We read more words than any generation before us, and understand fewer of them deeply. The shift is not a failure of will. It is the predictable result of reading on surfaces engineered to move us along. What follows is the evidence — from reading neuroscience, education research, and the science of learning — and the case it makes for building something different. Every empirical claim links to its source.
The shallowing
The screen rewards the scan.
For two decades, reading researchers have documented a consistent pattern. When the same text is moved from paper to a screen, comprehension of its argument tends to fall — not because the words change, but because the behavior around them does. A meta-analysis pooling fifty-four studies and more than 170,000 readers found a small but stubborn advantage for paper, and, tellingly, the gap was widening rather than closing as screens improved.5 The disadvantage concentrates exactly where it matters: under time pressure, on the dense, informational text an argument is built from.
Part of the cause is simply how we behave in front of a screen. As far back as 2005, researchers were describing a distinct “screen-based reading” style — browsing, scanning, keyword-spotting, reading once and moving on.1 Eye-tracking made the shape of it visible: on a web page, the gaze traces an “F,” taking the first lines and the left margin and skipping the rest.2 The page is consumed, but it is rarely studied.
The surfaces themselves invite it. Hyperlinks — the basic unit of the web — carry a hidden cost: every link is a small decision (click or not, where does it go, how do I get back), and a review of the research found this added load tends to lower comprehension versus plain, linear text, especially for readers with less working memory or background knowledge.3 Eye-tracking shows the effect in miniature: the instant a reader starts skimming, or the moment links become clickable, they fully process only the highlighted words and skim past everything around them.4
The dominant digital surfaces are tuned for a single metric — time on glass — and the fastest way to hold attention is to keep promising the next thing. Notifications interrupt; infinite feeds remove every natural stopping point. Together they train a reflex: arrive, sample, react, move. And the habit does not stay online. A 2025 meta-analysis of roughly 470,000 readers found that leisure reading on screens barely relates to comprehension at all — six to eight times weaker than the benefit of print, and actually negative for younger children.6 The reflex follows us onto the long-form page, where it does not belong.
What erodes is not the ability to decode words. It is the slower faculty that sits underneath comprehension — the willingness to hold a difficult idea in mind long enough to test it, to sit with ambiguity, to let an author’s argument finish before answering it. That faculty is the substance of reading deeply, and it is precisely what the modern surface gives us no room to practice.
What erodes is not the ability to read the words. It is the patience to let them mean something.
The slump
It is showing up in the numbers.
If this were only a laboratory finding, it would be easy to wave away. It is not. The same downward pressure is now visible in the largest measurements we have of how well whole populations read and think.
−10 pts7
Drop in OECD reading scores, 2018–2022 — the steepest single-cycle decline on record.
47 sec13
Average focus on one screen before switching — down from ~2.5 minutes in 2004.
≈40%11
Decline in Americans reading for pleasure on a given day, 2003–2023.
On PISA, the worldwide test of fifteen-year-olds, average reading fell about ten points between 2018 and 2022 — roughly three-quarters of a school year, and the steepest single drop the test has recorded. The OECD was careful to note the slide began about a decade before the pandemic.7 In the United States, the 2024 Nation’s Report Card told the same story: reading down at grades four and eight, the lowest twelfth-grade score ever recorded, and a record share of children reading below a basic level.8 Nor is it only children — when the OECD retested adults a decade on, literacy had improved in just two countries and was flat or falling almost everywhere else.9
The most striking signal comes from a measure that climbed for most of a century. Average intelligence scores rose for decades — the “Flynn effect” — then stalled and reversed in several wealthy countries. Among more than 730,000 Norwegian men, scores peaked for those born around 1975 and have fallen since; crucially, the decline appears between brothers in the same family, which rules out genetics and points squarely at a changing environment.10
Underneath the scores, the activity itself is receding. Across 236,000 Americans surveyed over twenty years, the share who read for pleasure on a given day fell by roughly 40 percent;11 federal arts data put book reading, and fiction reading especially, at their lowest rates in the survey’s three-decade history.12 And the attention reading requires is under measurable strain. Tracking real screen use, the average time a person spends on one screen before switching has collapsed — from about two and a half minutes in 2004 to some forty-seven seconds today.13 Heavy media multitaskers are measurably worse at filtering distraction,14 and lapses can be caught in the pupil and the brain’s electrical signals in the seconds before memory fails.15 Reviews link heavier phone use to weaker attention and shallower, less analytic thinking — with the honest caveat that this evidence is correlational.16
None of this proves screens single-handedly caused the decline; these are population trends with many causes, and careful researchers say so. But the direction is consistent across instruments that have little in common, and it converges on a single faculty — the sustained, effortful attention that deep reading both requires and builds.
The hopeful part
Attention is trained, not fixed.
If deep attention were a fixed trait, this would be a lament and nothing more. It is not. The same body of research that maps the decline also points to its reversal: sustained focus behaves like a trainable capacity, not an inheritance.
Start with the reading brain itself. Reading is a human invention only a few thousand years old — far too recent for evolution to have built an organ for it. The neuroscientist Stanislas Dehaene calls this the reading paradox, and resolves it with “neuronal recycling”: every script we ever invented had to be absorbed by brain circuits that evolved for vision.17 Imaging bears it out — reading reliably lands in one repurposed patch of cortex, the “visual word form area,” across every language and script.18 Nobody is born with a reading brain; practice carves one out.
And it keeps carving. When researchers scanned sixty-three adults with different reading histories, literacy had physically reshaped the brain — and most of those changes appeared even in people who learned to read as adults.19 The plasticity is not metaphorical: in a randomized reading program, struggling seven- to nine-year-olds who improved showed measurable changes in the white-matter wiring of their reading network within weeks.20 The circuit that supports close reading is assembled through use — and it weakens through disuse, in both directions.
Attention answers to practice too. In a controlled trial, a two-week focus course cut mind-wandering and raised both working memory and reading-comprehension scores, with the largest gains for the most distractible;21 a meta-analysis of 111 randomized trials confirms attention training reliably helps, though the gains are real-but-modest rather than magical.22 The crucial qualification is what does not work: generic “brain-training” drills improve the drill and little else,23 and even focused practice explains only part of any skill.24 The lesson is exact — to read better, and to hold attention while reading, you have to read, not run puzzles adjacent to it.
This reframes the problem. The question is no longer whether we can still read deeply — most of us once could — but whether the tools we read with are training the faculty or starving it. A surface that constantly interrupts is running the wrong exercise. A surface that protects a single unit of thought, removes the next-thing reflex, and asks something of the reader is running the right one. Environment is not neutral. It is the curriculum.
The faculty
What deep reading builds.
It is worth being precise about what “deep reading” names, because it is not merely reading slowly. Maryanne Wolf and Mirit Barzillai defined it as the sophisticated work that turns decoded words into understanding — inference, analogy, critical analysis, reflection, the arrival at insight — and warned that these cultivated processes can be crowded out by skim-heavy habits.25 In Wolf’s fuller account, the “deep-reading circuit” recruits attention, memory, background knowledge, analysis and empathy at once; it is built, not given, and it asks for what she calls cognitive patience.26
You can watch the difference. The brain network behind daydreaming — the “default mode network” — turns out to be central to comprehension too, which is why reading can either absorb you completely or let your mind drift, depending on whether that network stays coupled to the text.27 Deep narrative reading leaves a trace: daily scans found that reading a novel raised connectivity in regions tied to taking another’s perspective, with a bodily, sensory boost that lingered for days after the last page — suggestive, from a small study, but pointed.28
What the faculty pays back is large and unusually well-measured. The sheer amount a person reads predicts vocabulary, general knowledge and comprehension a decade later, even after accounting for early intelligence, because most of our words are absorbed through reading rather than taught;30 in a British cohort of six thousand, childhood reading for pleasure predicted gains in vocabulary and even mathematics larger than the advantage of having highly educated parents.31 Reading tracks lower cognitive decline in older age,36 and a Yale study following thousands of older adults found that book readers lived, on average, almost two years longer than non-readers.29
Some of the most appealing claims deserve the most caution. The idea that fiction is a “flight simulator” for social life is a compelling theory,32 and a famous 2013 experiment found that reading literary fiction briefly sharpened the ability to read others’ emotions.33 But a 792-person replication across three labs found no such instant effect; what predicted social skill was not a single passage but a lifetime of reading.34 The careful conclusion is the more interesting one: depth compounds. A large adolescent study links early reading for pleasure to better cognition and mental health,35 and structured reading can ease depression about as well as seeing a therapist.37 The dose is a habit, not a hit.
The design bet
Against the one-tap summary.
Which is where most new reading tools take the wrong turn. The reflexive application of AI to reading is the summary — the offer to compress the book into its conclusions and hand them over. It is fast, and it feels like progress, and it removes the one thing that mattered. A summary delivers the author’s claim while skipping the reader’s work of arriving at it.
Learning science has studied this exact trade for half a century, and the verdict is consistent: the effort is not a cost of understanding — it is the mechanism. People remember words they generate themselves far better than the same words handed to them.38 Testing yourself on a passage beats re-reading it — a week later, by something like 61 percent recalled to 40 — even though re-reading feels more productive.39 That feeling is the trap: when a method is smooth, learners reliably mistake fluency for mastery and cannot tell which approach is actually working.40 The broad principle has a name — “desirable difficulties” — and its corollary is blunt: the convenience of a shortcut is precisely the signal that little is being learned.41 Even taking notes by hand beats typing them, because the slower hand forces you to reframe instead of transcribe.42
The newest evidence aims straight at the summary. A study of 666 people found heavier reliance on AI tools correlated with weaker critical thinking, the link running through “cognitive offloading” — letting the tool do the thinking.43 An MIT experiment wired up essay-writers and found that those leaning on a chatbot showed the weakest brain connectivity, and afterward most could not quote a single sentence they had just “written” — a pattern the researchers call cognitive debt.44 Both come with honest caveats — one is correlational, the other a small preprint — but they point the same way the fifty years of learning research do.
IO is built against that shortcut. The personas do not condense the page; they argue with it. The Critical Thinker tests the passage’s logic and finds the condition the famous line leaves out. The Interpreter builds the bridge from a difficult sentence to what you already know. The Translator carries meaning across a language without flattening the music of the original. Each is built to leave the reader sharper than the page found them — to ask more, not less.
That is the bet IO is built on. The capacity for sustained, deliberate thought is becoming the scarcest human asset — and the evidence says it can still be rebuilt, one paragraph at a time, by anyone willing to do the work and any tool honest enough to ask for it.
References
The evidence, in full.
Every empirical claim above links to its source: peer-reviewed studies, meta-analyses pooling tens or hundreds of thousands of readers, and official national and international datasets. Where a finding is contested, correlational, or drawn from a small sample, we say so rather than rounding it up.
Liu (2005)survey
Reading behavior in the digital environment: Changes in reading behavior over the past ten years. Journal of Documentation, 61(6).
A landmark survey documenting the rise of “screen-based reading”: as screens spread, people spend more time browsing, scanning, keyword-spotting and reading once — and less time in deep, concentrated reading.
View source ↗Nielsen (2006)eye tracking
F-Shaped Pattern for Reading Web Content. Nielsen Norman Group (eye-tracking study).
Eye-tracking of 232 users found people scan web text in an “F” shape rather than read it — taking the first lines and the left edge, skipping most of the rest. Follow-ups find the pattern has persisted for ~20 years.
View source ↗DeStefano & LeFevre (2007)review
Cognitive load in hypertext reading: A review. Computers in Human Behavior, 23(3).
A review concluding that the constant decisions hyperlinks demand — whether to click, where it leads, how to get back — add mental load and tend to lower comprehension versus plain linear text, worst for readers with less working memory or prior knowledge.
View source ↗Fitzsimmons, Jayes, Weal & Drieghe (2020)eye tracking
The impact of skim reading and navigation when reading hyperlinks on the web. PLOS ONE, 15(9).
Eye-tracking showed that the moment readers skim — or links become clickable — they fully process only the hyperlinked words and give the surrounding text far shallower attention, using links as flags for “the important bits.”
View source ↗Delgado, Vargas, Ackerman & Salmerón (2018)meta analysis
Don’t throw away your printed books: A meta-analysis on the effects of reading media on reading comprehension. Educational Research Review, 25.
A meta-analysis of 54 studies and 171,000+ readers: comprehension was reliably better on paper than screen (g ≈ −0.21), with the gap widening under time pressure — and growing, not shrinking, in more recent studies. The effect is strongest for informational text, not narrative.
View source ↗Altamura, Vargas & Salmerón (2025)meta analysis
Do new forms of reading pay off? A meta-analysis on the relationship between leisure digital reading habits and text comprehension. Review of Educational Research, 95(1).
A meta-analysis of ~470,000 readers found that how much people read for fun on screens barely relates to comprehension (r ≈ 0.05) — six-to-eight times weaker than print reading — and was actually negative for primary- and middle-school children.
View source ↗OECD (2023)survey
PISA 2022 Results (Volume I): The State of Learning and Equity in Education. OECD Publishing, Paris.
On the largest international test of 15-year-olds, average reading across OECD countries fell ~10 points (about three-quarters of a school year) from 2018 to 2022 — the steepest single-cycle drop on record, a slide the OECD says began roughly a decade before the pandemic.
View source ↗Nation’s Report Card (NCES) (2025)survey
The Nation’s Report Card: 2024 NAEP Reading Assessment Results. U.S. Dept. of Education, IES (NAEP).
The 2024 U.S. assessment found reading down ~5 points since 2019 at grades 4 and 8, the lowest-ever 12th-grade score, and record shares of students — roughly 40% of 4th-graders — reading below the “Basic” level.
View source ↗OECD (2024)survey
Do Adults Have the Skills They Need to Thrive in a Changing World? Survey of Adult Skills 2023. OECD Publishing, Paris (PIAAC).
Retesting adults a decade on, average literacy improved in only two countries (Finland and Denmark) and was flat or falling everywhere else; the weakest readers slipped further behind in most countries, widening the gap.
View source ↗Bratsberg & Rogeberg (2018)observational
Flynn effect and its reversal are both environmentally caused. PNAS, 115(26).
Across 730,000+ Norwegian men, IQ rose for decades, peaked for those born around 1975, then fell ~0.2 points a year. The decline appeared between brothers in the same family — ruling out genetics or immigration and pointing to a changing environment.
View source ↗Bone, Bu, Sonke & Fancourt (2025)survey
The decline in reading for pleasure over 20 years of the American Time Use Survey. iScience, 28(9).
Across 236,270 Americans (2003–2023), the share reading for pleasure on a given day fell from ~28% to ~16% — down roughly 40% — with the steepest drops among lower-income and less-educated groups.
View source ↗National Endowment for the Arts (2024)survey
Federal Data on Reading for Pleasure: All Signs Show a Slump. NEA (2022 Survey of Public Participation in the Arts).
Federal survey data: U.S. adults reading any book in the prior year fell from 54.6% (2012) to 48.5% (2022), and novel- or short-story reading dropped to 37.6% — the lowest fiction-reading rate in the survey’s 30-plus-year history.
View source ↗Mark (2023)book
Attention Span: A Groundbreaking Way to Restore Balance, Happiness and Productivity. Hanover Square Press.
Tracking real screen use, the average time on one screen before switching fell from ~2.5 minutes in 2004 to about 47 seconds today (median 40 seconds) — a figure independent studies have replicated.
View source ↗Ophir, Nass & Wagner (2009)observational
Cognitive control in media multitaskers. PNAS, 106(37).
Heavy media multitaskers were worse at ignoring distractions and slower to switch tasks — they multitask more but do it less well. (Later replications find the effect small and inconsistent, so read it as suggestive.)
View source ↗Madore, Khazenzon, Backes, Jiang, Uncapher, Norcia & Wagner (2020)neuroimaging
Memory failure predicted by attention lapsing and media multitasking. Nature, 587.
Pupil size and brain (EEG) signals caught attention “drifting” in the seconds before people forgot — and heavier media multitaskers showed weaker sustained attention and remembered less, a concrete marker linking distraction to memory failure.
View source ↗Wilmer, Sherman & Chein (2017)review
Smartphones and cognition: A review of research exploring the links between mobile technology habits and cognitive functioning. Frontiers in Psychology, 8.
A review linking heavier smartphone use to poorer attention, weaker memory for offloaded information, and less analytic thinking — while stressing the evidence is correlational and cannot yet prove phones are the cause.
View source ↗Dehaene (2009)book
Reading in the Brain: The New Science of How We Read. Viking / Penguin.
The neuroscientist’s account of the “reading paradox”: writing is far too recent for evolution to have built a reading organ, so every script humans invented had to be absorbed by “recycling” brain circuits evolved for vision.
View source ↗Dehaene & Cohen (2011)review
The unique role of the visual word form area in reading. Trends in Cognitive Sciences, 15(6).
Reading reliably activates one repurposed spot — the “visual word form area” — across all languages and scripts. Nobody is born with a reading brain; practice carves it out of equipment built for other jobs.
View source ↗Dehaene, Pegado, Braga et al. (2010)observational
How learning to read changes the cortical networks for vision and language. Science, 330(6009).
Brain scans of 63 adults with different reading histories showed literacy physically reshapes the brain — and most of those changes appeared even in people who learned to read as adults, direct evidence the reading brain rewires well past childhood.
View source ↗Meisler, Gabrieli & Christodoulou (2024)rct
White matter microstructural plasticity associated with educational intervention in reading disability. Imaging Neuroscience, 2.
In a randomized reading intervention, struggling 7–9-year-olds who improved showed measurable changes in the white-matter wiring of the brain’s core reading network — teaching the skill visibly altered the brain’s reading pathways within weeks.
View source ↗Mrazek, Franklin, Phillips, Baird & Schooler (2013)rct
Mindfulness training improves working memory capacity and GRE performance while reducing mind wandering. Psychological Science, 24(5).
A two-week, eight-session attention course cut mind-wandering and raised working memory and GRE reading-comprehension scores (about a 16-percentile jump) — most for the people who started out most distractible.
View source ↗Zainal & Newman (2024)meta analysis
Mindfulness enhances cognitive functioning: A meta-analysis of 111 randomized controlled trials. Health Psychology Review, 18(2).
Across 111 randomized trials and 9,538 people, attention training reliably improved focus and thinking. The gains were small-to-moderate (about g = 0.37) — genuinely real, but meaningful rather than magical.
View source ↗Melby-Lervåg, Redick & Hulme (2016)meta analysis
Working memory training does not improve performance on measures of intelligence or other measures of “far transfer”. Perspectives on Psychological Science, 11(4).
A meta-analysis of 87 studies: generic “brain-training” improves the drilled task but shows no convincing transfer to broader skills like reasoning or reading comprehension — the practice has to resemble the thing you want to improve.
View source ↗Macnamara, Hambrick & Oswald (2014)meta analysis
Deliberate practice and performance in music, games, sports, education, and professions: A meta-analysis. Psychological Science, 25(8).
Across domains, focused practice explained only part of the difference in skill — about 21% in music and just 4% in education — a balanced corrective: practice genuinely builds ability, but it is one factor among several, not a guarantee.
View source ↗Wolf & Barzillai (2009)review
The importance of deep reading. Educational Leadership, 66(6).
The influential essay that defined “deep reading” — inference, analogy, critical analysis, reflection, insight — and warned these cultivated processes can be crowded out by fast, skim-heavy digital habits.
View source ↗Wolf (2018)book
Reader, Come Home: The Reading Brain in a Digital World. Harper / HarperCollins.
Wolf’s case that the “deep-reading circuit” — attention, memory, inference, analysis, empathy working at once — is built, not given, and can weaken under screen-driven skimming. Her remedy is “biliteracy”: build deep reading first, then add fluent digital reading.
View source ↗Smallwood, Gorgolewski, Golchert et al. (2013)neuroimaging
The default modes of reading: Modulation of posterior cingulate and medial prefrontal cortex connectivity associated with comprehension and task focus while reading. Frontiers in Human Neuroscience, 7.
The brain network behind daydreaming — the “default mode network” — is also central to comprehension, which is why reading can either pull you in or let your mind drift. Deep comprehension means keeping that internal network harnessed to the text.
View source ↗Berns, Blaine, Prietula & Pye (2013)neuroimaging
Short- and long-term effects of a novel on connectivity in the brain. Brain Connectivity, 3(6).
Daily brain scans found that reading a novel raised connectivity in regions tied to perspective-taking, with a sensory-motor boost lingering for days after the book was finished. (The sample was small — 19 readers — so treat it as suggestive.)
View source ↗Bavishi, Slade & Levy (2016)observational
A chapter a day: Association of book reading with longevity. Social Science & Medicine, 164.
Yale researchers followed 3,635 older adults for up to 12 years: book readers had a 20% lower risk of dying and lived on average ~23 months longer than non-readers, even after adjusting for health, wealth and education.
View source ↗Cunningham & Stanovich (1997)observational
Early reading acquisition and its relation to reading experience and ability 10 years later. Developmental Psychology, 33(6).
Following children for a decade, the sheer amount of print a student read predicted later vocabulary, general knowledge and comprehension — even after accounting for early IQ. Most vocabulary is absorbed through reading, not direct teaching.
View source ↗Sullivan & Brown (2015)observational
Reading for pleasure and progress in vocabulary and mathematics. British Educational Research Journal, 41(6).
Across ~6,000 Britons, reading for pleasure between ages 10 and 16 predicted greater progress in vocabulary, spelling and even mathematics — and the vocabulary boost was larger than the effect of having highly educated parents.
View source ↗Mar & Oatley (2008)review
The function of fiction is the abstraction and simulation of social experience. Perspectives on Psychological Science, 3(3).
The theory that stories are a “flight simulator” for social life: by living inside characters, readers rehearse understanding other minds. (A synthesis, not one experiment; the causal empathy claim remains debated.)
View source ↗Kidd & Castano (2013)rct
Reading literary fiction improves theory of mind. Science, 342(6156).
Five experiments found that reading a few pages of literary fiction briefly improved “theory of mind” — reading others’ emotions — more than popular fiction or nonfiction. Much-cited, but contested (see the replication that follows).
View source ↗Panero, Weisberg, Black et al. (2016)rct
Does reading a single passage of literary fiction really improve theory of mind? An attempt at replication. Journal of Personality and Social Psychology, 111(5).
A 792-person, three-lab replication found no instant empathy boost from one literary passage. What predicted social skill was a lifetime of reading, not the assigned text — the honest counterweight to the 2013 result.
View source ↗Sun, Sahakian, Langley et al. (2024)observational
Early-initiated childhood reading for pleasure: Associations with better cognitive performance, mental well-being and brain structure in young adolescence. Psychological Medicine, 54(2).
Across 10,000+ adolescents, those who began reading for pleasure early scored higher on cognition, reported better mental health, and showed differences in brain structure. Correlational, but in a very large, diverse sample.
View source ↗Yang, Xu, Guo, Zhang, Wang & Li (2022)meta analysis
Effect of leisure activities on cognitive aging in older adults: A systematic review and meta-analysis. Frontiers in Psychology, 13.
Pooling longitudinal studies of older adults, more cognitive leisure activity — reading prominent among it — was linked to roughly 23% lower risk of cognitive decline, fitting the “use it or lose it” pattern of cognitive reserve.
View source ↗Cuijpers (1997)meta analysis
Bibliotherapy in unipolar depression: A meta-analysis. Journal of Behavior Therapy and Experimental Psychiatry, 28(2).
Pooling controlled trials, working through a structured self-help book eased depression substantially (effect size ≈ 0.82) — and about as much as individual or group therapy with a clinician. Early evidence that the act of reading itself can heal.
View source ↗Slamecka & Graf (1978)rct
The generation effect: Delineation of a phenomenon. J. Experimental Psychology: Human Learning and Memory, 4(6).
The foundational “generation effect”: words you produce yourself are remembered reliably better than the same words handed to you. The mental effort of generating an answer, not receiving it, is what cements it in memory.
View source ↗Roediger & Karpicke (2006)rct
Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3).
Self-testing beat re-reading: a week later the retrieval group recalled ~61% versus ~40% for repeat-readers — yet the re-readers felt more confident. The easier method that felt better actually taught less.
View source ↗Karpicke & Roediger (2008)rct
The critical importance of retrieval for learning. Science, 319(5865).
Once material was learned, extra restudying did almost nothing while extra self-testing produced large gains — and students couldn’t tell which was working. This is the “illusion of knowing”: fluency feels like mastery but isn’t.
View source ↗Soderstrom & Bjork (2015)review
Learning versus performance: An integrative review. Perspectives on Psychological Science, 10(2).
The review of “desirable difficulties”: tactics that smooth the moment — cramming, re-reading, being handed answers — boost short-term performance but undercut durable learning. The convenience of a shortcut is the signal that little is being learned.
View source ↗Mueller & Oppenheimer (2014)rct
The pen is mightier than the keyboard: Advantages of longhand over laptop note taking. Psychological Science, 25(6).
Students who typed notes transcribed verbatim and understood concepts less than those writing longhand, who had to reframe in their own words. (A 2019 multi-lab replication found the headline effect weak — but the “process, don’t transcribe” principle holds.)
View source ↗Gerlich (2025)survey
AI tools in society: Impacts on cognitive offloading and the future of critical thinking. Societies, 15(1).
Among 666 people, heavier AI-tool use correlated strongly with weaker critical-thinking scores (r = −0.68), a link explained by “cognitive offloading” — letting the tool do the thinking. Correlational: the author cautions it cannot prove cause.
View source ↗Kosmyna, Hauptmann, Yuan et al. (2025)neuroimaging
Your Brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task. MIT Media Lab (arXiv:2506.08872).
An EEG study of 54 writers: ChatGPT users showed the weakest brain connectivity, and 83% couldn’t quote a sentence from the essay they’d just produced — what the authors call “cognitive debt.” A small, not-yet-peer-reviewed preprint, so suggestive.
View source ↗
