This long-form article requested by Fair Observer summarizes the wisdom of decades of neuroscience research as ten "Principles of Plasticity," which govern human learning and machine learning alike. Together they suggest, nay demand, that children's growing nervous systems be protected from artificial interference, especially by mentally manipulative technology.
Monetizing Children’s Brains Means the End of Our Species
Ten basic principles of human and machine learning, also known as neuroplasticity, explain which kinds of online “learning” damage children’s brains, and how. By William Softky
My editor at Fair Observer, Atul Singh, told me of his own friend Pankaj, a father of three daughters. Pankaj was concerned that his daughters have been subjected to three to seven hours a day of online classes, and felt it was deeply wrong. Atul invited me to write an article addressing Pankaj’s heartfelt concern, concerns I know are shared by parents worldwide. Children are precious, and their nervous systems are delicate enough to conform to what goes into them. Is online exposure harmful to kids?
For a decade I researched with media theorist Dr. Criscillia Benford, my wife, the problem of how brains react to artificial inputs. Her explanation of how business incentives and psychological vulnerabilities interact in “educational technology” [coming soon, linked here] agrees entirely with the neuroscience news I have, because they spring from our unified, quantified understanding.
This article is long and dense because it provides what anyone who cares about kids—parents, teachers, governors--needs to defend children’s budding brains against this summer’s looming threat. If these ideas make sense to you, a yet better understanding comes from a second reading, on paper in good light, pencil in hand. That’s the whole point: brains work best hooked directly to the senses, with time to digest.
Abraham Maslow’s famous “Hierarchy of Human Needs,” updated using biophysics to show how humans’ most urgent needs are structurally simple (chemical and physical, lower left), while our more long-term needs, built upon those, are structurally complex (biological, sensory, and interactive, upper right). Complexity is where neuroplasticity matters.
Decades of experiments have outlined several general principles governing learning, which together we neuroscientists call “neuroplasticity,” principles based on the presence, quality, duration, and order of sensory inputs and sensory-motor interaction. Unfortunately many of us are muzzled by obligations to colleagues, employers, and investors. At the moment I have no such obligations, so I am free to give you ten widely-recognized principles of learning, along with what those principles say parents and educators must do, immediately, to protect our children and their descendants. But first, allow me to praise our human species.
We are the most elaborately, intimately social animal ever to roam the planet. Dog-eat-dog doesn’t hold for pack animals, like dogs, whose nervous systems perform synchronized hunting to survive. Humans dominate the synchronization Olympics. Our distant runners-up are bonobos, constantly cuddling and caressing in countless ways, collaborating, conspiring, and cooperating through high-bandwidth neuromechanical interaction among all possible pairs. Their hands and bodies are built for touch, and their brains for processing it. Humans have all bonobos’ physio-social circuitry, and then some: white eyeballs to show where we look, bare skin sensitive to bare hands, perfect balance for expressive dance, octaves of various vocalizations for singing, intricate facial expressions, and eyebrows to amplify them. Forget words, tools, and symbols. Homo sapiens excels at continuous connection.
Neuroscience studies animals because animal nervous systems are like human ones. That essential similarity is why we perform experiments in the first place. We already assume by default that what applies to cats and dogs applies to us. The principles of neuroplasticity below were discovered fifty years ago through grisly experiments on kittens. Since then, during the computer revolution, they have been rediscovered in various mathematical forms by physicists, signal-processing engineers, data scientists, and machine learning specialists.When human brains acquire skills, we say they “learn.” When autonomous computer models like neural nets and self-driving cars acquire skills, we say they “self-calibrate.” But the underlying principles are the same.
The successes and failures of both learning and self-calibration depend on data quality, speed and timing in straightforward ways, ways that ultimately originate in laws of information flow through space and time. The mathematical generality of those laws means that “neuroplasticity” concepts are universal, applying everywhere in the universe to all self-calibrating instruments, even artificial ones. If robots existed on other planets, these laws of neuroplasticity would explain how they learn too.
Maslow’s same Hierarchy of Needs as above, here quantified.
Below are ten principles of learning, followed by guidelines on their use defending children everywhere.
1. Critical periods are critical. Period.
Babies have to learn to walk before they learn to run, to communicate before they learn to read, to grab before they learn to write, and to hug before they learn to love. Learning takes time, and must build on previous, simpler learning.
Each little chunk of the brain is hard-wired to learn from specific inputs. If those inputs don’t show up in time, during a “critical period” for acquiring that skill, the brain gives up on those inputs and makes do with other ones. That’s actually a good strategy. A baby born with a bad eye, for example, will learn by age two to use only the good one.
The strategy can backfire if an otherwise-good eye isn’t aimed right, as with “lazy eye” or “wall eye,” in which case the brain learns to ignore a functional but mis-aimed eye. A typical treatment is to cover the good eye, a therapy forcing the brain to learn to use the weaker one. During critical periods, it is important not to deprive the brain of its native inputs, and especially important not to replace them with anything worse. As athletes and coaches know, it’s at least ten times harder to un-learn a bad habit than to learn it right from scratch.
Critical periods contain and constrain the essence of neuroplasticity, by turning the often-vague term “developmentally appropriate” into very specific guidelines regarding the presence, quality, duration, and order of sensory inputs and sensory-motor interaction. The shortest critical periods, say for vision in kittens, take weeks. The longest ones, for emotional nuance in humans, take decades, well into young adulthood (see this articleon The Brain’s Emotional Development).
Learning after a critical period is not impossible. But the longer after the critical period a child goes without acquiring full function, the harder the child will have to work to improve, and the less improvement it will make.
2. Primary processing precedes perception
The simplest, most granular processing is the processing closest to raw input, such as edge-detection among pixels in computer vision, or their human equivalent, “orientation columns” in the primary visual cortex of mammals. The general strategy is to learn these “low-level” features first and fast, harvesting the most information from the tiniest pieces using the highest bandwidth. More abstract, broader, slower features build upon those, so of course they are learned later, and more slowly. Or not learned at all, if the low-level primary features weren’t learned right first.
The fastest and most granular low-level computations, denoted in microseconds and micrometers, occur a hundred thousand times faster than conscious thought. We need that deeply unconscious raw data most, yet by its nature it is easiest to damage, and impossible to sense directly.
3. Sensory fusion is the rule
The structure of brains is simple. Most brain regions and micro-regions (“cortical columns”) share the same five-layer structure and all nerve fibers transmit the same pulses. The sensory wires aren’t labeled by where they came from. Because a piece of brain can’t tell which input came from where, sensory input from different sources is inevitably mixed together into a single, coherent, multi-sensory perception. If any sources or senses are missing, or mis-timed, the integrated whole will suffer.
4. Learning requires constant objects
Those of us with brains know they make sense of the world, and those of us who live in the world know it is filled with three-dimensional objects.Babies learn about individual objects by moving their eyes and fingers (and maybe lips and tongue!) across them as the object stays still. If the object itself moved too fast, the baby couldn’t learn about it at all. Still, once it’s out of range, the baby may forget.
After a year or so babies learn that objects continue to exist even when not observed. This crucial discovery is called “object constancy,” a foundational stage of cognitive development. If for some reason a child never learns this fact—for example, if instead of looking at a live face, they looked at face-shaped pixels—they would never learn the difference between pixels and real life, and could never learn to trust their senses. Babies can’t yet tell the difference between broken-up representations such as images and continuous real-live objects right in front of them. Exposing babies to both, intermixed, impedes learning object-permanence.
The brain’s basic need for truly continuous targets means a baby couldn’t learn to perceive motion if it lived in a stroboscopic world. Or in screen-world, which flashes ten times faster, and is also flat and pixelated. Screen-based images are not objects, they are blinking dots, micro-timed to provoke object-recognition in mature visual systems, yet utterly incapable of calibrating immature systems. Do check this reasoning with someone who programs self-driving cars. Please don’t experiment on your children.
5. Perception comes from “Serve and return” timing
Trust in one’s senses comes from interaction, not transaction. (My wife and I wrote the singular peer-reviewed article about the physiological basis of trust,Sensory Metrics of Neuromechanical Trust).
At the finest internal level, every mammal brain sends out pulses and gets pulses back. It uses the out-and-back pulse timing—with exactness down to microseconds, thanks to temperature-stabilized brains—to make three-dimensional pictures of a target region of space, whether containing muscles inside the body or surfaces outside it. A brain’s precision in spatial mapping is directly proportional to the timing precision of its pulses.
In psychology this autonomous back-and-forth interaction, say a smile initiated by the baby and reflected by the mother, is called “serve and return.” In neuroscience, it’s called “sensorimotor contingencies.” In medicine, it’s called “biorhythm synchronization.” And in spatial imaging, it’s called “time-domain ultrasonic tomography.” All represent the same sensory algorithm, but at different timescales.
6. Big brains need play
“Serve and return” only works if the brain gets to choose when to serve. In other words, implementing a brain’s learning algorithm requires timing autonomy. The brain fiddles with its sensory-motor world in order to detect significant returns. The exact timings of its micro-actions, like private keys in cryptography, let the brain distinguish what it caused from what would have happened anyway. For immature mammals, this exploration looks like play. Play is children’s work. (Such work, including make-believe, is harmed by commercial coercion, as described in the book Consuming Kids).
The better a brain becomes at predicting returns, the smaller and more frequently it serves. Practice and success dial down amplitude into the deep subconscious of the microsecond realm. This virtuous cycle starts in the womb as the baby learns to anticipate the mother’s heartbeat, voice, and body tremor, ultimately synchronizing its biorhythms with hers. Ideally the two signals match, a sonic synchrony undergirding the mimicry humans do so well later in life.
In fact, the same continuous active synchronization underlies all animal society.Gnats swarm, fish school, birds flock, packs hunt, mates dance, and humans cuddle, play, wrestle, cry, laugh, and sing. The tiniest flickers will always synch first, but they’re damaged by digitization the worst.
7. Natural appetites need natural statistics
The principles above, which derive from the physics of matter and energy, tell us brains need continuous physical targets nearby to return them good sensory data. The principles below, regarding attention and motivation, derive instead from a brain’s hard-wired expectations of what will be common vs. rare as it grows up.
In particular, brains are hard-wired to seek rare things to make sure we get enough. Our native appetites for sugar, fat, and nutrients are well-known, but our native appetites for certain informational patterns--saturated colors, shiny things, novel shapes, sudden changes, sharp contours, and recognizability in general—also correspond to their rarity and usefulness in evolutionary times. Those expected (and formerly rare) patterns are collectively part of the “natural statistics” of the environment the brain was born to explore. Our senses enjoy “interesting things” because they are rare in Nature. But if saturated, a brain becomes desensitized and dependent, as if in thrall to a kind of addiction (as explained in Sensory Metrics).
8. Dopamine drives decisions
Dopamine is the motivating neurochemical which makes us want to learn, by rewarding difficult predictions. But as theoretical biologist Thomas Hills proved in 2006, the principle of dopamine pre-exists the chemical. Even bacteria have circuits rewarding successful prediction as they move about, and dopamine descended from those. This computational purpose of dopamine is to focus resources and make decisions. By design, patterns of inputs which capture dopamine circuitry control the nervous system via its most basic motivations. If the controlling pattern comes from an organic animal or inert object, we say the nervous system is “learning useful habits.”If the pattern is artificial, we say the nervous system is “being hacked.” Utter defeat.
Dopamine can be elicited, or rather administered, by countless artificial stimuli from drugs to porn to gambling to video games to social media. But dopamine only works as designedwhen its owner moves autonomously in a natural environment.
9. Data asymmetries permit parasitism
There was informational warfare long before humans, perception vs. deception. Colored plumage, colored flowers, lures, camouflage, and bandwidth arms-races.Whichever creature has more, better, faster information always gets to manipulate or prey upon the others. We humans have always been the ones to conquer or domesticate other animals. Now, by the numbers, edu-tainment-tech is domesticating us.
10. Covert biometrics = adversarial biofeedback
The most beneficial bio-measuring technologies, such as cochlear implants, heart-rate monitors, and stress biofeedback, provide the user with real-time, high-quality, unbiased data streams their nervous systems could not ever provide. Those signals boost the user’s self-knowledge, and remain under the user’s control. The worst forms do the opposite, secretly monitoring heart rate, blood perfusion, pupil dilation, typing speed, emotional affect, target of gaze, stress levels, etc., then using those signals to replace natural ones or subtly alter the user’s experience. The signals are far worse than useless, because they actively impede perceptual learning. Not only do corrupt return signals lower perceptual accuracy all by themselves, but their unpredictability destabilizes the algorithm for learning object constancy. Exposure to such signals makes self-awareness, motivation, and choice progressively impossible.
The principles [1-10] above directly answer many important questions about what happens when young humans are exposed to artificial input, including and especially educational technology. Since COVID has throttled most human interaction down to screens, parents are rightly concerned. It’s worth stating these heartfelt human questions and their answers using the uncontested neuroscience principles outlined above.
While the principles have been known for decades, their implications for growing minds are only recent. Furthermore, most of us are embedded in cultures or situations where this unambiguous advice is impractical, much like medical advice to avoid chemical additives and preservatives can be impractical. I personally regret how I exposed my own kids to technology, just a decade ago, before I knew better. And in my own life I still can’t follow all the advice below. So the following advice does not judge any given parent or parental choice. Instead, it provides for decision-makers a clear, compact, unambiguous description, directly traceable to decades of scientific certainty, about which inputs are healthy vs. harmful to growing minds. No sugar-coating.
How can I know if my kid is OK?If your kid can remain happy and functional interacting socially with adults and/or peers for 24 contiguous hours each week, with everyone completely tech-free, they’re fine. Any dependence on tech, being driven by dopamine, will share the typical symptoms of dependence and addiction: malaise, tantrums, skulking, wheedling, compulsive use, impaired reasoning, lying.
Which ages are safe for screens, and for tech in general?No child should be exposed to artificial stimuli before the skill is established organically and the critical period has long passed. Thus, even the most visually skilled child should avoid screen exposure until about age four, and then only for picturing real things; should avoid cartoons and video games until they perceive the three-dimensional mechanical nuance of real live people and animals, say age eight; should avoid interacting with people by video until they master emotional face-reading, say age twelve; and should not manage emotional relationships (like romance) remotely until they can in real life, say early twenties.
The learning of dexterity follows similar rules. A child should establish fine-motor coordination first, using paper and crafts, before learning to write, say earliest age seven; should write easily and well before learning to type with moving keys, say earliest age ten; should type well before using any touch-screen, say earliest age thirteen; and should not operate computers by automatic motion-capture or face-reading until adulthood, if at all.
The above rules assume the technology is optimized for human benefit (in terms of quantifiable results and not just good intentions, for example by minimizing all artificial enhancements, even video edge-enhancement and cartoonification). But finding such ideal tech is rare.Absent that, the most general observations possible about kid-safe tech are, as before, based directly on the neuroscience principles [numbered in brackets]: immature nervous systems will become mis-wired by technology which  separates image, sound, and/or touch, which  generates artificial stimuli, which  superimposes timing variations on human outputs, which  undermines or coerces autonomous outputs, which  synthesizes unrealistic reward profiles (e.g. “gamification”), which  elicits dopamine responses, or which [9, 10] does not display all monitored biometrics directly to the child, and preserve them for the child’s benefit alone.
Examples of “bad” educational technologies can easily be found by reading ads proclaiming how they elicit dopamine or make assessments using biometric signals. An example of a “neurosafe” technology could be a video/audio recording or teleconference platform with stereo sound and high-definition video, each separately registered to reality, and to each other, at microsecond resolution. Neurosafe technology preserves and protects maximal human bandwidth in native human format, not in formats convenient for monitoring, manipulating, or monetizing.
What is the best thing I could do for my kids right now? Get your whole family away from all technology for at least 24 continuous hours a week, then try to expand.Reminding yourselves collectively what real life feels like is the first step in reclaiming it. The adults, being grownups, have to go first, first because kids imitate, second because only we adults have both ability and willpower. With your family, try any kind of fun active play. That is, any attractive, symmetrical neuromechanical interaction, such as back-and-forth, billiards, badminton, baseball, or banter. Those provide more benefit than “games” which move symbols on boards according to rules. Which in turn are still better than tech. Inventing brand-new forms of voice-based fun, say laughing games, is best of all.
Triple-bonus points for massage, especially scalp massage, which can even be done in quarantine. Kids can learn to massage as sensitively as adults, and competence in giving pleasure is the best skill a human could ever learn. In particular, fingernails pressed gently into or dragged along the midline—brow, crown, occiput, sternum, lower back—help wake up the most emotionally central myo-fascial pathways on a human body. Supportive touch both relaxes and feels wonderful.
Never give kids tech to keep them busy. Their nervous systems need your live attention, and nothing else will do. Read to them, play with them, talk to them, do chores together.
What kind of education is best for my child?Small-scale (mesoscale) learning, like “home-schooling” by attentive tutors, gives groups of children a good social experience. Small-scale learning can also be educationally optimal, but only if it maximizes each child’s ability to confront, appreciate, and autonomously interact with the real diversity of the physical world and the people in it. (As opposed to the opposite motivation, protecting children from dangerous people and ideas). Unfortunately the modern caricature of “traditional values” mocks the constraints imposed by honor, family, integrity, and respect, providing nothing in return.That view forgets those attitudes are essential for human social function. Societies don’t last long without them.
A handful of internet sites, like Fair Observer, keep honor and integrity alive by avoiding the corrosive influence of ads and corporate sponsorship. Otherwise the old innocent internet is dead, replaced by bots, incentivized propaganda and surveillance. In such a corrupt environment, nothing worth asking can be honestly asked, nor honestly answered. Fake-ness reigns.
Yet real life still exists, and feels wonderful. The more you and others like me rediscover real life, the more time you will spend there. Therefore the less of your life you will spend online, or recording it for online consumption. Collectively, that out-migration will leave “the internet” as before, filled mostly with bots and humans posting criticisms to provoke yet more online discussions. Sadly, technophiliacs and neuroscientists alike share an allergy to simple, unambiguous conclusions, regardless of obvious truth. Like the internet, both those groups cheer discussion for its own sake in an endless loop.
What can schools do? Proponents of ed-tech love open discussion, but only as long as their technology is presumed safe by default. Now, neuroscience has turned the tables. The medical burden of proof lies in meticulously proving how tech could be safe in the first place. That will prove difficult, because all reasonable evidence-based investigations so far agree with fundamental neuroscience, fundamental data science, and plain parental common sense in concluding that technological interference with organic brain development is dangerous at best, toxic at worst.
Meanwhile, schools and their vendors are suddenly commencing a bold new experiment on children’s brains en masse, one with irreversible results. Since proponents brag that tech affects minds, then technology must be evaluated like a drug. In America, drugs must be proved both safe and effective before being dispensed. Let proponents prove their case according to medical safety standards, not bureaucratic line-item calculations. If ed-tech were a drug, they can show why it should not be banned.
Now let me warn our species.
This summer, an unholy alliance of risk-averse spreadsheet-bound bureaucrats and risk-loving, pixel-pushing Disaster Capitalists is poised to take our schools, and thus our children’s brains, by storm. It doesn’t matter if the evil is in the people or in the spreadsheets they obey, the result would be catastrophic for homo sapiens.
Never in human history has an entire generation of innocent young humans, rich and poor alike, had the nipple of loving neuromechanical connection yanked from their mouths, to be replaced by a dripping pap of pixels laced with dopamine, administered under glass. Never before have games been gamified and children’s play been played, for profit. Never before have the unfurling neurons of babies been clear-cut for short-term revenue and long-term prediction. Never before have all our young ones been experimented on with profitable mind-altering drugs and profitable mind-altering technology, at once. Not even once before in history has any single parent, much less an entire generation, looked into screens instead of into children’s eyes, then watched in horror as the children imitate that disconnection for the rest of their lives.
The disaster is more potent than attractive technology undermining attraction, more chilling than cold spreadsheets overruling warm, live hearts. It is the worst possible calamity to strike the most elaborately, intimately social animal ever to walk the planet, we humans who love continuous connection. It is an infection of communications itself, an infection whose violent, virulent, viral spread is fueled by data, dough, and dopamine. Left unchecked, this infection of disconnection will fill the world with psychic orphans (like the abandoned Romanian babies), impervious to love and unaware that it exists. Such cold souls could not nurture a last generation.
Please prove me wrong.