This might interest people. I saw it in a newsletter but didn’t look closely yet.
Here’s what we discovered: The very educational reforms that promised to make us smarter—particularly the widespread rejection of memorization as “rote learning” and the shift toward “critical thinking skills,” combined with our increasing reliance on external devices—may actually be undermining our cognitive abilities.
When we constantly offload thinking to calculators, smartphones, and AI, we’re not just changing where information is stored. We’re bypassing the brain’s natural learning mechanisms that build the neural architecture necessary for genuine understanding and insight.
The chapter traces how this “cognitive offloading” disrupts everything from forming robust mental frameworks to developing intuitive expertise. When students never memorize math facts because “they can always use a calculator,” or when we “just Google it” instead of wrestling with information internally, we may be weakening the very cognitive foundations that enable advanced reasoning. The timing is telling: the cohorts showing declining IQ scores were precisely those educated when memorization was not just abandoned but actively demonized, and digital dependence took hold.
In an era of generative AI and ubiquitous digital tools, human memory faces a paradox: the more we offload knowledge to external aids, the less we exercise and develop our own cognitive capacities. This chapter offers the first neuroscience-based explanation for the observed reversal of the Flynn Effect—the recent decline in IQ scores in developed countries—linking this downturn to shifts in educational practices and the rise of cognitive offloading via AI and digital tools. Drawing on insights from neuroscience, cognitive psychology, and learning theory, we explain how underuse of the brain’s declarative and procedural memory systems undermines reasoning, impedes learning, and diminishes productivity. We critique contemporary pedagogical models that downplay memorization and basic knowledge, showing how these trends erode long-term fluency and mental flexibility. Finally, we outline policy implications for education, workforce development, and the responsible integration of AI, advocating strategies that harness technology as a complement to – rather than a replacement for – robust human knowledge.
“This invention [writing], O king, will create forgetfulness in the learners’ souls, because they will not use their memories; they will trust to the external written characters and not remember of themselves. The specific which you have discovered is an aid not to memory, but to reminiscence, and you give your disciples not truth, but only the semblance of truth; they will be hearers of many things and will have learned nothing; they will appear to be omniscient and will generally know nothing; they will be tiresome company, having the show of wisdom without the reality.”
Every time a new technology arises, we, as human beings, lose some kind of cognitive ability. Now, we are losing the ability to think, to focus, and to remember.
Thanks for that link. Interesting (and somewhat alarming) developments - most of it makes intuitive sense and I already suspected that reliance on LLMs would be detrimental to deep understanding and thinking. I stopped using them as a coding assistant pretty quickly as I could sense my thinking and decision-making abilities start to atrophy already…
One thing that seemed odd was the implication that teaching “critical thinking skills” could be responsible for this inverse Flynn effect (dropping IQs), but upon skimming the paper it seems moreso that there was an educational shift, trading away rote memorisation to make room for more abstract topics including critical thinking.
I wonder if there are easy fixes that can undo some of that damage without giving up whatever benefits were gained.
Wow! This is very interesting stuff. This kind of reminds me of math class.
So, I do a lot of math competitions. And I’ve noticed that there’s a huuuuge difference between public school math vs. competition math. In public school math, all you do is just memorize, memorize, and memorize more formulas. Whereas in competition math, you usually have to actually understand stuff. Competition math deals mostly with a deep understanding of elementary math and applying clever uses of it. I think competition math is a great way to keep a high IQ.
Another factor is google maps and navigation. Exercising the skill of navigation is important in the development of the hippocampus. This is also impaired by the fact that children are always under supervision now so they are never responsible for finding their own way around.
Things look pretty bleak for humanity at the moment, we’re being prodded into a state of learned helplessness, and all investment has been directed towards it. We can avoid it on an individual level but I don’t know how the ship can be turned around on a civilizational level. I keep seeing people use LLMs to confirm their biases.
I don’t doubt that critical cognitive skills are in decline, especially as passivity becomes the norm. The loss of memorization from the public schooling system, especially for younger students is definitely something to lament as well (I submit I have some nostalgia attached here).
However, I do have a number of qualms with the source provided. It is generally a good rule of thumb to approach all pre-prints (i.e. anything from arXiv, SSRN, etc.) with caution, as they are not peer-reviewed. The front page states the chapter is to be published in Springer Nature, but this does not provide any particularly outstanding merit, given there is no listed associated journal (with a peer-review process). Springer is a publisher that owns a number of journals. Evidently, the quality of work published from each journal varies.
A good deal of logical jumps are made, even in the first few pages of the paper, with a corresponding lack of citations for a number of their claims. Furthermore, upon reading any portion of the paper it becomes strikingly obvious that generative AI was used to write the entire body of the text. Concerns regarding ethics and credibility (and the inherent irony of using AI to both do a literature review and write your criticism of AI supplementing the decline of IQ scores) aside, LLMs do not have the current capacity to produce scientifically rigorous work, as evidenced stunningly by this paper.
TLDR; you make a great point, but this paper fails to do so (or back said point up)
I’m not sure that they used LLMs to write the entire paper but they do have a disclosure at the top of the PDF stating that ChatGPT etc were used to help with the literature review and improve readability. I agree that it’s a bit ironic given the topic of the paper, and personally find the overly-verbose, hollow prose of LLMs less readable than almost any human writing.
After years as a student and teacher over the past few years, you develop a sort of spidey-sense for this kind of thing. I am aware that intuition and past experience (attempting to generate and marking such papers) is not sufficient grounds for deeming an entire paper (and I am very cautious to accuse a student/article of relying solely on AI), so I have listed some of the biggest giveaways below, in increasing order of damning capability.
The paper uses AI buzzwords frequently in contexts that are not used in typical norms of academic writing. For example, “delve” is used three times, the infamous em dash, and the sentence structure "X is not just Y, but also Z is beaten to death. Hallmarks of AI-generated text, especially when they show up far too often, but perhaps it was written by someone unfamiliar with academic writing. We move on.
“… we (the authors—not the omniscient ‘we’ that, yes, has snuck into this very chapter) analyzed phrases associated with less emphasis on memorization in published literature over time.” Attempting to crack jokes/be cute is the calling card of the terrified undergraduate, or an LLM. You decide.
The complete and utter lack of citations. No source is provided for such claims as below. A quick search through PubMed brings up this source claiming the exact opposite: alcohol consumption during encoding harms memory recall while it improves memory recall during consolidation. (In other words, for you legal drinkers, have that glass of wine after your daily reviews.) I am not saying that this paper is the sole literature on alcohol’s effects on cognitive performance, but in the absence of a source, you have to wonder if this is just an LLM attempting to match its output to a list of pre-provided sources (provided by some other LLM) or, giving them the benefit of the doubt, a very confused intern attempting to meet a word count. The lack of citations is an issue of LLMs, yes, but also an issue with scientific rigor, which is what I am trying to get at here. Even if it was written by a human, this human is proving to be largely oblivious as to what makes a proper commentary on the current literature (which is what is attempting to be done here, but the lack of literature makes it difficult to judge). I have an issue with the large jumps between evidence (if provided) and conclusion (often wildly different, or wrong, as below).
“A related phenomenon is context-dependent learning: if you encode
information while drunk, for example, you can recall it better when drunk than when sober. This phenomenon demonstrates how context affects memory retrieval—though it’s not a strategy educators would recommend adopting.” (NO SOURCE)
I have an issue with the use of IQ as a metric. It is not a standard metric we can rely on. Looking at the Norwegian data of an observed inverse Flynn effect from ~1960 to ~1990 the paper references, we see that IQ was derived from the “aggregate stanine score given each conscript based on three speeded tests of arithmetic (30 items), word similarities (54 items), and figures (36 items)”. You might see the problem here. If you shift the curriculum away from learning these specific skills, it will be no surprise when test scores drop. The Danish data relies on letter matrices, verbal analogies, number series, and a geometric figures test. One has to consider that both of these are based on military conscription data. None of the possible issues of skew are discussed. This is another issue. In attempting to make a claim one must consider the counter-arguments, which this paper fails to do.
And so on. It doesn’t help the authors’ case that when putting in any segment of the paper into an “AI Checker” (I am aware these are never the final verdict on these things, but compounded with the above evidence I have reason to believe it) a result of 100% AI is returned.
Either way, this is not a very well-written paper, and while the premises are interesting, the fact that the only actual ‘analysis’ done is a single subsection on pages 29 and 30 was a sentiment analysis of the word ‘rote’ in the COHA (with a beautiful Google Sheets graph, but that is just a personal peeve) does damage to its credibility and the extent to which it can be taken seriously. Always exercise caution with pre-prints.
No shock to anyone who’s read my comments in other “the kids aren’t alright” threads but I have all the quibbles @vermin had and more, lol. I’m not convinced by any of the claims made, and even if I were I probably wouldn’t be compelled to any action.
Plato was quoted upthread… he was clearly a smart guy but if the tradeoff for “losing our cognitive ability” is our current level of technology and continued progress…I’m on team ‘hell yeah.’ Like, he thought the moon had a soul. We stepped on its surface. in the two-and-a-half millennia since he published Phaedrus, we’ve completely rewritten the world’s technology how many times… 3? 4? You show that guy a lightbulb and ask him if he still thinks writing is gonna make us all stupid, let alone calculus or a steam engine or an AUTOMOBILE or Evangelion or an iPhone or, yes, ChatGPT. The people of that time would have worshipped it (I mean… even more than the people of our time do.)
Could Plato and his pals (meaning people with nothing to do but “think” all day) beat me at speed cards? I mean, maybe. Definitely, if they practiced, and also someone told them what speed cards (and cards) were. But in the modern world most of us are infinitely better off than we would have been then and it’s mostly technology to thank. Are we on the earth at exactly the right time to stop inventing things, that it would be better for the future of humanity’s collective brains? Feels unlikely.
The IQ scores translating to assess human intelligence is in itself questionable. It is similar to many questioning GRE exam to find person’s aptitude for grad education. Many univirsities waived GRE as prerequisite.
The thing of brain seem to be it’s adoptability. Remember when you dad/mom or anyone in community could remember atleast 20 phone numbers without any problem. When save feature came to the phone, the memorizing capability of the parents are only slightly better to ours.
Many of these research papers highly funded by corporates have grandiose claims but when you see their research method it is not even above average in many situations. If you have a knack for studying research paper, you realize some university guy’s research paper which was partially funded by university and mostly covered by his/her part time job along with university studies utilize most regorous methodoloy with consistent results closer to truth.
The children and very young youths generally do not fall for many youtube videos developed by sora and say it is AI generated outright without second thought, which if I show to my parents would not only believe but have strong emotional response to it.
Same will be of AI. The novelty and hype to the AI will wear out soon. Why is this on hype? It is because it has greatest financial investment of humanity. Just do a google research and you will see for yourself how much investment has been made. I myself left chatting with AI and am active in this forum because these convrsations are more relatable to me. People have their own view on it. The most important to me is their capability to suprise me. They have one of the most original answer to the questions I put. If I make such conversations with AI, the answers are always agreeable and so shallow. The enjoyable part is only on the first few chats because we seem to get newer perspective to our research.
This could be highly opiniated but I do not consider research of Barbara even remotely good. I have studied many of her research. The youths have never failed to amaze me with their approach to life.
To go back to Phaedrus, I think I have read it argued (I can check the commentaries and books I’ve used, as well as the text itself to see if this reading is justified) that the point was more that really learning can’t be just a passive process of consuming information, say through reading, that we don’t even store within our minds or potentially even fully understand, but that debating and examining the points made with another person is necessary to reach either truth or versimilitude, whichever one believes is possible. I think the issue of whether an author might be misunderstood after he dies and cannot defend or explain himself might have been a part of this.
That might not be relevant, but perhaps it is - we don’t have to be opposed to all new technology, but we shouldn’t forget how to memorize or think by letting something else do it for us. We should be cautious, not all for or all against.
Also, technology as defined above is too broad a category to make grand statements - it’s not good or bad; some of it is useful, some of it does more harm than good. Lightbulbs are nice, but are we so much better off with iPhones? What do we do on iPhones that is really necessary that cannot be done on a flip phone or paper? The only things I can think of are the expense of international calls or texting if you can’t use a messaging app (and have friends or relatives in other countries) and the problem of reading, say, Greek on public transport without an online dictionary (a decent physical one is probably to heavy for everyday commuting). What about the distraction and disconnection caused by smartphones?
Is the convenience provided by some types of technology good? Too much convenience harms resilience, I think, and makes some people act a bit crazy when they lose access to something, like Google maps or clothes-drying machines, that they have become reliant on.
This. We know that merely rereading notes is the worst way to review (if it can be called review at all). What we need is recall.
If students are going to just reread the textbook before exams and fool themselves that they have studied, then saying “reading/writing is bad” makes sense in that context. Because it allows the students to be busy without actually learning.
It’s very easy to criticise people from older times saying “ha ha look how dumb/ignorant they were” to make ourselves feel better and patting ourselves on the back saying how much better our society is now. But often it’s because we miss the point and don’t actually understand what they said and why.
And also because we tend to forget that more knowledge doesn’t equate more intelligence/wisdom.
For every new tool, it makes a divide between the people who can learn to use it well, and the people who can’t or won’t learn to use it well.
Same for the internet. It opened an amazing treasure trove of free ressources to learn countless new things. But the idea that I don’t need to memorise anything because I can always Google it, doesn’t do anybody any favours. Not to mention that it means that I will tend to accept as true the first things the algorithm decided to push forward. How many people actually check the sources of an internet search? How many people have the basic knowledge to evaluate these sources?
It’s like AI, it’s an amazing tool that is going to be a huge help if you know how to learn to use it, how to prompt it and how to check the results. But if you don’t already know how to do that, it will only make you less efficient and get worst results. And a lot of people are not at all curious to learn how to use it well (some because they’re not even aware they need to).
So the most educated/curious people are becoming better, and the others are left behind.
This has little to do with intelligence (and age) but mostly it’s about (access to) education and the drive to learn and improve.
And modern life is dangerous in that regard because often we are comfortable enough that we are not motivated to strive for more (more than just “consuming” technology and knowledge).