5ef00-75ff-4ba4-a4b6-84899c3ea968.pdf

52. Source: worldbank.org
Link:https://www.worldbank.org/ext/en/topic/education/digital-technologies-in-education

53. Source: blogs.worldbank.org
Title: From chalkboards to chatbots Transforming learning in Nigeria
Link:https://blogs.worldbank.org/en/education/From-chalkboards-to-chatbots-Transforming-learning-in-Nigeria

54. Source: documents1.worldbank.org
Link:https://documents1.worldbank.org/curated/en/099734306182493324/pdf/IDU152823b13109c514ebd19c241a289470b6902.pdf

55. Source: thedocs.worldbank.org
Title: Hausman 4
Link:https://thedocs.worldbank.org/en/doc/b7ef34ba76abc3c8765d7da5c97e9a96-0070062026/original/Hausman-4.pdf

56. Source: openknowledge.worldbank.org
Link:https://openknowledge.worldbank.org/entities/publication/df09910b-a3d9-4630-b611-9aa7e862b35f

57. Source: worldbank.org
Link:https://www.worldbank.org/en/news/video/2025/08/19/ai-revolution-in-education

58. Source: worldbank.org
Link:https://www.worldbank.org/en/news/press-release/2025/10/30/new-report-offers-evidence-based-solutions-to-address-global-literacy-crisis-among-children

59. Source: worldbank.org
Title: ending learning poverty
Link:https://www.worldbank.org/en/topic/education/brief/ending-learning-poverty

60. Source: Wikipedia
Title: Bloom’s 2 sigma problem
Link:https://en.wikipedia.org/wiki/Bloom%27s_2_sigma_problem

61. Source: facebook.com
Link:https://www.facebook.com/groups/1048019523623103/posts/1447738773651174/

62. Source: unesco.org.uk
Link:https://unesco.org.uk/resources/guidance-for-generative-ai-in-education-and-research

63. Source: edtechhub.org
Link:https://edtechhub.org/world-bank-ai-in-education-global-community-of-interest/

64. Source: unesco-asp.dk
Link:https://unesco-asp.dk/wp-content/uploads/2025/02/AI-Competency-framework-for-teachers_UNESCO_2024.pdf

Additional References

65. Source: youtube.com
Link:https://www.youtube.com/watch?v=ziOVjdr5jyM

66. Source: youtube.com
Title: How AI Could Save (Not Destroy) Education | Sal Khan | TED Tech
Link:https://www.youtube.com/watch?v=gA-dYPDv1qo

67. Source: researchgate.net
Link:https://www.researchgate.net/publication/392839220_AI_tutoring_outperforms_in-class_active_learning_an_RCT_introducing_a_novel_research-based_design_in_an_authentic_educational_setting

68. Source: researchgate.net
Link:https://www.researchgate.net/publication/396808798_Leveraging_Khanmigo_Generative_AI-Powered_Tool_for_Personalized_Tutoring_to_Learn_Scientific_Concepts

69. Source: researchgate.net
Link:https://www.researchgate.net/publication/358358065_Technology_in_Education_Background_Paper_for_2023_Global_Education_Monitoring_Report

70. Source: aricafoundation.org
Link:https://aricafoundation.org/news/iniciativa-educacao/70-percent-of-10-year-olds-unable-to-read-and-understand-a-simple-text

71. Source: facebook.com
Link:https://www.facebook.com/worldbankgroup/posts/70-of-10-year-olds-in-low-middle-income-countries-are-unable-to-read-understand-/488962366604745/

72. Source: scribd.com
Link:https://www.scribd.com/document/998181886/AI-Tools-for-Accessible-Learning

73. Source: povertyactionlab.org
Link:https://www.povertyactionlab.org/initiative-project/ai-powered-tutoring-unleashing-full-potential-personalized-learning-khanmigo

74. Source: facebook.com
Link:https://www.facebook.com/100091956776588/posts/a-new-stanford-study-suggests-that-ai-assisted-tutoring-when-designed-to-support/470107319397823/