The Boiling Frog Was Wrong: AI Cognitive Harm Arrives In Minutes

The room is unremarkable. A clean desk, a laptop, a printed sheet of arithmetic and short reading passages, and a pair of EEG electrodes resting against the scalp like patient fingers. The participant has just finished ten minutes of working through problems with the help of an AI assistant. The screen is closed. The chatbot is gone. A research assistant slides a fresh page across the desk and asks, politely, for the subject to answer the next set of questions alone.

The subject reads. The subject thinks. The subject, by every behavioural and neural measure the researchers can capture, performs measurably worse than a control participant who never touched the assistant. The effect is not subtle. It is there in the response latencies, in the error rates, in the EEG traces that show a dampened pattern of frontoparietal engagement which, ten minutes earlier, was healthy and robust.

That, in essence, is the claim of a multi-institution study widely reported across science media in April 2026, attributed to researchers from UCLA, MIT, Oxford and Carnegie Mellon, which proposes the first causal evidence that brief AI use is sufficient to produce immediate, measurable cognitive impairment in the unaided performance of equivalent tasks. The reporting has been breathless and the framing predictably apocalyptic, but the scientific stakes, if the finding survives replication, are genuinely large. Earlier work in this area had described a slow drift, a kind of boiling-frog dependency in which years of cognitive offloading might thin out a person's capacity to think for themselves. The newer claim is something different and arguably more disturbing: that the cost shows up in minutes, not years.

The distinction is not academic. If the harm is gradual, you can argue, with some plausibility, that informed adults using AI in the privacy of their own choices are merely making a long-term trade-off they are entitled to make. If the harm is acute, then the deployment of AI assistants in classrooms, clinical consulting rooms, courtrooms, contact centres and welfare offices, often without disclosure and almost always without anything resembling informed consent, looks rather different. It looks like a very large and largely unmonitored field experiment.

What does the evidence actually show? What can be defended, and what cannot? And once we are honest about both, who has the responsibility to act?

The Boiling Frog and Its Discontents

For the past three years, the dominant frame for thinking about AI and cognition has been the boiling frog, the apocryphal creature that fails to leap from a gradually heating pot. The framing made sense because the foundational evidence in cognitive science was itself longitudinal and slow.

Eleanor Maguire's work at University College London on the hippocampi of London taxi drivers, beginning in 2000, established that the brain regions used to navigate a complex city physically thicken with use. Subsequent imaging work, including a 2017 study in Nature Communications by Hugo Spiers and colleagues, suggested that turn-by-turn satnav use suppressed activity in the same hippocampal circuits. Capacity follows demand: ask the brain to navigate, and it grows the apparatus for navigation; ask it to follow instructions from a phone, and the apparatus quietens.

In 2011, Betsy Sparrow, then at Columbia, with Jenny Liu and the late Daniel Wegner of Harvard, published a paper in Science showing that participants who expected to look information up later remembered the information itself less well, but remembered where to find it. A 2024 meta-analysis in the journal Memory found the Google effect real but more modest than early coverage suggested.

Together, this literature painted a picture of slow, accumulative externalisation. Bit by bit, certain cognitive functions migrated from the wet hardware in the skull to the dry hardware in the pocket. The implicit settlement was that the costs were chronic and perhaps reversible if you put the phone down.

Generative AI complicated this picture, but for the first eighteen months of the consumer chatbot era the public discussion still defaulted to the chronic frame. Even Michael Gerlich's much-cited 2025 paper in Societies, which surveyed 666 participants and reported a strong negative correlation between AI tool use and critical thinking scores, was best read as a snapshot of ongoing erosion rather than a claim about acute injury.

Acute injury is what the newer reporting is now claiming. And acute injury, scientifically and ethically, is a different beast.

What An Acute Effect Would Have to Mean

To understand why the reported April 2026 finding has provoked the reaction it has, it is worth being precise about what an acute cognitive effect would, and would not, be. An acute effect appears rapidly after exposure and is measurable on a short timescale. A chronic exposure might gradually wear down an organ over decades; an acute exposure produces a measurable change within minutes or hours.

In the cognitive context, the equivalent claim is that ten minutes of AI-assisted maths or reading leaves a measurable footprint on the brain's ability to perform similar tasks unaided immediately after. The footprint, if it exists, is not memory loss in any everyday sense. It is more like a transient state shift, a cognitive tone that has slackened.

This is not biologically implausible. Cognitive psychology has long documented carry-over effects between tasks. Mental set, the tendency to apply a problem-solving strategy beyond its useful range, is a textbook example. So is the well-replicated finding that performing a task in a state of high external scaffolding can degrade subsequent independent performance, a phenomenon educators have long known as the assistance dilemma.

What the reported study would add is an EEG-level signal, that the brain is not merely behaving as if it has just been scaffolded but is in some quantifiable sense still in the scaffolded state, with reduced engagement in the networks that would ordinarily be doing the work. If that signal replicates, the implication is that AI use is not merely a labour-saving device whose benefits and costs balance out neatly. It is a state-altering one.

This is where the strongest existing evidence in the literature, the MIT Media Lab paper Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task, becomes essential context. Authored by Nataliya Kosmyna and seven co-authors and posted to arXiv in 2025 as preprint 2506.08872, the paper studied 54 participants in the Boston area aged 18 to 39, who wrote SAT-style essays under one of three conditions: with a large language model, with a search engine, or with no tools at all. EEG recordings during writing showed that brain-only participants exhibited the strongest, most distributed network engagement; search engine users showed moderate engagement; and LLM users displayed the weakest connectivity. Eighty-three per cent of LLM users were unable to quote from the essays they had just produced. In a fourth session, when LLM users were reassigned to brain-only writing, they continued to show weaker neural connectivity than the consistent brain-only group. The MIT authors called this carry-over cognitive debt.

The MIT preprint was not peer reviewed when it was posted, the sample was modest, and the authors themselves cautioned against the most sensational interpretations. But the basic shape of its finding, that there is a residual neural signature after the AI is taken away, is precisely the shape of the claim that the April 2026 reporting is now amplifying. The newer study, on the description circulating in the science press, extends the logic to elementary cognitive tasks rather than essay writing, and to far shorter exposures.

It is too early to know whether the April 2026 work will hold up under peer review and replication. It is not too early to ask what the world should do if it does.

Cognitive Offloading, Without Romance

The mechanism most often cited for both the chronic and acute findings is cognitive offloading: the use of external tools to reduce the demands on internal cognition. The concept predates the AI debate by decades. Writing things down is cognitive offloading. So is asking a colleague. Offloading reduces working-memory load and frees attention. Under certain conditions, it also reduces depth of processing, weakens encoding into long-term memory, and degrades the capacity to do the offloaded task without the tool.

What seems to be different about generative AI is the scope, the seamlessness and the ambient nature of the offload. A calculator does arithmetic. A search engine fetches documents. A large language model writes the paragraph, generates the answer, structures the argument and presents the result in finished form. The cognitive task it performs is not retrieval but synthesis, the very thing that, in classical accounts, is supposed to constitute the active work of thinking.

The Microsoft Research and Carnegie Mellon paper presented at CHI 2025 in Yokohama, authored by Hao-Ping Hank Lee and colleagues and based on a survey of 319 knowledge workers analysing 936 real-world AI-assisted tasks, gives this dynamic empirical shape in the workplace. The paper found that higher confidence in AI was associated with less critical thinking, while higher self-confidence in one's own abilities was associated with more critical thinking. The authors described a shift in the nature of cognitive work itself, from information gathering toward verification, from problem-solving toward integration of AI output, from doing toward supervising. They warned of what they called cognitive atrophy.

The proposed mechanism for an acute effect, then, is not mysterious. During AI-assisted work, the cognitive networks responsible for evaluating and integrating outputs remain active. The networks responsible for original generation, planning and synthesis quieten down. When the tool is taken away, the still-quietened networks do not instantly come back online. There is a lag. The lag is what the EEG picks up. The lag is what shows up in error rates and response latencies on the unaided task that follows.

This is, importantly, not a permanent change. Nothing in the existing literature suggests that ten minutes of AI use causes structural damage to the brain. The relevant concern is not about lasting injury but about state, about the cognitive tone in which the next task is begun, and how quickly that tone recovers when the scaffolding is withdrawn.

Reversibility, And What We Do Not Yet Know

Here the honest answer is that the science is at the very beginning of being able to say anything precise. The MIT preprint hints at carry-over within and across sessions, but its design does not isolate the time course of recovery. The reported April 2026 work claims acute impairment immediately after use; it does not, on the publicly available descriptions, characterise the recovery curve in detail. We have evidence of a measurable effect on the order of minutes after AI use, and we do not yet have systematic evidence about whether that effect dissipates within an hour, a day or a longer period, nor about whether repeated daily exposures produce cumulative residue.

The plausible space of outcomes is not fanciful. If the effect resolves quickly and completely after each exposure, it is roughly analogous to the post-meeting fog that anyone who has spent two hours in a video call recognises, an irritation that fades. If it resolves slowly, or if repeated exposure produces cumulative dampening, the deployment-context implications become substantial. A nurse consulting an AI scribe before a complex assessment, a teacher grading with an AI marker before lesson planning, a junior solicitor moving from AI-drafted briefs to in-court argument: all are scenarios in which acute carry-over, even if reversible, has the potential to land on the high-stakes unaided task that follows.

The claim that needs neither hyping nor dismissing is the modest one. There is evidence, from multiple research groups and instruments, that recent AI use leaves a footprint on subsequent unaided cognition. The size of that footprint, its time course, and its dependence on the type of task, the type of AI and the individual user, are all open questions.

Where The Footprint Lands

The deployment contexts in which acute carry-over would matter most are, helpfully, the same contexts in which AI is being most aggressively deployed. They are not the recreational ones. Nobody is particularly worried about the cognitive aftermath of asking a chatbot to write a wedding speech. The relevant contexts are workplaces where consequential decisions are made under time pressure, classrooms where developing minds are still acquiring the very skills that AI is offloading, healthcare settings where lapses cost lives, and public services where outcomes determine whether citizens are housed, fed, treated or detained.

Take healthcare. In the United Kingdom, AI scribes and clinical-decision-support assistants have proliferated in primary care since 2024, with the Department of Health and Social Care actively encouraging the use of approved tools to reduce administrative burden on general practitioners. The case for these tools is strong; clinician burnout is a public-health emergency in its own right, and time spent transcribing is time not spent with patients. But the consultation that follows the AI-assisted note is not a low-stakes task. It is the next patient. If the cognitive tone with which the clinician enters that next consultation is even slightly slackened by the immediately preceding offload, the relevant question is not whether the tool, on average, saves time. It is whether the unmonitored carry-over is being detected, accounted for, or even acknowledged.

In classrooms, the acute frame inverts the existing debate. The debate so far has largely been about whether students who use AI to do their homework will eventually fail to learn how to write or reason. The acute frame asks a different question: what does it mean to ask a student to use an AI in the first half of a lesson and then to demonstrate understanding in the second? If the cognitive state in which that demonstration happens is materially different from the state of a student who never used the tool, then the assessment is not measuring what it purports to measure. The Department for Education's June 2025 guidance on AI in schools acknowledged that students still needed a strong foundation in reading, writing and critical thinking to use AI effectively. The acute literature, if it stabilises, suggests the guidance does not go nearly far enough. It is not enough to know how to use the tool. The question is what happens when you put it down.

In workplaces more generally, the carry-over question intersects with the dynamic identified by Lee and colleagues at CHI 2025: workers shifting from generation to verification, from problem-solving to integration. If the verification mode itself depends on a cognitive state that is, in the moment, dampened by the just-preceding AI exposure, then verification is precisely the function being undermined. The dynamic is recursive.

In public services, the stakes are starkest. Algorithmic systems already mediate decisions about welfare entitlements, child-protection assessments, criminal-justice risk scoring and immigration triage in many jurisdictions. The case-workers operating those systems are increasingly being given AI-assisted summarisation, drafting and recommendation tools. The decisions they then make about real human lives are made, in some cases minutes after closing the assistant. Whether the cognitive tone in which those decisions are made is materially different from the tone of an unaided counterpart is not a niche concern. In the deployment contexts that matter most, it is the central one.

The ethical literature on technology adoption has historically operated on a strong presumption: that adults, when offered new tools, are entitled to choose to use them, and that the costs of choosing are theirs to bear. This presumption rests on a thicket of assumptions which the acute-impairment frame, if it survives, calls into question.

The first assumption is that the user is the one bearing the cost. In the workplace, that is rarely true. A nurse using an AI scribe is not the principal bearer of the risk if her cognitive tone in the next consultation is dampened. The patient is. A teacher using an AI marker is not the principal bearer if his unaided judgement in the next lesson is reduced. The student is. The deployment of AI in service contexts shifts the costs onto people who were not party to the decision and who, in many cases, do not even know the tool was used.

The second assumption is that the user has been informed. This is, in practice, almost never the case. The patient who interacts with a clinician using an AI scribe is not, in the United Kingdom or in most other jurisdictions, given any disclosure that the scribe was used, much less that recent research has suggested an acute carry-over effect on subsequent unaided cognition. The student whose teacher has just spent half an hour grading essays with an AI marker is not informed. The benefits claimant whose case-worker's notes were drafted by a generative system is not informed. There is, in most settings, no equivalent of the medical-imaging consent form, no equivalent of the data-protection notice, no equivalent of any of the layered consent infrastructures that exist for less consequential interventions in the same lives.

The third assumption is that the cognitive risk is well characterised. It is not. The literature on acute carry-over is, at the time of this writing in April 2026, weeks old in its strongest formulations and months old in its broader contours. Honest informed consent at present would have to read something like: research suggests, but has not yet established, that recent AI use may produce a transient reduction in unaided cognitive performance, the magnitude and duration of which are not yet well understood. That is not a notice that any organisation in any sector is currently required to provide.

The result is a deployment landscape in which a class of cognitive risk has been quietly normalised across millions of high-stakes interactions, on the strength of an implicit assumption that the science was either not real or not relevant, and without any of the consent infrastructure that would be required to make the deployment ethically defensible if the science turns out to be both.

Who Is Currently Responsible, And Who Currently Is Not

The question of who has the responsibility to act on the emerging evidence has, at present, no clean answer. There is a thicket of actors with partial responsibilities, and a great deal of empty space between them where the responsibility falls through.

Regulators are the obvious candidates, but their instruments are not shaped for the problem. The European Union's AI Act, which entered substantive force during 2025, classifies systems by risk and imposes obligations on developers and deployers. It does not require disclosure of cognitive carry-over effects to end-users, nor monitoring in deployment. The United Kingdom's pro-innovation framework prefers sector-specific guidance and avoids cross-cutting consent obligations. The United States, post the rescission of the Biden-era AI executive order in 2025, has effectively no federal framework at all.

Employers have a duty of care to employees and, in regulated sectors, a duty of care to clients and patients. That duty arguably already extends to an obligation to understand the cognitive risks that workplace tools might impose. The General Medical Council in the United Kingdom and equivalent professional bodies elsewhere have begun to issue guidance on AI use in clinical practice, but these documents focus overwhelmingly on data protection, accuracy and clinical accountability. They do not, at the time of writing, address acute carry-over.

Educators bear a different but related duty. The Department for Education's guidance and the curricular adjustments under way in many school systems are mostly oriented toward whether AI use degrades the development of skill over time. They do not address whether AI use during an assessment, or in the hour before one, materially changes what the assessment measures.

Platform vendors are commercially positioned to be most relevant and culturally positioned to be least. The major AI labs (OpenAI, Anthropic, Google DeepMind, Microsoft) have all published responsible-use guidance of varying depth, and have all engaged, with varying degrees of seriousness, with concerns about cognitive effects. None of them, at the time of writing, surfaces information about cognitive carry-over to end-users in the products themselves. The prevailing commercial logic, in which engagement and frequency of use are positive metrics, does not align with cognitive-risk disclosure, and there is no regulatory instrument forcing the alignment.

Individual users carry the residual responsibility, the way they carry it for every consumer product whose risks have been imperfectly disclosed. That is a thin reed to lean on in any sector where the user is, in fact, the patient or the student or the citizen rather than the operator of the tool.

The honest map of responsibility is therefore a sparse one. There is no regulator currently obliged to act, no employer currently obliged to act, no educator currently obliged to act, no platform currently obliged to act, and no user adequately positioned to act. The gaps are not bugs in the system; the system was not designed for the problem, because the problem was framed, until very recently, as chronic.

Replication, Caveats, And The Cost Of Waiting

It would be irresponsible to leave this account without flagging that the strongest version of the acute-impairment claim still rests on a small number of studies, much of it unreplicated and some of it not yet peer reviewed. The MIT preprint by Kosmyna and colleagues has the limitations its own authors acknowledged: a sample of 54 participants in a single geographic region, no peer review at the time of its initial release, and a fourth-session reassignment design that, while suggestive, is not definitive. The CHI 2025 paper by Lee and colleagues is a survey of self-reported behaviour, not a controlled experiment. The Gerlich 2025 paper in Societies is correlational and was subsequently corrected by the publisher in September 2025 for unrelated issues.

The reported April 2026 multi-institution study would be the strongest causal evidence yet, but its full methodological detail is not, at the time of writing, available for the kind of scrutiny that allows confident claims. It will need to be peer reviewed. It will need to be replicated. It will need to be tested against the standard battery of cognitive-experiment objections: demand characteristics, expectancy effects, the difficulty of isolating the AI-use intervention from time-on-task confounds, the question of whether the post-test deficit is a real cognitive change or a motivational artefact.

These caveats matter, and the article that elides them does the public no favours. They do not, however, license inaction. The asymmetry of the situation is consequential. The cost of acting on a finding that turns out to overstate its case is, mostly, the modest inconvenience of disclosure obligations that would have been good practice anyway. The cost of failing to act on a finding that turns out to be robust is the continued silent conversion of millions of high-stakes interactions into a field experiment whose subjects never agreed to participate.

The right posture, on the present evidence, is therefore neither alarm nor dismissal. It is the unfashionable posture of taking research seriously while it is still emerging, of treating disclosure and consent as prudent defaults under uncertainty, and of designing deployment contexts to be measurable, monitorable and reversible. None of these are dramatic interventions. None require believing that the strongest claims in the literature are true. They require only believing that they might be.

What The Evidence Demands, And What It Does Not

What the evidence demands is modest, and would be modest even if every study cited above were fully replicated and beyond serious dispute. It demands, first, that the deployment of AI assistants in high-stakes settings be accompanied by disclosure to the individuals whose welfare depends on the cognitive performance that follows. The patient is entitled to know that the clinician is using a scribe. The student is entitled to know that the teacher is grading with a marker. The benefits claimant is entitled to know that the case-worker has just closed a chatbot.

It demands, second, that organisations deploying these tools begin to monitor outcomes in a manner sensitive to acute carry-over. Quality-assurance audits exist; they have not, until now, been designed with the carry-over hypothesis in mind, but they could be without much trouble.

It demands, third, that regulators, professional bodies and educators begin to update their guidance with the acute frame in view, and stop treating the cognitive consequences of AI use as a problem of long-term skill development alone. It demands, fourth, that platform vendors stop pretending the question of cognitive effects is somebody else's department, and begin to surface, in their products, the relevant information that emerging research has produced.

What the evidence does not demand is panic. It does not demand that AI be removed from clinics, classrooms or public-service settings. It does not demand that workers stop using tools that, on net, help them do their jobs. It does not demand the kind of moral-panic legislation that would, if enacted on the present evidence, almost certainly do more harm than good.

What it asks of us is the harder thing: to live, as adults, in the uncomfortable middle ground where evidence is suggestive but not yet conclusive, where the costs of action are real but bounded, and where the costs of inaction are uncertain but potentially large. The history of technology regulation is mostly the history of arriving at this middle ground decades after the relevant tools have already reshaped the landscape. The unfashionable possibility, this time, is to arrive earlier.

The Smaller, Truer Claim

Strip the press coverage of its more lurid framings and what remains is a claim that is smaller and harder to dismiss. The claim is not that AI is rotting our brains. The claim is not that ten minutes of ChatGPT will leave you intellectually impaired for the rest of the day. The claim is not even that the acute effect, if it exists, is large enough to matter in the average use case.

The claim is that there is a measurable carry-over effect from recent AI use to subsequent unaided cognitive performance, that the effect appears on the order of minutes rather than years, that the existing deployment of AI in high-stakes contexts has not been designed with that effect in mind, and that the consent and disclosure infrastructure required to make that deployment ethically defensible has not been built. The reported April 2026 study strengthens the first proposition. The MIT, Microsoft, Carnegie Mellon and Swiss Business School literatures of the past eighteen months have already strengthened the second. The third is empirical and visible to anyone who looks. The fourth is a matter of public policy that we have, until now, declined to address.

The room from the opening of this article, the desk and the laptop and the EEG electrodes, is not a metaphor. It is a research site, one of a small but growing number, in which the cognitive tone of recent AI users is being measured against the cognitive tone of unaided controls. Whether the field finds the effect to be small and easily managed, or large and policy-relevant, will become clearer in the months ahead. That it is being measured at all is the first piece of good news. That the measurements are not yet, in any meaningful sense, being relayed to the patients, students, clients and citizens whose welfare depends on the unaided performance that follows the use of the tools, is the part of the situation that does not require any further evidence to fix.

The technology will not pause for the science to catch up. The disclosure can.

References & Sources

  1. Kosmyna, N., et al. (2025). Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. arXiv preprint arXiv:2506.08872. MIT Media Lab. https://www.media.mit.edu/publications/your-brain-on-chatgpt/
  2. Lee, H.-P. H., et al. (2025). The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. Proceedings of the CHI Conference on Human Factors in Computing Systems (CHI '25), Yokohama. Microsoft Research / Carnegie Mellon University. https://www.microsoft.com/en-us/research/publication/the-impact-of-generative-ai-on-critical-thinking-self-reported-reductions-in-cognitive-effort-and-confidence-effects-from-a-survey-of-knowledge-workers/
  3. Gerlich, M. (2025). AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15(1), 6. SBS Swiss Business School. https://www.mdpi.com/2075-4698/15/1/6
  4. Gerlich, M. (2025). Correction: AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking. Societies, 15(9), 252. https://www.mdpi.com/2075-4698/15/9/252
  5. Sparrow, B., Liu, J., & Wegner, D. M. (2011). Google Effects on Memory: Cognitive Consequences of Having Information at Our Fingertips. Science, 333(6043), 776-778.
  6. Maguire, E. A., et al. (2000). Navigation-related structural change in the hippocampi of taxi drivers. Proceedings of the National Academy of Sciences, 97(8), 4398-4403.
  7. Javadi, A.-H., Spiers, H. J., et al. (2017). Hippocampal and prefrontal processing of network topology to simulate the future. Nature Communications, 8, 14652.
  8. Department for Education (UK). (2025, June). Generative artificial intelligence (AI) in education: policy paper. https://www.gov.uk/government/publications/generative-artificial-intelligence-in-education
  9. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/1689 (Artificial Intelligence Act). Official Journal of the European Union.
  10. UK Department for Science, Innovation and Technology. (2023). A pro-innovation approach to AI regulation. White paper. HM Government.
  11. TIME Magazine. (2025). ChatGPT's Impact On Our Brains According to an MIT Study. https://time.com/7295195/ai-chatgpt-google-learning-school/
  12. WBUR News. (2025, September 16). Using ChatGPT as a homework tool? MIT researcher says think twice. https://www.wbur.org/news/2025/09/16/ai-study-essays-brain-cognition
  13. Fortune. (2025, February 11). AI might already be warping our brains, leaving our judgment and critical thinking 'atrophied and unprepared,' warns new study. https://fortune.com/2025/02/11/ai-impact-brain-critical-thinking-microsoft-study/
  14. Policy Options / IRPP. (2025, September). How AI is eroding human memory and critical thinking. https://policyoptions.irpp.org/2025/09/ai-memory/
  15. Frontiers in Psychology. (2025). The cognitive paradox of AI in education: between enhancement and erosion. https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2025.1550621/full

Tim Green

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

Listen to the free weekly SmarterArticles Podcast

Discuss...