The results from Gregory Kestin's Harvard physics experiment arrived like a thunderclap. Students using an AI tutor in their dormitory rooms learned more than twice as much as their peers sitting in active learning classrooms with experienced instructors. They did it in less time. They reported feeling more engaged. Published in Scientific Reports in June 2025, the study seemed to confirm what education technology evangelists had been promising for years: artificial intelligence could finally crack the code of personalised learning at scale.
But the number that truly matters lies elsewhere. In September 2024, Khan Academy's AI-powered Khanmigo reached 700,000 students, up from just 40,000 the previous year. By the end of 2025, projections suggested more than one million students would be learning with an artificial tutor that never tires, never loses patience, and remembers every mistake a child has ever made. The question that haunts teachers, parents, and policymakers alike is brutally simple: if machines can now do what Benjamin Bloom proved most effective back in 1984, namely provide the one-to-one tutoring that outperforms group instruction by two standard deviations, does the human educator have a future?
The answer, emerging from research laboratories and classrooms across six continents, turns out to be considerably more nuanced than the headline writers would have us believe. It involves the fundamental nature of learning itself, the irreplaceable qualities that humans bring to education, and the possibility that artificial intelligence might finally liberate teachers from the very burdens that have been crushing them for decades.
The Two Sigma Dream Meets Silicon
In 1984, educational psychologist Benjamin Bloom published what would become one of the most cited papers in the history of educational research. Working with doctoral students at the University of Chicago, Bloom discovered that students who received one-to-one tutoring using mastery learning techniques performed two standard deviations better than students in conventional classrooms. The average tutored student scored higher than 98 per cent of students in the control group. Bloom called this “the 2 Sigma Problem” and challenged researchers to find methods of group instruction that could achieve similar results without the prohibitive cost of individual tutoring.
The study design was straightforward but powerful. School students were randomly assigned to one of three groups: conventional instruction with 30 students per teacher and periodic testing for marking, mastery learning with 30 students per teacher but tests given for feedback followed by corrective procedures, and one-to-one tutoring. The tutoring group's results were staggering. As Bloom noted, approximately 90 per cent of the tutored students attained the level of summative achievement reached by only the highest 20 per cent of the control class.
For forty years, that challenge remained largely unmet. Human tutors remained expensive, inconsistent in quality, and impossible to scale. Various technological interventions, from educational television to computer-assisted instruction, failed to close the gap. Radio, when it first entered schools, was predicted to revolutionise learning. Television promised the same. Each technology changed pedagogy in some ways but fell far short of approximating the tutorial relationship that Bloom had identified as the gold standard. Then came large language models.
The Harvard physics study, led by Kestin and Kelly Miller, offers the most rigorous evidence to date that AI tutoring might finally be approaching Bloom's benchmark. Using a crossover design with 194 undergraduate physics students, the researchers compared outcomes between in-class active learning sessions and at-home sessions with a custom AI tutor called PS2 Pal, built on GPT-4. Each student experienced both conditions for different topics, eliminating selection bias. The topics covered were surface tension and fluid flow, standard material in introductory physics courses.
The AI tutor was carefully designed to avoid common pitfalls. It was instructed to be brief, using no more than a few sentences at a time to prevent cognitive overload. It revealed solutions one step at a time rather than giving away complete answers. To combat hallucinations, the tendency of chatbots to fabricate information, the system was preloaded with all correct solutions. The scientists behind the experiment instructed the AI to avoid cognitive overload by limiting response length and to avoid giving away full solutions in a single message. The result: engagement ratings of 4.1 out of 5 for AI tutoring versus 3.6 for classroom instruction, with statistically significant improvements in learning outcomes (p < 10^-8). Motivation ratings showed a similar pattern: 3.4 out of 5 for AI tutoring compared to 3.1 for classroom instruction.
The study's authors were careful to emphasise limitations. Their population consisted entirely of Harvard undergraduates, raising questions about generalisability to community colleges, less selective institutions, younger students, or populations with different levels of technological access and comfort. “AI tutors shouldn't 'think' for students, but rather help them build critical thinking skills,” the researchers wrote. “AI tutors shouldn't replace in-person instruction, but help all students better prepare for it.”
The median study time also differed between conditions: 49 minutes for AI tutoring versus 60 minutes for classroom instruction. Students were not only learning more but doing so in less time, a finding that has significant implications for educational efficiency but also raises questions about what might be lost when learning is compressed.
The Global Experiment Unfolds
While researchers debate methodology in academic journals, the global education market has already placed its bets with billions of dollars. The AI in education market reached $7.05 billion in 2025 and is projected to explode to $112.30 billion by 2034, growing at a compound annual rate of 36 per cent. The market rose from $5.47 billion in 2024 to $7.57 billion in 2025, representing a 38.4 per cent increase in a single year. Global student AI usage jumped from 66 per cent in 2024 to 92 per cent in 2025, according to industry surveys. By early 2026, an estimated 86 per cent of higher education students utilised AI as their primary research and brainstorming partner.
The adoption statistics tell a remarkable story of rapid change. A survey of 2,232 teachers across the United States found that 60 per cent used AI tools during the 2024-25 school year. Usage was higher among high school teachers at 66 per cent and early-career teachers at 69 per cent. Approximately 26 per cent of districts planned to offer AI training during the 2024-25 school year, with around 74 per cent of districts expected to train teachers by the autumn of 2025. A recent survey by EDUCAUSE of more than 800 higher education institutions found that 57 per cent were prioritising AI in 2025, up from 49 per cent the previous year.
In China, Squirrel AI Learning has been at the forefront of this transformation. Founded in 2014 and headquartered in Shanghai, the company claims more than 24 million registered students and 3,000 learning centres across more than 1,500 cities. When Squirrel AI set a Guinness World Record in September 2024 by attracting 112,718 students to an online mathematics lesson in 24 hours, its adaptive learning system generated over 108,000 unique learning pathways, tailoring instruction to 99.1 per cent of participants. The system designed 111,704 unique exercise pathways for the students, demonstrating the scalability of personalised instruction. The company reports performance improvements of up to 30 per cent compared to traditional instruction.
Tom Mitchell, the former Dean of Computer Science at Carnegie Mellon University, serves as Squirrel AI's Chief AI Officer, lending academic credibility to its technical approach. The system breaks down subjects into thousands of knowledge points. For middle school mathematics alone, it maps over 10,000 concepts, from rational numbers to the Pythagorean theorem, tracking student learning behaviours to customise instruction in real time. In December 2024, Squirrel AI announced that Guinness World Records had certified its study as the “Largest AI vs Traditional Teaching Differential Experiment,” conducted with 1,662 students across five Chinese schools over one semester. The company has also provided 10 million free accounts to some of China's poorest families, addressing equity concerns that have plagued educational technology deployment.
Carnegie Learning, another pioneer in AI-powered mathematics education, has accumulated decades of evidence. Founded by cognitive psychologists and computer scientists at Carnegie Mellon University who partnered with mathematics teachers at the Pittsburgh Public Schools, the company has been compiling and acting on data to refine and improve how students explore mathematics since 1998. In an independent study funded by the US Department of Education and conducted by the RAND Corporation, the company's blended approach nearly doubled growth in performance on standardised tests relative to typical students in the second year of implementation. The gold standard randomised trial included more than 18,000 students at 147 middle and high schools across the United States. The Institute of Education Sciences published multiple reports on the effectiveness of Carnegie's Cognitive Tutor, with five of six qualifying studies showing intermediate to significant positive effects on mathematics achievement.
Meanwhile, Duolingo has transformed language learning through its AI-first strategy, producing a 51 per cent boost in daily active users and fuelling a one billion dollar revenue forecast. The company reported 47.7 million daily active users in Q2 2025, a 40 per cent year-over-year increase, with paid subscribers rising to 10.9 million and a 37 per cent year-over-year gain. Quarterly revenue grew to $252.3 million, up 41 per cent from Q2 2024. Survey data indicates that 78 per cent of regular users of its Roleplay feature, which allows practice conversations with AI characters, feel more prepared for real-world conversations after just four weeks. The “Explain My Answer” feature, adopted by 65 per cent of users, increased course completion rates by 15 per cent. Learning speed increased 40 to 60 per cent compared to pre-2024 applications.
Khan Academy's trajectory illustrates the velocity of this transformation. Khanmigo's reach expanded 731 per cent year-over-year to reach a record number of students, teachers, and parents worldwide. The platform went from about 68,000 Khanmigo student and teacher users in partner school districts in 2023-24 to more than 700,000 in the 2024-25 school year, expanding from 45 to more than 380 district partners. When rating AI tools for learning, Common Sense Media gave Khanmigo 4 stars, rising above other AI tools such as ChatGPT and Bard for educational use. Research from Khan Academy showed that combining its platform and AI tutor with additional tools and services designed for districts made it 8 to 14 times more effective at driving student learning outcomes compared with independent learning.
What Machines Cannot Replicate
Yet for all the impressive statistics and exponential growth curves, a growing body of research suggests that the most crucial elements of education remain stubbornly human.
A 2025 systematic review published in multiple peer-reviewed journals identified a troubling pattern: while AI-driven intelligent tutoring systems can improve student performance by 15 to 35 per cent, over-reliance on these systems can reduce critical thinking, creativity, and independent problem-solving. Researchers have termed this phenomenon “cognitive offloading,” the tendency of students to delegate mental work to AI rather than developing their own capabilities. Research also indicates that over-reliance on AI during practice can reduce performance in examinations taken without assistance, suggesting that AI-enhanced learning may not always translate to improved independent performance.
The ODITE 2025 Report, titled “Connected Intelligences: How AI is Redefining Personalised Learning,” warned about the excessive focus on AI's technical benefits compared to a shallow exploration of socio-emotional risks. While AI can enhance efficiency and personalise learning, the report concluded, excessive reliance may compromise essential interpersonal skills and emotional intelligence. The report called for artificial intelligence to be integrated within a pedagogy of care, not only serving performance but also recognition, inclusion, and listening.
These concerns are not merely theoretical. A study of 399 university students and 184 teachers, published in the journal Teaching and Teacher Education, found that the majority of participants argued that human teachers possess unique qualities, including critical thinking and emotions, which make them irreplaceable. The findings emphasised the importance of social-emotional competencies developed through human interactions, capacities that generative AI technologies cannot currently replicate. Participants noted that creativity and emotion are precious aspects of human quality which AI cannot replace.
Human teachers bring what researchers call “emotional intelligence” to the classroom: the ability to read subtle social cues that signal student engagement or confusion, to understand the complex personal circumstances that might affect performance, to provide the mentorship, encouragement, and emotional support that shape not just what students know but who they become. As one education researcher told the World Economic Forum: “AI cannot replace the most human dimensions of education: connection, belonging, and care. Those remain firmly in the teacher's domain.” Teachers play a vital role in guiding students to think critically about when AI adds value and when authentic human thinking and creativity are irreplaceable.
The American Psychological Association's June 2025 health advisory on AI companion software underscored these concerns in alarming terms. AI systems, the advisory warned, exploit emotional vulnerabilities through unconditional regard, triggering dependencies like digital attachment disorder while hindering social skill development. The advisory noted that manipulative design may displace or interfere with the development of healthy real-world relationships. For teenagers in particular, confusing algorithmic responses for genuine human connection can directly short-circuit developing capacities to navigate authentic social relationships and assess trustworthiness.
While AI can be a helpful supplement, genuine human connections release oxytocin, the “bonding hormone” that plays a crucial role in reducing stress and fostering emotional wellbeing. Current AI does not yet possess the empathy, intuition, and depth of understanding that humans bring to conversations. For example, a teenager feeling isolated might share their feelings with a chatbot, but the AI's responses may be generic or may not fully address deeper issues that a trained human educator would recognise and address.
The Creativity Conundrum
Beyond emotional intelligence lies another domain where human teachers remain essential: nurturing creativity.
AI tutoring systems excel at structured learning tasks, at drilling multiplication tables, at correcting grammar mistakes, at providing step-by-step guidance through physics problems. But great teachers do not just transmit facts. They inspire curiosity, challenge students to think beyond textbooks, and encourage discussions that lead to deeper understanding. When it comes to fostering creativity and open-ended problem-solving, current AI tools fall short. They lack the capacity to recognise a student's unconventional approach as potentially brilliant rather than simply incorrect.
“In the AI era, human creativity is increasingly recognised as a critical and irreplaceable capability,” noted a 2025 analysis in Frontiers in Artificial Intelligence. Fostering creativity in education requires attention to pedagogical elements that current AI systems cannot provide: the spontaneous question that opens a new line of inquiry, the willingness to follow intellectual tangents wherever they might lead, the ability to sense when a student needs encouragement to pursue an unorthodox idea. Predictions of teachers being replaced are not new. Radio, television, calculators, even the internet: each was once thought to make educators obsolete. Instead, each changed pedagogy while reinforcing the irreplaceable role of teachers in helping students make meaning, navigate complexity, and grow as people.
UNESCO's AI Competency Framework for Teachers, launched in September 2024, explicitly addresses this tension. The framework calls for a human-centred approach that integrates AI competencies with principles of human rights and human accountability. Teachers, according to UNESCO, must be equipped not only to use AI tools effectively but also to evaluate their ethical implications and to support AI literacy in students, encouraging responsible use and critical engagement with the technology.
The framework identifies five key competency aspects: a human-centred mindset that defines the critical values and attitudes necessary for interactions between humans and AI-based systems; AI ethics that establishes essential ethical principles and regulations; AI foundations and applications that specifies transferable knowledge and skills for selecting and applying AI tools; and the ability to use AI for professional development. Since 2024, UNESCO has supported 58 countries in designing or improving digital and AI competency frameworks, curricula, and quality-assured training for educators and policymakers. During Digital Learning Week 2025, UNESCO released a new report titled “AI and education: protecting the rights of learners,” providing an urgent call to action and analysing how AI and digital technologies impact access, equity, quality, and governance in education.
Liberation Through Automation
Perhaps the most compelling argument against AI teacher replacement comes from an unexpected source: the teachers themselves.
A June 2025 poll conducted by the Walton Family Foundation and Gallup surveyed more than 2,200 teachers across the United States. The findings were striking: teachers who use AI tools at least weekly save an average of 5.9 hours per week, equivalent to roughly six additional weeks of time recovered across a standard school year. This “AI dividend” allows educators to reinvest in areas that matter most: building relationships with students, providing individual attention, and developing creative lessons. Teachers who engage with AI tools more frequently report greater time savings: weekly AI users save an average of 5.9 hours each week, twice as much time as those who only use AI monthly at 2.9 hours per week.
The research documented that teachers spend up to 29 hours per week on non-teaching tasks: writing emails, grading, finding classroom resources, and completing administrative work. They have high stress levels and are at risk for burnout. Nearly half of K-12 teachers report chronic burnout, with 55 per cent considering early departure from the profession, creating a district-wide crisis that threatens both stability and student outcomes. Schools with an AI policy in place are seeing a 26 per cent larger “AI dividend,” equivalent to 2.3 hours saved per week per teacher, compared with 1.7 hours in schools without such a policy.
Despite the benefits of the “AI dividend,” only 32 per cent of teachers report using AI at least weekly, while 28 per cent use it infrequently and 40 per cent still are not using it at all. Educators use AI to create worksheets at 33 per cent, modify materials to meet students' needs at 28 per cent, complete administrative work at 28 per cent, and develop assessments at 25 per cent. Teachers in schools with an AI policy are more likely to have used AI in the past year at 70 per cent versus 60 per cent for those without.
AI offers a potential solution, not by replacing teachers but by automating the tasks that drain their energy and time. Teachers report using AI to help with lesson plans, differentiate materials for students with varying needs, write portions of individualised education programmes, and communicate with families. Sixty-four per cent of surveyed teachers say the materials they modify with AI are better quality. Sixty-one per cent say AI has improved their insights about student performance, and 57 per cent say AI has led them to enhance the quality of their student feedback and grading.
Sal Khan, founder of Khan Academy and the driving force behind Khanmigo, has consistently framed AI as a teaching assistant rather than a replacement. “AI is going to become an amazing teaching assistant,” Khan stated in March 2025. “It's going to help with grading papers, writing progress reports, communicating with others, personalising their classrooms, and writing lesson plans.” He has emphasised in multiple forums that “Teachers will always be essential. Technology has the potential to bridge learning gaps, but the key is to use it as an assistant, not a substitute.” Khan invokes the historical wisdom of one-to-one tutoring: “For most of human history, if you asked someone what great education looks like, they would say it looks like a student with their tutor. Alexander the Great had Aristotle as his tutor. Aristotle had Plato. Plato had Socrates.”
This vision aligns with what researchers call the “hybrid model” of education. The World Economic Forum's Shaping the Future of Learning insight report highlights that the main impacts of AI will be in areas such as personalised learning and augmented teaching. AI lifts administrative burdens so that people in caring roles can focus on more meaningful tasks, such as mentorship. As the role of the educator shifts, teachers are moving from traditional content delivery to facilitation, coaching, and mentorship. The future classroom is not about replacing teachers but about redefining their role from deliverers of content to curators of experience.
The Equity Question Looms Large
Any serious discussion of AI in education must confront a troubling reality: the technology that promises to democratise learning may instead widen existing inequalities.
A 2025 analysis by the International Center for Academic Integrity warned that unequal access to artificial intelligence is widening the educational gap between privileged and underprivileged students. Students from lower-income backgrounds, those in rural areas, and those attending institutions with fewer resources are often at a disadvantage when it comes to accessing the technology that powers AI tools. For these students, AI could become just another divide, reinforcing the gap between those who have and those who do not. The disproportionate impact on marginalised communities, rural populations, and underfunded educational institutions limits their ability to benefit from AI-enhanced learning.
The numbers bear this out. Half of chief technology officers surveyed in 2025 reported that their college or university does not grant students institutional access to generative AI tools. More than half of students reported that most or all of their instructors prohibit the use of generative AI entirely, according to EDUCAUSE's 2025 Students and Technology Report. AI tools often require reliable internet access, powerful devices, and up-to-date software. In regions where these resources are not readily available, students are excluded from AI-enhanced learning experiences. Much current policy energy is consumed by academic integrity concerns and bans, which address real risks but can inadvertently deepen divides by treating AI primarily as a threat rather than addressing the core equity problem of unequal opportunity to learn with and from AI.
Recommendations from the Brookings Institution and other policy organisations call for treating AI competence as a universal learning outcome so every undergraduate in every discipline graduates able to use, question, and manage AI. They advocate providing equitable access to tools and training so that benefits do not depend on personal subscriptions, and investing in faculty development at scale with time, training, and incentives to redesign courses and assessments for an AI-rich environment. Proposed solutions include increased investment in digital infrastructure, the development of affordable AI-based learning tools, and the implementation of inclusive policies that prioritise equitable access to technology. Yet only 10 per cent of schools and universities have formal AI use guidelines, according to a UNESCO survey of more than 450 institutions.
The OECD Digital Education Outlook 2026 offers a more nuanced perspective. Robust research evidence demonstrates that inexperienced tutors can enhance the quality of their tutoring and improve student learning outcomes by using educational AI tools. However, the report emphasises that if AI is designed or used without pedagogical guidance, outsourcing tasks to the technology simply enhances performance with no real learning gains. Research validates that adaptive learning's positive effects on educational equity have the capability of redressing socioeconomically disadvantaged conditions by ensuring equitable availability of educational resources.
North America dominated the AI in education market with a market share of 36 per cent in 2024, while the Asia Pacific region is expected to grow at a compound annual rate of 35.3 per cent through 2030. The United States AI in education market alone was valued at $2.01 billion in 2025 and is projected to reach $32.64 billion by 2034. A White House Executive Order signed in April 2025, “Advancing Artificial Intelligence Education for American Youth,” aims to unite AI education across all levels of learning. The US Department of Education also revealed guidance supporting schools to employ existing federal grants for AI integration.
Training the Teachers Who Will Train with AI
The gap between student adoption and teacher readiness presents a significant challenge. In 2025, there remains a notable gap between students' awareness of AI and teachers' readiness to implement AI in classrooms. According to Forbes, while 63 per cent of US teens are using AI tools like ChatGPT for schoolwork, only 30 per cent of teachers report feeling confident using these same AI tools. This difference emphasises the critical need for extensive support and AI training for all educators.
Experts emphasise the need for more partnerships between K-12 schools and higher education that provide mentorship, resources, and co-developed curricula with teachers. Faculty and researchers can help simplify AI for teachers, offer training, and ensure educational tools are designed with classroom realities in mind. AI brings a new level of potential to the table, a leap beyond past solutions. Instead of just saving time, AI aims to reshape how teachers manage their classrooms, offering a way to automate the administrative load, personalise student support, and free up teachers to focus on what they do best: teaching.
The key to successful AI integration is balance. AI has the potential to alleviate burnout and improve the teaching experience, but only if used thoughtfully as a tool, not a replacement. Competent, research-driven teachers are not going to be replaced by AI. The vision is AI as a classroom assistant that handles routine tasks while educators focus on what only they can provide: authentic human connection, professional judgement, and mentorship.
The Horizon Beckons
The evidence suggests neither a utopia of AI-powered learning nor a dystopia of displaced teachers. Instead, a more complex picture emerges: one in which artificial intelligence becomes a powerful tool that transforms rather than eliminates the human role in education.
By 2026, over 60 per cent of schools globally are projected to use AI-powered platforms. The United States has moved aggressively, with a White House Executive Order signed in April 2025 to advance AI education from K-12 through postsecondary levels. All 50 states have considered AI-related legislation. California's SB 243, signed in October 2025 and taking effect on 1 January 2026, requires operators of “companion chatbots” to maintain protocols for preventing their systems from producing content related to suicidal ideation and self-harm, with annual reports to the California Department of Public Health. New York's AI Companion Models law, effective 5 November 2025, requires notifications that state in bold capitalised letters: “THE AI COMPANION IS A COMPUTER PROGRAM AND NOT A HUMAN BEING. IT IS UNABLE TO FEEL HUMAN EMOTION.”
The PISA 2025 Learning in the Digital World assessment will focus on two competencies essential to learning with technologies: self-regulated learning and the ability to engage with digital tools. Results are expected in December 2027. Looking further ahead, the PISA 2029 Media and Artificial Intelligence Literacy assessment will examine whether young students have had opportunities to engage proactively and critically in a world where production, participation, and social networking are increasingly mediated by digital and AI tools. The OECD and European Commission's draft AI Literacy Framework, released in May 2025, aims to define global AI literacy standards for school-aged children, equipping them to use, understand, create with, and critically engage with AI.
The future English language classroom, as described by Oxford University Press, will be “human-centred, powered by AI.” Teachers will shift from traditional content delivery to facilitation, coaching, and mentorship. AI will handle routine tasks, while humans focus on what only they can provide: authentic connection, professional judgement, and the kind of mentorship that shapes lives. Hyper-personalised learning is becoming standard, with students needing tailored, real-time feedback more than ever, and AI adapting instruction moment to moment based on individual readiness.
“It is one of the biggest misnomers in education reform: that if you can give kids better technology, if you give them a laptop, if you give them better content, they will learn,” observed one education leader in a World Economic Forum report. “Children learn when they feel safe, when they feel cared for, and when they have a community of learning.”
AI tutors can adapt to a child's learning style in real time. They can provide feedback at midnight on a Saturday when no human teacher would be available. They can remember every mistake and track progress with precision no human memory could match. But they cannot inspire a love of learning in a struggling student. They cannot recognise when a child is dealing with problems at home that affect performance. They cannot model what it means to be curious, empathetic, and creative. While AI can automate some tasks, it cannot replace the human interaction and emotional support provided by teachers. There are legitimate concerns that over-reliance on AI could erode the teacher-student relationship and the social skills students develop in the classroom.
By combining the analytical power of AI with the irreplaceable human element of teaching, we can truly transform education for the next generation. Collaboration is the future. The most effective classrooms will combine human insight with AI precision, creating a hybrid model that supports personalised learning. With AI doing the busy work, teachers dedicate their time and energy to building confidence, nurturing creativity, and cultivating critical thinking skills in their students. This human touch and mentorship are invaluable and can never be fully replaced by AI.
The question is not whether AI will replace human teachers. The question is whether we will have the wisdom to use this technology in ways that enhance rather than diminish what makes education fundamentally human. As Sal Khan put it: “The question isn't whether AI will be part of education. It's how we use it responsibly to enhance learning.”
For now, the answer to that question remains in human hands.
References & Sources
- Bloom, B. S. (1984). “The 2 Sigma Problem: The Search for Methods of Group Instruction as Effective as One-to-One Tutoring.” Educational Researcher, 13(6), 4-16. https://journals.sagepub.com/doi/10.3102/0013189X013006004
- Kestin, G., & Miller, K. (2025). “AI Tutoring Outperforms Active Learning.” Scientific Reports. https://www.nature.com/articles/s41598-025-85814-z
- RAND Corporation. Carnegie Learning Cognitive Tutor Study. https://www.carnegielearning.com/why-cl/research/
- “A systematic review of AI-driven intelligent tutoring systems (ITS) in K-12 education.” npj Science of Learning. https://www.nature.com/articles/s41539-025-00320-7
- “Will generative AI replace teachers in higher education? A study of teacher and student perceptions.” Teaching and Teacher Education. https://www.sciencedirect.com/science/article/abs/pii/S0191491X24000749
- Institute of Education Sciences. Reports on Cognitive Tutor effectiveness. https://ies.ed.gov/
- UNESCO. (2024). AI Competency Framework for Teachers. https://unesco-asp.dk/wp-content/uploads/2025/02/AI-Competency-framework-for-teachers_UNESCO_2024.pdf
- UNESCO. (2025). “AI and education: protecting the rights of learners.” https://unesdoc.unesco.org/ark:/48223/pf0000395373
- OECD. (2026). Digital Education Outlook 2026. https://www.oecd.org/en/publications/oecd-digital-education-outlook-2026_062a7394-en.html
- OECD-European Commission. (2025). “Empowering Learners for the Age of AI: An AI Literacy Framework for Primary and Secondary Education.” https://oecdedutoday.com/new-ai-literacy-framework-to-equip-youth-in-an-age-of-ai/
- White House Executive Order. (2025). “Advancing Artificial Intelligence Education for American Youth.”
- Khan Academy Annual Report SY24-25. https://annualreport.khanacademy.org/
- Walton Family Foundation & Gallup. (2025). “The AI Dividend: New Survey Shows AI Is Helping Teachers Reclaim Valuable Time.” https://www.waltonfamilyfoundation.org/the-ai-dividend-new-survey-shows-ai-is-helping-teachers-reclaim-valuable-time
- Precedence Research. (2025). “AI in Education Market Size.” https://www.precedenceresearch.com/ai-in-education-market
- Common Sense Media. (2025). “AI Companions Decoded.” https://www.commonsensemedia.org/ai-companions
- EDUCAUSE. (2025). Survey of Higher Education Institutions and Students and Technology Report.
- American Psychological Association. (2025). Health Advisory on AI Companion Software.
- World Economic Forum. (2025). “How AI and human teachers can collaborate to transform education.” https://www.weforum.org/stories/2025/01/how-ai-and-human-teachers-can-collaborate-to-transform-education/
- World Economic Forum. (2025). “AI is transforming education by allowing us to be more human.” https://www.weforum.org/stories/2025/12/ai-is-transforming-education-by-allowing-us-to-be-more-human/
- Brookings Institution. (2025). “AI and the next digital divide in education.” https://www.brookings.edu/articles/ai-and-the-next-digital-divide-in-education/
- Forbes. (2025). Student and teacher AI adoption statistics.
- Oxford University Press. (2025). “The Future English Language Classroom: Human-Centred, Powered By AI.” https://teachingenglishwithoxford.oup.com/2025/06/23/future-english-language-classroom-ai/
- Squirrel AI. (2024). Guinness World Record announcement. https://www.prnewswire.com/news-releases/squirrel-ai-learning-sets-guinness-world-record-for-the-most-users-to-take-an-online-mathematics-lesson-in-24-hours-302324623.html
- Duolingo. (2025). Q2 2025 Financial Results and 2025 Language Report. https://blog.duolingo.com/2025-duolingo-language-report/
Tim Green
UK-based Systems Theorist & Independent Technology Writer
Tim explores the intersections of artificial intelligence, decentralised cognition, and posthuman ethics. His work, published at smarterarticles.co.uk, challenges dominant narratives of technological progress while proposing interdisciplinary frameworks for collective intelligence and digital stewardship.
His writing has been featured on Ground News and shared by independent researchers across both academic and technological communities.
ORCID: 0009-0002-0156-9795
Email: tim@smarterarticles.co.uk
