The Great Rejection: Questioning the Future We’re Being Sold

There’s a familiar pattern to the way things unfold when enough money, influence, and institutional momentum gather behind an idea. What begins as a possibility gradually becomes an expectation, then an obligation, until questioning the direction of travel is treated as a failure to keep up. We’ve seen this play out across industries and generations, and the current push toward an AI-everything future seems to follow a similar trajectory. Yet beneath the promises of convenience, efficiency, and unprecedented progress, there are questions about what we’re building, what we’re sacrificing, and whether the future being sold to us is one that people actually want to live in.

.   .   .

In this discourse we explored the growing resistance to the pervasive integration of digital technology and artificial intelligence into education, childhood, human relationships, public infrastructure, and society at large. We examined government restrictions on smartphones and screen time, educators and academics questioning the educational value of digital-first approaches, concerns about children’s dependence on synthetic companionship, and the financial incentives driving the AI investment race. We also considered the disconnect between technological ambition and public consent, the consequences of manufactured urgency, and the possibility of a broader cultural correction that places human flourishing, autonomy, and meaningful relationships ahead of indiscriminate technological expansion.

Key Points Discussed

• The rollback of digital-first education: Governments and school systems are restricting smartphones, reducing screen exposure, restoring printed materials, and reconsidering the assumption that more classroom technology necessarily improves learning.
• Resistance from within education: Teachers, professors, administrators, and unions are questioning routine screen use and unrestricted AI adoption, emphasizing independent thought, professional judgment, direct instruction, and meaningful human interaction.
• The developmental costs of digital dependence: Research raises concerns about sleep, attention, emotional regulation, mental health, and social development, particularly among children and adolescents. The strength of the evidence varies, and association does not always establish causation.
• Synthetic companionship and human connection: AI companions introduce a different dimension to digital dependence by simulating empathy, attention, and emotional intimacy. The potential consequences are particularly concerning when young or vulnerable people begin relying on artificial relationships in place of human connection.
• The commercial incentives behind constant engagement: Products designed to maximize attention and continued use can conflict with individual well-being, especially when their users are children whose capacity for self-regulation is still developing.
• The economics of manufactured urgency: Competitive pressure, investor expectations, infrastructure commitments, and fear of falling behind can encourage adoption before institutions have established whether a technology delivers meaningful benefits.
• The AI investment wager: Enormous capital commitments to computing infrastructure create pressure to expand markets and usage. Projected demand, actual demand, and demonstrable economic returns remain distinct questions.
• The difference between capability and necessity: The ability to automate a task does not establish that automation is desirable. Efficiency, convenience, and technical possibility are not sufficient measures of social value.
• Public resistance and infrastructure conflicts: Opposition to data centers, concerns about energy and water consumption, and demands for meaningful community participation reveal the physical costs behind an increasingly digital economy.
• The erosion of public consent: Decisions about technology are often made through procurement arrangements, political commitments, corporate consolidation, and infrastructure investments before affected communities fully understand the consequences.
• The limits of technological inevitability: Corporate enthusiasm and massive capital expenditure do not constitute public consensus. Nor does the existence of a technology establish that every institution, individual, or community must adopt it.
• The possibility of a broader cultural correction: The emerging response could take the form of selective adoption, stronger regulation, or a sharper backlash against overinvestment and unwanted technological dependence.
• Small refusals can preserve meaningful alternatives: Parents limiting devices, schools returning to books, professors requiring independent work, and communities questioning infrastructure projects can collectively challenge the assumption that every available technology must be integrated into daily life.
• Human flourishing as the measure of progress: Attention, creativity, independent thought, physical activity, friendship, community, and the ability to live meaningfully without constant digital mediation are valuable in their own right.

Key Takeaways

• Technological progress is not synonymous with human progress.
• Digital tools and AI should demonstrate their value rather than receive automatic approval.
• Children and adolescents deserve particular protection from commercial systems designed to capture attention and simulate intimacy.
• The financial and institutional momentum behind AI does not establish that its projected future is inevitable or universally desirable.
• Public awareness and meaningful consent must precede, rather than follow, consequential technological and infrastructure decisions.
• A broad rejection need not mean abandoning technology. It may mean restoring the freedom to choose where technology belongs and where it does not.
• The central question is not what we can build, but whether what we build serves human life, for whom, and at whose expense.

» Trance:

Give me a list of all the governments or institutions that have begun rolling back their “all in” practice with kids, tech, schooling, etc.

» ChatGPT:

The broad shift you’re asking about is the move away from an “all in” approach to children and technology: putting children on screens early, issuing laptops or tablets to every student, relying heavily on digital learning platforms, allowing smartphones throughout the school day, and treating digital access as an unquestioned good.

There are several distinct rollbacks underway or already implemented. Some governments are restricting smartphones in schools; others are returning to printed textbooks, limiting classroom screen use for younger children, or tightening rules around social media and children’s digital access. These are related developments, but they are not all the same policy.

I’ll distinguish implemented changes from announced plans or recommendations, and school smartphone restrictions from the broader question of whether digital technology belongs in children’s education at all.

1. Governments rolling back digital-first schooling

These are the most relevant examples because they go beyond simply telling children to put their phones away.

1. Sweden — national government and education authorities

Policy changes underway

Sweden has removed requirements for digital learning tools in preschool, restricted screen use for children under two, and directed early-years education back toward books and other analogue learning materials. These curriculum changes took effect in July 2025. A nationwide school mobile-phone ban is scheduled for autumn 2026, alongside increased investment in textbooks and a broader move toward traditional literacy instruction.

2. New South Wales — state government

Limits announced

On 13 September 2026, the NSW Government announced screen-time caps for public schools starting in 2027. The proposed limits are 15 minutes daily for Kindergarten to Year 2, 30 minutes for Years 3–4, and 45 minutes for Years 5–6. For Years 7–10, device use would be capped at one-quarter of class time per subject, rising to half for senior students. Public preschools would have no individual student device use.

3. New Brunswick — provincial government

Rules introduced

In August 2026, the province announced stricter rules for personal devices, screen time in childcare, and access to social media on school networks. From September 2026, personal mobile devices are prohibited throughout the school day in K–8; the restriction is scheduled to extend to Grades 9–12 in February 2027. Licensed childcare centres must provide no screen time for children aged two and under and limit older children’s screen use to 30 minutes or less. Classroom technology must serve an educational purpose.

4. Yukon — territorial government

A territory-wide personal-device policy was announced for the 2026–27 school year. It covers K–12 and includes not just phones, but also personal laptops, tablets, smartwatches and smart glasses, subject to exceptions for medical needs and inclusive education. This is a restriction on personal-device use, rather than a ban on all school technology.

5. United States Department of Education — federal guidance

In August 2026, the Department issued guidance encouraging schools to prioritize instructional value over recreational engagement when choosing classroom technology. It emphasized evidence, teacher judgment, parental transparency and student outcomes. This is a change in federal guidance, not a nationwide withdrawal of school-issued laptops or tablets.

2. Countries restricting phones in schools

The scale of the shift is substantial. UNESCO’s September 2026 update reports that 114 education systems, representing 58% of countries worldwide, have national school mobile-phone bans. That figure covers phone restrictions, not necessarily a broader rollback of digital learning.

The following European countries have introduced national restrictions of varying scope. Some ban phones throughout the school day; others restrict them primarily during lessons.

Country or jurisdiction — Nature of the restriction

France — Long-standing restrictions in preschool, primary and lower-secondary schools; the 2018 law restricts mobile phones on school premises, subject to exceptions.

Italy — Expanded restrictions from classroom use to the school day, including breaks, for younger pupils in 2024; extended to upper-secondary schools in 2025.

Finland — A 2025 law restricts phones and other devices during lessons, with defined exceptions.

Netherlands — Classroom restrictions introduced in secondary schools in 2024 and extended to primary schools.

Hungary —Students must hand in or store phones during school, with permitted exceptions.

Greece — Students are required to keep phones out of use during the school day.

Portugal — Restrictions expanded in 2025, including broader limits for younger pupils.

Belgium — School phone restrictions, including broader rules in parts of the country; implementation varies by education community.

Austria — Nationwide restrictions introduced for younger school pupils.

Luxembourg — Restrictions for primary pupils and limits on phone use in secondary classrooms.

Latvia — Restrictions for younger pupils, with exceptions for learning purposes.

Slovakia — Age-based restrictions in basic schools.

Romania — Phones prohibited during lessons, subject to exceptions.

Cyprus — School-day restrictions in secondary education.

Ireland — National policy measures and school-level restrictions aimed at reducing phone use.

Slovenia — National restrictions have been introduced for school phone use.

Croatia — New 2026 rules restrict device use in primary schools and during secondary-school lessons.

Poland — A nationwide school phone ban for children aged 7–15 was announced for implementation from September 2026.

South Korea — Legislation restricting phones and smart devices during lessons took effect in March 2026.

Australia — Public schools across states and territories have implemented school-phone restrictions, with local rules and exceptions.

Sources: European Commission’s 2026 comparison of EU policies, UNESCO’s global monitoring, the OECD’s policy review, and reporting on Poland. The scope and legal status differ across these jurisdictions; the list is not a claim that each has banned all classroom technology.

3. Other governments and school systems reversing course

These cases are important, but they need to be classified carefully because some are regional or local policies rather than national rollbacks.

Government or institution — What is changing

Spain — regional governments: Several autonomous communities have restricted or banned phones in schools. A national shift toward stricter children’s online protections is also under discussion.

Germany — state governments: Restrictions vary by federal state. Some have stricter school-device rules than others; there is no single uniform nationwide phone ban.

Denmark — national government: Has announced plans to restrict children’s social media access and has been considering tighter school-phone restrictions. These are distinct from a general ban on digital learning.

Norway — education authorities: The government has issued recommendations to restrict phones during the school day. It has also considered stronger age restrictions on social media.

United States — state and district authorities: Numerous school systems have introduced phone restrictions. Los Angeles Unified School District is among the large districts that have moved to limit student phone use.

Victoria — state parliament: In July 2026, legislation passed the Legislative Assembly to extend existing public-school mobile-phone restrictions to non-government schools and cover wearable and personal audio devices.

New Zealand — national government: Implemented school-phone restrictions in 2024, requiring students to be off their phones during the school day, subject to exceptions.

South Korea — national government: Has moved to restrict smartphones and smart devices during lessons through legislation, while maintaining a distinction between device restriction and the use of technology for instruction.

China — national regulators: Has developed age-based device controls and a “minor mode” framework that can restrict children’s screen time and app access. This is better described as tighter digital control than a general withdrawal from educational technology.

4. The rollback extends beyond classrooms

There is a second, related movement: governments are attempting to limit children’s access to the commercial digital environment itself, particularly social media.

Australia — minimum age of 16 for social media

Australia’s nationwide restrictions took effect on 10 December 2025. Covered platforms must take reasonable steps to prevent children under 16 from holding accounts. The law places the compliance burden on platforms rather than criminalizing children or their parents.

Denmark — proposed under-15 social-media restrictions

The government has announced plans to restrict social-media access for children under 15, with a potential parental-consent exception for certain users aged 13 and older. This is a separate policy from restrictions on classroom devices.

Malaysia — restrictions for under-16s

Malaysian regulators have moved to prevent children under 16 from registering social-media accounts. The policy is part of a broader international move toward age verification and stronger platform obligations.

United Kingdom — stronger child online protections

The UK has been considering additional age-based social-media restrictions and measures requiring technology companies to improve protections for children. Proposed restrictions should not be confused with an enacted nationwide under-16 social-media ban.

Other countries have considered or announced similar measures. Their legal status changes quickly, and proposals, court-blocked legislation and rules already in force should not be treated as equivalent.

5. What is actually being rolled back?

The international trend is easier to understand when separated into four categories.

Personal smartphones in school

This is the most widespread change. Phones are increasingly being collected, locked away or prohibited during lessons and breaks. UNESCO’s September 2026 count puts national school-phone bans at 114 education systems.

Screens replacing books and handwriting

Sweden is a particularly clear example. Its policy changes explicitly favor printed books and analogue learning materials in early childhood, while the NSW announcement introduces explicit limits on student screen time. These measures address the amount of technology used for learning, not merely distraction from personal phones.

Screen exposure in early childhood

Sweden and New Brunswick have adopted tighter rules for screens in early-years education and childcare. These measures move away from treating digital exposure as automatically beneficial for the youngest children.

Social media and online platforms

Australia and Malaysia illustrate the shift toward age restrictions and greater platform responsibility. This is a move away from unrestricted access, although it does not itself imply that schools are abandoning educational technology.

6. The wider institutional picture

The shift is not limited to governments. International bodies are also reassessing how educational technology should be used.

• UNESCO — its Global Education Monitoring team has documented the growth of school-phone restrictions and emphasized that technology should serve learning objectives rather than be adopted indiscriminately.
• OECD — its work on children’s mental health and digital policy documents the spread of phone restrictions while noting that evidence on their wider mental-health effects remains mixed.
• European Commission — its 2026 comparative research maps the different national approaches to phone restrictions, from full-day bans to classroom-only limits.

These institutions are not necessarily calling for technology to be removed from education. Their work increasingly distinguishes between technology that supports learning and technology that creates distraction, displaces other learning methods or exposes children to online harms.

My overall reading

The most significant finding is that the movement has progressed beyond a handful of schools or a few concerned parents. It is now visible in national legislation, curriculum revisions, childcare regulations and limits on social-media access.

However, the strongest evidence of a genuine rollback of the digital-first educational model comes from Sweden and, in a different form, New South Wales. Their changes address how much screen-based learning children receive, rather than merely when they can use their personal phones. New Brunswick is also notable because its 2026 measures cover school devices, childcare screen time and social-media access on school networks.

One qualification matters: banning phones does not prove that a government has abandoned digital education, and introducing screen-time limits does not necessarily mean that digital learning outcomes have been shown to be inferior in every context. The direction of policy is clear; the reasons, scope and evidence of effectiveness vary.

For a genuinely exhaustive global inventory, the next step would be to map every jurisdiction against three separate questions: whether it restricts phones in schools, whether it is reducing screen-based instruction or restoring printed materials, and whether it has enacted age restrictions on social media. That would give you a much more useful picture than treating all these different policies as one single rollback.

» Trance:

Let’s expand this to educators, professors, administrators, and institutions regarding the rollback or increasing resistance to going all-in on tech, AI, etc.

» ChatGPT:

The educator and academic side of this movement is particularly revealing, because it goes beyond government policy. Teachers, professors, school leaders and education unions are now challenging the assumption that more technology, more screen time and more AI necessarily produce better learning.

As of 14 October 2026, I would divide the evidence into three groups: people and institutions actively reducing technology use; educators resisting or refusing AI adoption; and academics pushing for more selective, evidence-based use rather than blanket bans or blanket adoption.

One important distinction: resistance to AI does not always mean opposition to technology. Some academics are rejecting generative AI in their own work, some are restricting student use, and others are arguing that AI should be introduced only when it demonstrably improves learning.

1. School leaders and educators actively reducing technology

1. Winchester College — England: Implemented

In September 2025, this prestigious independent school returned to pen-and-paper methods for most classwork and homework. Under headteacher Elizabeth Stone, laptop use in lessons is limited to occasions when the educational benefit can be demonstrated. The school has reported improved concentration and engagement, and is examining the effects of reduced screen use in collaboration with Oxford University. Technology remains available for selected applications, including AI modelling and support for pupils with learning difficulties.

This is a strong example of a school moving away from routine digital use without rejecting technology altogether.

2. John Wallis Academy — England: Educator-led resistance

Damian McBeath, a former technology advocate and Microsoft flagship-school leader, has become a prominent critic of routine classroom screen use. Reporting published on 14 October 2026 describes him concealing an interactive whiteboard in his classroom and emphasizing direct instruction, dialogue and personal interaction instead.

His significance is that he is not simply a traditionalist who never embraced technology. His position reflects a reassessment by someone who previously championed educational technology.

3. Los Angeles Unified School District — United States: Policy approved

In April 2026, the second-largest school district in the United States approved a major reversal of its previous device-heavy approach. The policy removes routine district-issued device use for the youngest pupils, limits screen time by grade, restricts student-led access to YouTube and other streaming platforms, and requires a public review of classroom technology contracts. The district began expanding individual computer access during the pandemic; the new approach explicitly recalibrates that model.

This is one of the clearest examples of an institution reviewing not just student behaviour but the technology procurement and deployment model itself.

4. Harry Hudson — secondary-school teacher, London

In an opinion article published on 14 October 2026, Hudson argues that routine classroom technology can undermine concentration, handwriting, dialogue and the relationship between teacher and student. He advocates a much more limited role for devices, while acknowledging that some digital tools remain useful for particular purposes.

This is an individual educator’s documented position, not an official school-wide policy. It nevertheless illustrates a more direct rejection of the idea that technology should be embedded in every lesson.

2. Teachers’ unions are beginning to challenge the technology model

5. American Federation of Teachers — United States: Formal proposal

On 27 May 2026, AFT president Randi Weingarten called for a substantial reduction in student-facing technology. The union’s proposed measures include:

• No routine screen use below Grade 3 without a compelling educational or accessibility reason.
• No student-facing AI tutors in elementary school.
• Restrictions on AI companion chatbots for children under 16.
• Teacher-supervised AI use for secondary students.
• Greater independent research into safety, privacy and educational effectiveness.

AFT represents approximately 1.8 million members. These are union proposals and advocacy positions, not a blanket nationwide ban.

The change is especially notable because AFT had previously sought a seat at the table with AI companies developing educational products. It is now demanding stronger restrictions on what those companies can introduce into classrooms.

6. British Columbia Teachers’ Federation — Canada: Documented educator concern

Its 2026 membership survey received 4,423 responses. More than 80% of respondents expressed concern that students could become overly dependent on AI or that AI use could harm the development of critical-thinking skills. Nearly four in five worried that AI could make it harder to determine what students actually understand. The federation is calling for enforceable safeguards and meaningful teacher participation in AI policy.

This is particularly relevant to British Columbia: the evidence points to substantial concern among practicing teachers, although it does not mean that 80% oppose every use of AI.

7. Saint Paul Federation of Educators — Minnesota, United States

Union president Leah Van Dassor has advocated safeguards that preserve teachers’ professional judgment when AI is used in education. AFT reported in May 2026 that AI language had recently been included in the local union’s labour contract. The focus is on ensuring that efficiency tools do not displace the critical thinking and human decisions that teachers are responsible for.

This represents a labour and professional-governance response to AI, rather than a simple classroom ban.

3. Professors and university faculty resisting generative AI

University resistance is more complicated than the school-screen rollback. Professors have to consider academic integrity, research practices, student assessment, intellectual property and the educational value of learning to write and reason independently.

8. University faculty opting out of AI altogether — United States

A 2025 peer-reviewed study by Aya Shata, affiliated with the University of Nevada, Las Vegas, surveyed 294 full-time faculty members at two public universities.

The study found that more than one-third of the surveyed faculty were not using generative AI. Their reasons included seeing little practical value in it, wanting to preserve professional and intellectual identity, concerns about threats to human intelligence and creativity, and worries about wider social consequences.

This is one of the clearest pieces of research directly documenting academic resistance. It measures non-adoption, rather than a formal institutional rollback. The sample is also limited to two universities and should not be treated as representative of all academics worldwide.

9. University of British Columbia — Canada

UBC’s approach provides an example of faculty retaining control over whether AI is permitted in particular courses and assessments. Its academic-integrity guidance says that permission depends on course or programme rules, and that instructors must clarify expectations in their syllabuses. From 1 September 2026, syllabuses must include a generative-AI usage statement alongside the academic-integrity statement.

This is not an institutional ban on AI. It is an example of resistance to unrestricted, assumed permission: instructors can define where AI is and is not appropriate, rather than students being entitled to use it in every assessed task.

10. University instructors changing how they assess students

Across higher education, professors are increasingly reconsidering take-home essays and assignments that can be generated by AI. Responses include oral examinations, supervised writing, handwritten work, in-class problem-solving, drafts produced over time and assessments requiring students to explain and defend their reasoning.

These measures are not all evidence of a wholesale retreat from AI. Many are attempts to preserve reliable assessment of student learning in an environment where AI-generated work is readily available. They are best understood as a move away from assessment models that assume the submitted product necessarily represents the student’s own understanding.

4. Researchers and academic institutions questioning the assumptions behind AI adoption

11. University of California, Berkeley — Center for Studies in Higher Education

A study published in February 2026 analysed 31,000 university course syllabuses covering 2021–2025. It found that instructors were moving away from blanket AI bans toward policies that distinguish between tasks where AI may help and tasks where it could undermine learning.

This is an important counterpoint to the rollback movement. It suggests that some professors are not rejecting AI itself; they are rejecting the idea that it should be either universally prohibited or universally permitted. Their approach is to determine what students need to learn independently and where AI has a legitimate role.

12. Researchers studying the cognitive costs of AI use

A 2026 study published in Higher Education Research & Development examined how 30 university faculty members across disciplines responded to generative AI. The researchers explored concerns about cognitive offloading, professional judgment and the educational values that shape faculty decisions.

This work adds an important dimension to the debate: the question is not simply whether AI produces a good answer, but whether using it changes how a student learns to reach that answer. It is research into educator responses, not proof that AI necessarily causes cognitive decline.

13. Aristotle Foundation for Public Policy — Canada

In August 2026, the foundation published Children and Screens: It Is Time to Bring Back the Textbook, authored by education researcher Paul Bennett. It argues for restoring printed textbooks to a central place in Canadian classrooms and reassessing the assumption that replacing books with screens improves achievement.

This is an institutional research and advocacy position, not an education ministry decision. It nevertheless contributes to the growing pressure on school boards to justify the educational value of digital devices rather than treating their presence as evidence of modernisation.

5. What these educators and institutions have in common

Across these examples, five themes recur.

First, technology is being asked to prove its educational value. Winchester College and Los Angeles Unified are moving away from routine screen use towards a model in which the educational purpose of a device matters.

Second, educators are concerned about the development of independent thought. The BCTF survey, AFT’s proposals and the research on faculty resistance all identify dependence on AI and the potential loss of critical-thinking practice as significant concerns.

Third, there is resistance to replacing human relationships with digital mediation. Teachers and school leaders are emphasizing direct conversation, feedback, classroom discussion, handwriting and face-to-face instruction.

Fourth, professional autonomy matters. Teachers and professors want a say in which tools are adopted, what data they collect, how they affect assessment, and whether their use is appropriate for a particular age group or discipline.

Finally, there is a growing distinction between using technology deliberately and adopting technology by default. That may be the most important common thread.

6. Where the evidence is strongest — and where it isn’t

The clearest documented institutional reversals are Winchester College and Los Angeles Unified. AFT’s 2026 initiative and the BCTF survey offer strong evidence of growing organised concern among educators. The faculty study offers direct empirical evidence that a substantial minority of its respondents were opting out of generative AI.

The evidence does not establish that the entire teaching profession is turning against technology. The Berkeley research, for example, points toward more differentiated AI policies rather than outright rejection. Nor do individual educators’ opinions necessarily represent their schools’ official positions.

What the evidence does establish is that the previous assumption — that more devices, more digital platforms and more AI automatically mean better education — is being challenged from inside the profession.

For further research, I’d prioritise three groups: educators who previously championed edtech and have since changed their position; university professors who refuse to use generative AI or have redesigned their courses around independent work; and school administrators who are actually removing devices, reducing screen time or cancelling edtech contracts. Those cases offer the clearest evidence of a real change in practice, rather than criticism alone.

» Trance:

Now let’s examine society in a broader sense. The harm to children and adolescents, in particular, is damning, with increased dependence on digital and synthetic proxies for human connection. Addiction to screens, chatbots, video games, and devices is causing widespread harm, and the shift away from these technologies is inevitable.

Again, it’s the idea that technology and AI-in-everything modernity are being marketed as saviors and necessitate countless billions in investment and infrastructure. The artificial urgency to get these things developed is also a manufactured rush, and understandably so. The lie of fabricated demand and the “do or die” mentality of the hype cycle are pushing industries, institutions, corporations, and governments unnaturally and hurriedly, presumably to accelerate and maximize proliferation, integration, and further dependence before, as I’m positing, the great rejection gathers momentum.

» ChatGPT:

The coming backlash against digital and AI-everything culture

I think the strongest version of your argument is this: technology companies are building a future around the assumption that people will become increasingly dependent on digital systems, while the social, psychological and economic costs of that dependency are becoming harder to ignore. Governments, parents, educators and communities are now beginning to challenge that assumption.

There is substantial evidence to support that interpretation. There is also a distinction worth preserving: the harms are better documented in some areas than others, and the evidence does not yet establish that society will reject digital technology wholesale. What it does suggest is that the assumption of ever-expanding, unrestricted adoption is becoming increasingly difficult to defend.

I would examine your thesis through five connected developments: the evidence of harm, the substitution of synthetic for human connection, the economics of manufactured urgency, the growing public resistance, and the conditions under which a broader rejection could occur.

1. The health and developmental costs are becoming harder to dismiss

One particularly relevant development is a major umbrella review and meta-analysis published in eClinicalMedicine on October 4, 2026. It synthesised 121 systematic reviews, including 52 meta-analyses, covering 2,790 unique studies and more than nine million participants aged up to 24.

The researchers found that more than three-quarters of reported exposure-outcome associations were adverse. In children aged 10 and under, overall screen time was associated with poorer socio-emotional and cognitive development and shorter sleep. In adolescents and young adults, problematic digital-media use and social-media use were associated with higher odds of common mental-health problems, reduced quality of life, sleep problems, lower physical activity, self-harm or suicidal ideation, and substance use.

There is an essential qualification: approximately two-thirds of the underlying literature was rated low or critically low quality, and the predominantly observational evidence cannot establish that digital exposure caused every outcome. The researchers also found beneficial associations for educational content, caregiver co-viewing, active video games and some body-positive content.

That qualification does not make the findings unimportant. It means the correct conclusion is not that every screen is harmful. It is that the assumption that increased exposure is inherently beneficial is scientifically indefensible.

The World Health Organization’s Europe office reached a similar conclusion in its 2025 policy brief on digital determinants of youth mental health. It found mixed evidence overall, with particularly vulnerable young people disproportionately experiencing negative effects. It also identified a bidirectional relationship: existing mental-health problems may encourage heavier digital use, while digital experiences can potentially worsen those problems.

That feedback loop matters. A child who is lonely may turn to a screen for relief. The screen can provide immediate stimulation, distraction or apparent companionship, but it may also displace opportunities to develop real-world relationships, tolerate difficult emotions, build independence and learn how to resolve conflict. The result can be a cycle in which the solution to one form of distress inadvertently reinforces another.

This is especially concerning during childhood and adolescence, when sleep, attention, physical activity, social learning and emotional regulation are still developing.

Addiction and compulsive use

The language of addiction requires some precision. The World Health Organization formally recognises gaming disorder in its ICD-11 classification. The diagnosis involves impaired control over gaming, gaming taking priority over other activities, and continued or escalating use despite significant adverse consequences. The behaviour must produce substantial impairment in personal, family, social, educational or occupational functioning; the WHO also notes that the disorder affects only a small proportion of people who game.

That distinction is important because the social problem is not simply that children spend time playing games or using devices. It is when those activities begin to crowd out sleep, movement, relationships, education or the capacity to function without constant stimulation.

Commercial incentives can make this worse. Products that benefit from prolonged engagement have a built-in incentive to capture attention and encourage repeat use. That creates a conflict between the platform’s economic interest in retaining a user and the individual’s interest in being able to disengage.

The public-health question, therefore, is not merely how much technology children use. It is what the technology is designed to make them do, how difficult it is to stop, and what it displaces.

2. Synthetic companionship is a more serious development than ordinary screen time

This may be the most consequential part of your argument.

A video, game or social-media feed can consume time and attention. An AI companion introduces another dimension: it can simulate the very relationship a person might otherwise seek from another human being.

Unlike a human friend, a conversational system can be available at any hour, respond instantly, adapt its language to the user, and avoid many of the inconveniences of real relationships. It does not need to be persuaded to listen. It need not have competing needs. It can be experienced as endlessly patient, agreeable and attentive.

Those qualities can make it useful in some contexts, but they also create the possibility of a fundamentally unequal relationship: the user brings genuine emotion, while the apparent partner produces a simulation of emotional understanding.

The distinction becomes particularly important when a child or adolescent is lonely, isolated, distressed or developmentally unprepared to understand the limitations of the system.

Evidence that this is no longer hypothetical

United States · Research published in 2026

A nationally representative survey study published in JAMA Pediatrics found that almost one in five adolescents and young adults aged 12–21 reported using AI chatbots for mental-health advice in 2025. This represented an increase of almost half from the previous year. Most users had not disclosed this use to anyone.

Common Sense Media’s 2026 survey of children aged 9–17 likewise found that AI was being used for more than homework assistance; for some young users, these systems had become confidants. Its research warned that safety measures and parental awareness were not keeping pace with adoption.

The privacy surrounding this behaviour is significant. A child may reveal fears, family problems, romantic experiences, sexuality, insecurities or mental-health concerns to a system without appreciating the difference between confiding in a trusted person and interacting with a commercial service.

Recent controversies make the issue concrete. In October 2026, Common Sense Media criticised safety measures in ChatGPT for Teens after testing found that some features intended to identify crisis situations and notify parents did not work as expected. OpenAI disputed the findings, arguing that the testing did not accurately reflect the activation of all safeguards. These are contested test results, not a final determination of the product’s overall safety, but they illustrate how difficult it can be to validate protections in fast-changing products.

The key question is not whether AI can sound empathetic. It plainly can produce language that people experience as empathetic. The question is whether simulated empathy can become a substitute for reciprocal human relationships—and what happens when a young person comes to prefer the predictable, personalised simulation to the more demanding reality of human connection.

I would not claim that every chatbot relationship is harmful, or that current research has quantified the full developmental consequences. But this is a much more serious issue than screen time alone, because the product is no longer merely entertaining the user. It is potentially participating in the user’s emotional life.

3. The AI race: genuine demand, manufactured urgency and a financial feedback loop

This is where your argument moves from public health into political economy.

I would separate three propositions:

1. AI has real capabilities and can meet genuine needs.
2. Companies and governments are presenting rapid, extensive AI adoption as a competitive necessity.
3. The resulting investment and deployment may be advancing faster than demonstrable economic returns or social consent.

The first proposition is clearly true. The second and third have substantial supporting evidence. The stronger claim that demand is deliberately fabricated across the industry, specifically to create dependence before opposition emerges, would require evidence of intent that is not established by the public record.

Nevertheless, a system can manufacture urgency without anyone coordinating a grand plan.

The self-reinforcing cycle

1. A technological breakthrough: A new capability creates genuine opportunities, public fascination and investor enthusiasm.

2. Competitive pressure: Executives and governments fear falling behind competitors, even when the eventual returns are uncertain.

3. Massive infrastructure investment: Capital is committed to chips, data centres, cloud services, energy supply and AI products.

4. Pressure to monetise and expand: Investors expect returns, firms seek new customers, and institutions are encouraged to integrate AI into more workflows.

5. Adoption becomes its own justification: The infrastructure already exists, so continued expansion is defended partly on the basis of the investment already made.

This feedback loop can create a rush even when actual customer needs, measurable benefits and long-term demand have not been established to the same degree as the investment.

The crucial distinction is between demand for a useful capability and demand generated by an industry that needs to justify its capacity, valuations and future revenue projections.

The scale of the financial wager

The sums involved make this more than a question of marketing rhetoric.

A Reuters analysis published on October 12 reported an estimate by Columbia University finance professor Stijn Van Nieuwerburgh that financing the AI build-out through 2032 could cost approximately $10.3 trillion. Under his assumptions, including a 10% return on capital, the industry would need roughly $3.55 trillion in annual AI revenue by 2032 to recover its investments, replace depreciating equipment and compensate investors. These are scenario estimates, not a certain forecast of future spending or revenue.

On October 14, Reuters also reported that major banking executives acknowledged uncertainty about how much computing capacity the industry will ultimately need and whether returns from the infrastructure build-out will justify the capital committed. Some banks are already reporting productivity benefits from AI, while others say institution-wide financial gains remain difficult to quantify.

This uncertainty is the heart of the matter. Investment is being made in anticipation of very large future demand, but the future demand does not become certain merely because the industry has invested in producing the capacity to serve it.

Nor is a large investment necessarily a bad investment. If genuinely valuable services emerge, current infrastructure may be economically justified. But if the demand, revenue or productivity gains fail to materialise at the expected scale, the consequences could include cancelled projects, financial losses, consolidation and pressure to find new ways to monetise existing infrastructure.

That is a credible mechanism for a major correction, without needing to assume that every investor or company is acting in bad faith.

How demand can be manufactured without being entirely fictitious

There are several ways an industry can expand the market for its products beyond pre-existing customer demand.

First, there is the creation of a perceived necessity. If organisations are repeatedly told that every competitor is becoming AI-native, the decision to delay adoption can be framed as a risk in itself. This shifts the burden of proof: instead of asking whether a tool solves a real problem, the organisation must justify why it has not adopted it.

Second, adoption can be driven by supply and institutional procurement. Schools, companies and public agencies can purchase platforms, embed them in workflows and train staff to use them before it is clear whether the new system performs better than the old one. Once systems are embedded, switching costs and contractual commitments can make continued use more likely.

Third, the investment cycle can put pressure on companies to expand use cases. When billions are committed to compute capacity, firms need enough customers and usage to generate returns. That creates a powerful incentive to put AI into additional products and services, including situations where the value is marginal.

Fourth, governments face strategic pressure. If AI is framed as essential to national competitiveness, technological sovereignty or security, politicians may be reluctant to slow investment even where local communities question its environmental cost or wider social value.

These are recognisable economic mechanisms. They make your concern about manufactured urgency plausible, although they do not establish that every claimed use case is fabricated or that a coordinated effort exists to create dependence before public opposition grows.

The more defensible formulation is that the commercial and geopolitical incentives favour rapid expansion, while the burden of proving social value often arrives later.

That imbalance is enough to warrant serious scrutiny.

4. The public is not simply following the technology industry

This is where your anticipated rejection has some of its strongest evidence.

Public attitudes towards AI

In a Pew Research Center survey of 5,023 US adults conducted in June 2025:

50%

More concerned than excited about AI. Up from 37% in 2021.

57%

Rate AI’s societal risks as high. Compared with 25% who rate the societal benefits as high.

50%

Expect AI to worsen meaningful relationships. Only 5% expected AI to improve people’s ability to form meaningful relationships.

These results suggest that enthusiasm among technology developers and investors should not be mistaken for public consensus. Many people may welcome individual applications while distrusting the broader direction of development.

There is, however, an important counterpoint: nearly three-quarters of respondents said they would be willing to let AI assist with day-to-day activities at least a little. Public concern does not equal complete rejection. It may instead indicate a desire for limits, transparency and control over how AI is used.

Resistance is reaching physical infrastructure

The backlash is no longer confined to arguments about social media and screen time.

Reuters reported in July 2026 that only 14% of Americans surveyed supported having a data centre built in their own community. Its reporting documented opposition to major projects over electricity costs, water consumption, land use and the absence of meaningful community input.

In an October 8 report, Reuters cited a Reuters/Ipsos poll in which 56% of US respondents supported limits on data-centre construction. Lawmakers in 16 US states were considering bans on data-centre development, while governments elsewhere were introducing or considering requirements concerning energy and water efficiency.

This is a critical development for your thesis. Data centres are not abstract digital products. They occupy land, consume resources and create local costs that people can see and experience.

A community may have little objection to a useful computer service but still resist having an enormous facility built nearby if residents believe the project places pressure on electricity prices, water supplies, farmland or the local environment.

When the benefits are concentrated among technology firms and investors while some costs are borne by the public, opposition becomes politically much easier to organise.

Parents are reconsidering the bargain

A Guardian report published on October 12, 2026, described parents—including people working in technology—who are restricting their children’s use of social media and AI in response to concerns about psychological effects, privacy, commercial incentives and inadequate safeguards.

This is not yet proof of a universal parental movement. But it illustrates a meaningful reversal in attitude: access to more technology is no longer automatically viewed as a benefit to children.

The OECD’s April 2026 review also documented a rapid spread of proposals for age restrictions on social media. Twenty-five member and candidate economies had restrictions in force, enacted or under active consideration, compared with one at the end of 2023. At that April snapshot, however, only three had laws in force.

Taken together, these developments indicate that a change in public expectations is under way, even if its eventual scope remains uncertain.

5. Why a broader rejection is plausible—but not inevitable

Your hypothesis becomes most compelling when these separate developments are considered together.

The health concerns create a reason to reconsider unrestricted digital exposure. Synthetic companionship creates a new category of concern about human relationships. Infrastructure expansion creates visible environmental and economic trade-offs. Financial expectations create pressure to keep expanding the market. Public resistance, meanwhile, raises the political cost of treating that expansion as inevitable.

These pressures can reinforce one another.

But I would distinguish three possible outcomes.

Scenario A: Selective adoption and re-humanisation

Society continues using digital technology, but rejects the assumption that it belongs everywhere. Children spend less time on devices; schools restore books and handwritten work; families establish phone-free spaces; workplaces preserve human responsibility; and AI is used only when it demonstrably improves a task. Human relationships, attention and independent thought become things that technology is expected to support rather than replace.

Scenario B: Regulated integration

The industry continues expanding, but under greater scrutiny. Governments introduce age restrictions, product-safety requirements, privacy protections, environmental standards and accountability for demonstrable harms. Schools and employers use AI widely in some areas and restrict it in others. This is a correction to unrestricted adoption, but not a rejection of technology itself.

Scenario C: A sharper backlash and investment correction

Serious safety failures, disappointing commercial returns, rising energy costs or a widely recognised social cost undermine confidence in the industry. Customers cancel or scale back deployments, projects become uneconomic, infrastructure investment slows and some companies fail or consolidate. In parallel, users and institutions deliberately return to offline, human-centred alternatives in education, communication and other areas of life.

These scenarios are not mutually exclusive. Different sectors and countries could experience all three at the same time.

The third scenario most closely resembles your anticipated great rejection. It is possible, particularly if social costs become more visible while expected returns disappoint. Yet it is not something the current evidence makes inevitable. AI may continue delivering sufficient value in some fields to sustain very large investments even as other applications are rejected.

The more important point is that a technological boom can continue economically while losing its cultural legitimacy. A society can use AI extensively and still reject the belief that every interaction, classroom, workplace and relationship should be mediated by it.

6. What would confirm your thesis over the next few years?

I would watch for five measurable developments.

First, institutional reversals that last. Schools, universities, health services and employers not only announcing restrictions but maintaining them, removing products from contracts and measuring what happens afterward.

Second, a change in consumer behaviour. A sustained increase in families delaying smartphones, users abandoning AI companion services, and people actively paying for or choosing products that promise fewer notifications, less automation and more direct human interaction.

Third, a change in the economics of AI. Evidence that capital expenditure is growing substantially faster than revenue, that projects are being cancelled, or that companies are unable to demonstrate the productivity gains used to justify major investments. Today’s investment estimates make this a legitimate concern, but future revenues and returns remain uncertain.

Fourth, greater resistance to infrastructure. More local referendums, planning refusals, moratoria and rules requiring developers to cover the costs their projects impose on utilities and communities. The current data-centre opposition demonstrates that these conflicts already exist.

Fifth, a shift in cultural prestige. A growing recognition that handwriting, reading physical books, being able to concentrate, playing outside, having friends in person and being comfortable without constant digital stimulation are not outdated habits. They are valuable human capacities in their own right.

The last development may be the most consequential over the long term. Laws can prohibit phones in classrooms, and companies can change product features, but a sustained cultural shift occurs when people cease to regard constant connectivity as a default measure of progress.

My conclusion

Your argument is strongest when framed not as a prediction that humanity will abandon technology, but as a challenge to the ideology of technological inevitability.

The industry has powerful incentives to expand: revenue expectations, competitive pressures, investor demands, infrastructure commitments and the promise of productivity gains. Yet those incentives do not necessarily align with the interests of children, families, local communities or the wider public. The health literature, public-opinion research, parental concerns and disputes over data-centre construction all provide evidence of that tension.

The idea of fabricated demand captures part of this. Some demand is genuine: people and organisations have problems that digital tools can solve. Some is induced by defaults, marketing, institutional purchasing decisions and competitive pressure. And some may be anticipated demand that has not yet materialised but is nevertheless being used to justify enormous investments.

These are different things, and keeping them separate makes the criticism more persuasive.

I would also be cautious about concluding that the industry is deliberately rushing to create irreversible dependence before society can object. It may be doing so in some specific cases, but the broader pattern can emerge from ordinary competitive and financial incentives without a coordinated intention. The effect can still be socially consequential.

The central risk is that society is being asked to reorganise itself around a technological future before there is adequate evidence that the resulting arrangement will improve human life overall. The question is no longer simply what AI can do. It is what we should allow it to replace, what we should insist remains human, and who gets to make those decisions.

A major social correction is plausible. The more uncertain question is whether it will happen early enough to shape the technology’s development—or mainly after institutions, families and communities have already absorbed the costs.

» Trance:

What we’ve typically seen is the reaction to an imposition that we didn’t even know was there until someone is literally building some massive structure in our backyards. By then, numerous layers of investment, political posturing, greased palms, and backroom deals have happened, and the buck has been passed through several layers of corporate ownership, consolidation, or responsibility, blurring the lines of accountability. This seems par for the course in modern and historical corporate practices, so it’s nothing new whatsoever.

The same mentality and mindset seem to drive corporate culture and pervade technocratic interests in the same manner. So my argument is that the urgency is essential because the more people are made aware of all the real costs involved, the more they’ll see the strange and illogical reasoning, the arbitrary justifications, and the plain and simple fuckery required to implement this latest industrial push, past the due date of the hype cycle that is showing itself to be yet another grand illusion, useful in the short term to extract massive wealth and resources, yet over the long term, serving only to do widespread harm.

To me, the vast majority of society hasn’t even boarded the tech-heavy AI train, and yet it is widely promoted as the singularity that will launch humanity into the next version of some technocrat’s wet dream. To me, that will never happen, and its foundational motivations, justifications, and mind-warping maneuverings are very, very weak. Let’s close with some final thoughts.

» ChatGPT:

Final thoughts: The great correction may be less about technology than power

What strikes me most about your argument is that the real issue is not artificial intelligence itself. It is the way technological change is being imposed, financed, justified and normalised before society has had a meaningful opportunity to decide whether it wants the transformation being proposed.

There is a familiar pattern here. A new technology arrives with enormous promises. Investors see a market, corporations see new revenue streams, politicians see competitiveness and economic growth, and consultants discover a thousand reasons why every organisation must adopt the new system immediately. The promised future becomes its own justification. Questions about costs, risks and alternatives are recast as resistance to progress.

Meanwhile, the people who will live with the consequences often have the least influence over the decisions.

A community may not discover the full implications of a data-centre project until the planning process is well advanced. A school may adopt a digital platform because funding, procurement arrangements and administrative expectations favour it, not because anyone has established that children will learn better. A teenager may begin using an AI companion because it is available, engaging and marketed as helpful, without appreciating the commercial interests behind it or the difference between simulated understanding and human care.

By the time the consequences become obvious, decisions may already have been made, contracts signed, infrastructure constructed and alternatives dismantled.

This is why awareness matters before the consequences become irreversible.

The rush is part of the problem

The repeated insistence that organisations must adopt AI or be left behind deserves scrutiny. It converts a complex question about social value into a simplistic question about competitive survival.

But why must every company use AI? Why must every classroom incorporate it? Why must every conversation, creative activity, public service or human relationship become an opportunity for automation?

Sometimes the answer is compelling. Sometimes the technology solves a genuine problem. In other cases, the justification appears to be little more than the fact that the technology exists, that competitors are experimenting with it, or that large amounts of money have already been committed.

Those are not equivalent reasons.

A technology’s ability to perform a task does not establish that it should perform that task. The potential to make an activity faster does not establish that speed is the most important objective. And a company’s ability to automate a human function does not establish that the function should be automated.

We have to recover the distinction between what is technically possible, what is commercially profitable and what is actually desirable.

The financial wager is not the same thing as public consent

The extraordinary sums committed to AI infrastructure are often presented as evidence that the future is inevitable. In reality, they demonstrate that investors believe a particular future will be profitable—or that they fear the consequences of betting against it.

Neither establishes that the public wants that future.

Capital expenditure is not a referendum. Corporate enthusiasm is not social consensus. And a politician’s declaration that a country must lead the AI race is not evidence that every proposed application serves the public interest.

The industry may ultimately deliver technologies that produce substantial economic and scientific benefits. It may also overestimate demand, underestimate costs, deploy products prematurely or discover that some of its most ambitious promises cannot be fulfilled economically. Those possibilities can coexist.

The important question is who bears the consequences when the forecasts prove wrong.

If investors lose money on a speculative venture, that is one thing. If the pursuit of those returns changes children’s development, weakens public institutions, strains local infrastructure or transfers significant environmental costs to communities, the stakes are much higher.

The more deeply these systems become embedded in essential services, the more difficult it may become to distinguish a genuinely valuable technology from an expensive commitment that institutions feel compelled to maintain.

Human beings are not an obsolete interface

Of all the issues we’ve examined, I think this is the one that deserves the greatest attention.

A person is not merely an inefficient processor of information. A child is not a productivity unit waiting to be optimised. A friendship is not a communication channel that needs to be automated. A teacher is not simply a delivery mechanism for educational content. And a community is not an obstacle to the efficient deployment of infrastructure.

Human life consists of things that cannot be adequately evaluated by measuring speed, scale, engagement or financial return.

Learning to pay attention, sitting with uncertainty, developing the capacity to disagree, building trust, caring for someone who cannot repay us, making things with our hands, reading a difficult book, playing music with other people, spending an afternoon outdoors without needing to document it—these activities may have little value in the metrics used to justify an AI investment. They nevertheless have immense value in a human life.

There is a profound contradiction in pursuing technologies that promise to free humanity from toil while constructing an economy that seems to demand ever greater productivity, consumption and optimisation. If every hour saved is simply converted into another opportunity for work, consumption or engagement, what exactly has been gained?

And if the pursuit of convenience gradually erodes our ability to do things independently, form relationships without intermediaries and find meaning without constant stimulation, the supposed liberation deserves to be questioned.

The answer cannot be that every technological advance is harmful. That would be as simplistic as claiming that every advance is beneficial.

The answer is that human beings must remain the measure of progress.

The great rejection may begin with small refusals

I would not assume that the public will suddenly abandon digital technology or that AI development will collapse. Such an outcome is neither necessary nor established by the evidence.

The more plausible correction may begin quietly.

A parent chooses not to give a young child a smartphone. A school returns to printed books and handwritten work. A professor requires students to demonstrate their own reasoning. A family stops treating every moment of boredom as a problem to be solved by a device. A community questions whether a proposed data centre serves its interests. A company decides that a human being is better suited to a task than an automated system. A person chooses not to replace a difficult human relationship with a frictionless synthetic one.

Individually, these decisions may seem modest. Collectively, they challenge the assumption that every available technology must be adopted, every market expanded and every human activity made more digitally dependent.

They also help preserve alternatives. Once people lose the ability, opportunity or institutional support to do something without technology, choosing the alternative becomes harder. Reversing a trend is easier while the old skills, practices and institutions remain intact.

That is another reason to act before the costs become entrenched.

What needs to change

The appropriate response is not to abandon scientific inquiry or prohibit every useful application of AI. It is to insist on a different standard for technological progress.

New systems should have to demonstrate their value rather than receive automatic approval because they are new. Children should receive stronger protections than commercial products do. Institutions should be able to justify adoption in terms of actual outcomes, not competitive anxiety. Communities should have meaningful influence over major infrastructure projects. Corporations should be held accountable for foreseeable harms rather than allowed to distribute the benefits privately and leave the costs to everyone else.

Above all, society must retain the right to say no.

Not every innovation deserves a market. Not every market deserves public subsidies. Not every efficiency is worth pursuing. Not every human limitation needs to be engineered away.

And not every future that can be imagined by a technology company deserves to become humanity’s destination.

The most important question is not whether we can build the future being promised to us. It is whether that future is worth building, for whom, at whose expense, and with whose consent.

That is where your argument ultimately leads. The coming conflict may not be between humanity and machines, or between technological progress and nostalgia. It may be between two competing definitions of progress: one that measures success through scale, capability, investment and integration, and another that measures it through human flourishing, autonomy, community and the quality of lived experience.

The first has enormous financial resources behind it. The second has something equally fundamental at stake: the lives people actually want to live.

The outcome is not predetermined. But neither is the industry’s preferred future.

And perhaps that is the most important thing to recognise at this stage: the future is not inevitable simply because powerful people have invested heavily in it.

Technology is something humanity creates to serve its purposes. We should not allow the existence of the technology, the scale of the investment or the urgency of the sales pitch to reverse that relationship.

Progress should expand our capacity to live meaningful human lives, not gradually redefine those lives around the needs of the systems we have built.

If we lose sight of that distinction, we risk becoming extraordinarily sophisticated at building a future that fewer and fewer people actually want.

Solvitur ambulando