Superstimuli: The Ingenious Deception of our Senses
A horizon scan on hyper-engineered stimuli, what three years of evidence and regulation have done to the picture, and what we are watching next.
In the 1950s the Dutch ethologist Niko Tinbergen put a thin red stick with three white bands in front of herring gull chicks. They pecked at it harder than at an accurate model of their own parent's head. He found the same thing again and again. Oystercatchers abandoned their own eggs to sit on a giant dummy one. Male grayling butterflies, across something like fifty thousand field tests, preferred black cardboard to actual females.
Tinbergen called these supernormal stimuli. His point was narrow and precise: an animal responds to a few abstracted features and not to the real thing itself, so an artificial object that exaggerates those features can beat the real thing at its own game.
Deirdre Barrett carried the idea into human behaviour in her 2010 book Supernormal Stimuli. Eliezer Yudkowsky gave it the shorter name and the economic framing in a 2007 essay, arguing that competitive markets ratchet products a few percent more compelling with every iteration, and that willpower is a finite resource being asked to hold a line that keeps moving. He did not cite Tinbergen, and most people who use the word now have inherited that gap.
The thing to keep from Tinbergen is the mechanism. A superstimulus works because it is more than real, along the specific dimension the response evolved to track.
Food, and the myth of the formula
Food is where the engineering is most deliberate, and where it has been going on longest.
The bliss point is Howard Moskowitz's term. It describes the concentration of a single sensory attribute, most often sugar, at which liking peaks: too little and a product is bland, too much and it is cloying. It is measured product by product, attribute by attribute, and often segment by segment. It is a method for finding an optimum, and it is not a recipe.
So there is no single formula, and the comparison people reach for does not hold either. Mature human milk takes about 48% of its energy from fat, 44% from lactose and 7% from protein, and that carbohydrate is lactose, only mildly sweet and a long way from the sucrose and glucose syrups in a packet of crisps.
What is true is more prosaic and more effective. Food scientists optimise combinations of fat, sugar and salt for maximum palatability, product by product, and they are very good at it. Human milk is fat-dominant. Our responses to energy-dense combinations were built for a world where energy density was scarce and is now abundant. A packet engineered against all three lands on a system with no upper stop.
What the health picture actually looks like
The IDF's 2025 Atlas puts 589 million adults aged 20 to 79 living with diabetes, roughly one in nine, projected to reach 853 million by 2050. Forty-three per cent are undiagnosed, and four in five live in low- and middle-income countries.
On weight, WHO gives 2,5 billion adults with overweight, which is 43% of everyone aged 18 and over, and 890 million with obesity, 16%. Adult obesity has more than doubled since 1990 and adolescent obesity has quadrupled. A Lancet forecast published in 2025 puts adults with overweight or obesity at 3.8 billion by 2050, around 60%.
Non-communicable diseases killed at least 43 million people in 2021, about three quarters of all deaths outside the pandemic, and WHO names unhealthy diets among the behavioural drivers. The eleven million diet-attributable deaths that get quoted alongside this come from a 2019 Lancet study using 2017 data, so the number is worth dating whenever it is used.
Attention, where the evidence is contested
Screens are the version people argue about, and the argument has grown sharper without growing clearer.
Riehm and colleagues, in JAMA Psychiatry in 2019, found that adolescents using social media more than three hours a day had about a 60% higher risk of internalising problems, rising to roughly double for the narrower category of combined internalising and externalising problems. The shorthand that three hours doubles mental health problems comes from the second figure being applied to the first, a compression that entered circulation through the US Surgeon General's 2023 advisory and has been travelling ever since. The underlying data are observational, collected between 2013 and 2016, and predate TikTok.
Eighth and tenth graders averaged 3,5 hours a day on social media in 2021. Higher figures, in the five to eight hour range, describe total screen time, which is a different measure and a different argument.
The field itself has polarised rather than converged. Jonathan Haidt and Zach Rausch argued in the World Happiness Report 2026 that population-level effects are now visible in national data, drawing on natural experiments in broadband rollout and randomised trials of sustained reduction. Candice Odgers, Andrew Przybylski and others maintain that the causal case is weak, the effect sizes small, and the moral panic a distraction from other drivers. A 2026 review concluded that associations are consistent but modest, that total time online predicts little while passive consumption predicts more, and that most of the evidence remains observational.
The harm is plausible and contested at once, and policy has moved considerably faster than the evidence.
Three things that have changed since 2023
A pharmacological answer arrived. GLP-1 receptor agonists have gone from a diabetes drug to a mass consumer product in three years. Gallup found 11% of US adults currently taking one for weight loss in mid-2026, up from 3% in 2024. The FDA approved an oral semaglutide for obesity in December 2025, and a US pricing deal in November 2025 cut monthly costs from roughly a thousand dollars to the low hundreds, with Medicare covering obesity indications for the first time.
The pattern of what people stop buying is the part that bears on this scan. Cornell research published in the Journal of Marketing Research in December 2025 found households cutting grocery spending by 5,3% within six months of starting a GLP-1, and the steepest falls were in savoury snacks and sweets, at around 10%. The same drug class reduced heavy drinking days from nine to five a month in a Lancet trial published in April 2026. One intervention, acting on the reward pathway generally, moves both food and alcohol. That is about as close to a test of the superstimuli hypothesis as anyone was likely to get, and it cuts both ways: it supports the mechanism and it suggests the fix may turn out to be pharmacological instead of environmental.
Regulation moved on minors, and the first results are not encouraging. Australia's under-16 social media ban passed in November 2024 and came into force in December 2025. An eSafety Commissioner study reported in August 2026 found more than 80% of under-16s still using social media, with daily use essentially unchanged, and about half saying platforms never verified their age. The UK, France, Brazil, Indonesia, Malaysia, Denmark and Greece have all legislated or announced similar measures. The policy wave is real. The evidence that it works is, so far, not.
Addictive design became a regulated category. In July 2026 the European Commission issued preliminary findings that Meta's platforms breach the Digital Services Act through addictive design, naming infinite scroll, autoplay, push notifications and personalised recommenders, and arguing that Meta failed to assess the risks to users' physical and mental wellbeing. These are preliminary findings and not a ruling. Alongside them, the European Parliament has called for a default minimum age of 16 and bans on infinite scrolling, autoplay and gamification for minors, and a Digital Fairness Act covering dark patterns and addictive design is expected as a Commission proposal in late 2026.
On food, the Lancet published a major ultra-processed food series in November 2025, finding 92 of 104 long-term studies reporting increased chronic disease risk and calling for labelling, marketing restrictions and taxation. Chile's warning-label law remains the best natural experiment available, with household sugar purchases down 20% against the counterfactual. The UK's ban on HFSS advertising online and before 9pm became statutory in January 2026. It is worth noting that the NOVA classification underpinning much of this is itself contested, on the grounds that the ultra-processed category is too heterogeneous to work as a causal unit.
The signal we are watching
The thing that interests us most is not any single one of these developments but the shape they make together.
Three years ago the available responses to superstimuli were individual restraint, regulation, or resignation. All three now look different. Restraint has a pharmaceutical substitute that roughly one in nine American adults is already using. Regulation has arrived faster than its evidence base and, on the first real test, has been widely circumvented by teenagers. And the environments themselves are being formally investigated as engineered, and that changes the argument from personal failing to product safety.
What might follow is a sorting that runs along lines we have written about before: one population living comfortably inside optimised environments with pharmacological assistance, and another building deliberate friction into their lives through restriction, biohacking and analogue substitution. Neither is a good story on its own, and the interesting question is whether the sorting turns out to run along economic lines, because that would make it a question about who can afford to opt out.
What we are watching next
Whether GLP-1 effects hold outside food. Alcohol has trial evidence. Gambling, nicotine and compulsive shopping have signals and no adequate trials. If controlled evidence appears in a second non-food domain, the general-reward mechanism moves from plausible to established.
Whether any under-16 ban produces a measurable behavioural change. Australia is the test case and the first read was poor. Watch enforcement design and age assurance rather than the headline legislation.
The outcome of the EU's addictive design case. A final decision against Meta would make interface design a regulated safety matter across the bloc, which is a larger shift than any age limit.
Whether "ultra-processed" survives as a regulatory category, or whether the heterogeneity objection defeats it and policy falls back on nutrient thresholds.
Price and access to GLP-1s outside the US. If the sorting above happens, this is the variable that decides whether it runs along income lines.
Reformulation as a business signal. Food companies shifting toward protein and fibre are telling you what they expect demand to do.
Originally published in 2023. Revised September 2026.
Sources
N. Tinbergen and A. C. Perdeck, work on supernormal stimuli in herring gulls, oystercatchers and grayling butterflies, 1930s to 1950s
Deirdre Barrett, Supernormal Stimuli: How Primal Urges Overran Their Evolutionary Purpose, W. W. Norton (2010)
Eliezer Yudkowsky, "Superstimuli and the Collapse of Western Civilization", LessWrong, 16 March 2007
Nicklas Brendborg, Vanedyr, Gyldendal (2023); English edition Super Stimulated, Hodder & Stoughton (June 2025)
IDF Diabetes Atlas, 11th edition (2025)
WHO, Obesity and overweight fact sheet (updated December 2025) and Noncommunicable diseases fact sheet (updated September 2025)
Riehm et al., "Associations Between Time Spent Using Social Media and Internalizing and Externalizing Problems Among US Youth", JAMA Psychiatry (2019)
US Surgeon General, Social Media and Youth Mental Health advisory (2023)
Haidt and Rausch, World Happiness Report 2026, chapter 3; and Candice Odgers, review of The Anxious Generation, Nature (2024)
Hristakeva, "The No-Hunger Games: How GLP-1 Medication Adoption is Changing Consumer Food Demand", Journal of Marketing Research (December 2025)
Semaglutide and alcohol use disorder trial, The Lancet (April 2026); and Hendershot et al., JAMA Psychiatry (February 2025)
The Lancet series on ultra-processed foods (November 2025)
European Commission, preliminary findings on addictive design under the Digital Services Act (July 2026)
Australian eSafety Commissioner, evaluation of the social media minimum age (August 2026)
