Last night, somewhere around 11 pm, you probably did something your nervous system filed under danger. You read the news, or scrolled X, or checked your phone after a notification arrived from someone you have never met and will never meet, and felt a spike of dread or agitation, a low hum of threat that took longer to leave than it had any right to. You were safe, you knew you were safe, and the knowing made no difference.
This is not an anxiety disorder or a weakness, and it is not peculiar to the digital age, though the digital age has made it considerably worse. For most people it is the predictable output of a brain doing exactly what it was built to do, in an environment so different from the one that built it that the machinery is running in permanent misfire.
The standard framing for this problem reaches for the word “bias”, as in negativity bias, optimism bias or confirmation bias, as though the brain were a generally functional instrument with a list of documented glitches. A bias implies deviation from correct function, whereas the evolutionary record points to correct functions operating in the wrong context. The name for this is evolutionary mismatch: the environment you live in was built without reference to the mind running in it.
The theoretical framework comes from John Tooby and Leda Cosmides: rather than a general-purpose reasoning machine or a blank slate written by culture, the brain is a collection of domain-specific psychological mechanisms, each with its own evolutionary history, each shaped by selection to solve specific problems in a specific environment. That environment has a name, the “environment of evolutionary adaptedness”, a term coined by the psychiatrist John Bowlby and developed by Tooby and Cosmides. For most of our mental machinery it was the Pleistocene, the stretch of time from roughly 2.5 million years ago to about 12,000 years ago. The twelve millennia since are a thin slice of that span.
If the mismatch is real and its costs are significant, it is fair to ask why 12,000 years of selection pressure in agricultural, then industrial, then digital environments have not produced a better-calibrated brain.
Part of the answer lies in what selection rewards. Selection operates on reproductive success in specific environments, not on wellbeing, rationality or suffering. A mechanism that made you anxious and distracted but kept your genes in the pool is, from selection’s perspective, a success. The larger part is speed: biological evolution cannot keep pace with cultural change. Cultural and technological change now operates on timescales of years to decades. Natural selection can tune simple traits quickly, and human adaptive evolution accelerated over the past 40,000 years, but even accelerated selection works generation by generation, and only under selection pressure that is strong, consistent, and differential, while culture now remakes the environment several times within a single lifetime. The agricultural revolution is roughly 500 generations ago, the industrial revolution ten, the internet one.
None of this is genetic determinism, and it does not foreclose change: environmental redesign is where this argument leads. Three mechanisms show the pattern most clearly, each an evolved solution to a real problem that now misfires.
The Negativity Bias and the Attention Economy
Cortisol is synthesized in the adrenal cortex and released in response to perceived threat. Working in tandem with adrenaline, which floods the body within seconds of a threat, it raises blood glucose, suppresses non-essential systems such as digestion, immune function and reproductive physiology, and keeps the body on alert for as long as the brain perceives danger. The problem Robert Sapolsky has spent a career studying is what happens when the system that releases it cannot distinguish between threats that are acute and physical and threats that are chronic and symbolic, arriving through a screen.
Sapolsky’s central observation sets humans apart from most other animals in a way that works to our disadvantage: we can activate the stress response with thought. A zebra’s stress response surges when a lion appears and subsides once the lion leaves or the zebra outruns it; the threat is real, and it passes. A human can activate the same physiological cascade by reading about a lion attack that happened to someone else in another country, or by anticipating a difficult conversation that hasn’t happened yet, or by dwelling on a threat that is years away or purely hypothetical. The response fires either way.
This is the cost of a brain capable of symbolic reasoning and future-modeling. The same cognitive capacity that lets humans plan and cooperate is the capacity that makes them uniquely susceptible to chronic stress activation. The downstream effects are physical. Chronically elevated cortisol suppresses immune function, accelerates cardiovascular damage, degrades memory consolidation, disrupts sleep architecture and, in conditions of prolonged glucocorticoid excess such as Cushing’s syndrome and severe recurrent depression, is linked to a measurably smaller hippocampus.
In an environment where threats such as predators, rival groups, contaminated food and injury were physical, local and immediate, asymmetric attention to negative stimuli was strongly adaptive. Missing a threat could cost a life; missing an opportunity usually cost a meal. The brain evolved accordingly: negative stimuli register faster than positive stimuli, recruit more neural processing resources, and are encoded in memory more readily, even though, over the years, the emotional sting of bad memories tends to fade faster than the glow of good ones. Bad is stronger than good in nearly every domain of human psychology that has been examined, from attention, memory and learning to social evaluation and emotional experience. The negativity bias is the human baseline, present in the calm as well as in the anxious or the depressed.
The attention economy is built on top of this baseline. Engagement-optimized platforms did not invent the negativity bias, but their ranking systems learned to exploit it. In thousands of Upworthy headline experiments, each additional negative word raised a typical headline’s click-through rate by about 2.3 percent. Posts about a political out-group are shared far more readily than other posts. From 2017, Facebook’s ranking formula counted emoji reactions, “angry” included, as worth five times a like. In early 2023 Twitter’s engagement-based ranking, set against a plain chronological feed, amplified emotionally charged, out-group-hostile content. None of this required anyone to set out to spread outrage; optimizing for engagement against real human behavior was enough, at a scale and with a precision that no previous information environment came close to.
The ethologist Nikolaas Tinbergen described what happens when an animal’s evolved response is triggered by a stimulus that exaggerates the features the response was calibrated for. Oystercatchers offered a choice between their own egg and a giant artificial one will clamber onto the giant, trying to brood an egg they can barely sit on; the preference for larger has been exploited past the point where it tracks anything real. In a separate line of experiments, herring gull chicks, which peck at the red spot on a parent’s bill to solicit food, pecked more at a thin red rod banded with white than at a realistic model of a gull’s head, the innate releasing mechanism responding more strongly to an exaggerated, high-contrast signal than to the stimulus it evolved to detect. Tinbergen called these supernormal stimuli. The information environment produced by engagement-maximization is a supernormal stimulus for the human threat-detection system: it delivers a continuous stream of worst-case events, curated from a global population of eight billion people, selected for engagement, which in practice rewards threat and outrage, and arriving far more often and with more force than news of danger ever did in an ancestral setting. The underlying system has little defense against it because that system was calibrated in a world where news of a threat came from nearby, now and then, and through people one knew.
A person living in one of the safest societies in recorded human history, with lower rates of violent death, longer life expectancy and more reliable food access than at any previous point in the species’ existence, who consumes a standard modern diet of news and social media can end up running a threat-response system designed for real emergencies as a background condition. Their nervous system, responding correctly to an engineered stimulus environment, tells them that the world is dangerous and deteriorating.
Status Circuitry and the Infinite Audience
In the early 1990s, the anthropologist and evolutionary psychologist Robin Dunbar noticed something in the data on primate neocortex size. Across the primate species he examined, the ratio of neocortex volume to the rest of the brain tracked not environmental complexity, tool use, or foraging strategy but social group size. This is the “social brain hypothesis”, still influential, though later and larger datasets suggest diet matters at least as much. The bigger the neocortex, the larger the stable social group the species maintains. Dunbar ran the human neocortex ratio against the primate regression line and arrived at a number: approximately 150. That, he argued, is the number of individuals a human being can maintain meaningful, trust-based, socially embedded relationships with simultaneously, the number at which you know who everyone is, what their relationships to each other are, and where you stand relative to them.
Dunbar then went looking for the number in real human groups and found it recurring at one particular level of organization: the clan or village community of hunter-gatherers and traditional farmers, which averaged about 150 in his sample and sits above the 30–50-person bands that camp together; Neolithic farming villages; the Roman maniple and the modern army company; and the size at which, by management folklore, companies start needing formal hierarchy. Dunbar reads this as a cognitive load limit, the point at which the primate social-tracking machinery runs out of bandwidth. The exact figure is contested. A later re-analysis of the primate data produced confidence intervals so wide, the widest running from 4 to 520, that its authors called any single figure futile, and Dunbar’s own confidence interval ran from roughly 100 to 230. But the core insight survives — the brain evolved to track a community of acquaintances numbered in the dozens or low hundreds, not millions.
In the ancestral environment, status was a running measurement of one’s position within a known group, among people whose opinions of you were directly tied to your access to food, mates, coalition support, and protection. The brain’s status-monitoring system, which links perceived social rank to the body’s stress and mood chemistry, was tuned to that environment. High perceived rank produced neurochemical conditions conducive to confidence, risk-taking, and social engagement. Low perceived rank, or rank instability, produced the opposite. The system ran continuously because rank in the ancestral environment changed continuously and the consequences of missing a shift were real.
Sapolsky’s long-term research on baboon troops in Kenya is among the most detailed physiological data we have on what social rank does to a mammalian body. In stable hierarchies, low-ranking baboons show chronically elevated cortisol, a sluggish recovery after stress, higher blood pressure, and suppressed immune function; when the hierarchy is in flux, the cortisol burden shifts to the high-ranking males whose position is being challenged from below. Sapolsky himself draws the parallel to our own species: the health costs of low rank in primates echo the socioeconomic gradient in human health, where those at the bottom carry a heavier burden of cardiovascular and other disease. Which animals pay depends on how the hierarchy is organized, but in every version chronic social stress takes a physical toll.
Today those circuits run on very different input. The status-monitoring system built for a community of known faces is receiving signals from an audience that is, in principle, infinite. Follower counts, like counts, comment engagement and repost velocity all work as public status signals. The brain can tell a band member, whose opinion bears directly on your survival and coalition access, from a stranger in another country, and it values a friend’s response more. But a notification from someone who has never met you, and whose opinion of you has no bearing on anything, can still register as a social signal.
The scale mismatch alone would be disorienting. An organism built to track its position in a group of a hundred or so cannot process meaningful rank information from an audience of millions: the signal becomes noise, the comparison cohort has no stable ceiling, and the question of where you stand relative to “everyone” has no answer. On top of that, the delivery mechanism has been engineered to maximize the frequency and emotional intensity of status-relevant events. Variable-ratio reward schedules, in which rewards arrive after an unpredictable number of tries, produce the highest, steadiest response rates and the most stubborn resistance to extinction of the classic reinforcement schedules. They are why slot machines are more compulsive than vending machines, which pay out every time. Follower counts that move up and down, engagement that arrives in unpredictable bursts, the possibility at any moment of either viral visibility or public humiliation — this is a variable reward schedule running against status circuitry that evolved to care intensely about every signal it received from the people it knew personally.
Sean Parker, Facebook’s founding president, said publicly in 2017 that this was by design: the goal was to consume as much of users’ time and attention as possible, likes and comments delivered “a little dopamine hit”, and the whole system was “a social-validation feedback loop” that exploited “a vulnerability in human psychology” — knowingly. The business model requires sustained attention. A reliable route to sustained attention runs through the brain’s status-monitoring system, delivering signals at a frequency and unpredictability that keeps the system perpetually activated. Whether this produces in users anything like the physiological toll Sapolsky documented in stressed baboons has not been established, and across large samples the average link between screen use and wellbeing is small. What Parker described, though, is a design aimed deliberately at that status system.
Hyperbolic Discounting and the Civilizational Horizon
In a standard temporal discounting experiment, participants are asked to choose between a smaller reward now and a larger reward later. Richard Thaler’s classic thought experiment runs like this. Offered one apple in a year or two apples in a year and a day, hardly anyone would take the single apple. Offered one apple today or two apples tomorrow, plenty of people would. The same one-day delay, the same doubling of the reward — yet the choice flips. The inconsistency is not ignorance of arithmetic, and it has been found again and again in laboratory and field studies, most strongly for immediate experiences such as food, effort and pleasure, more modestly for money.
In many choices the discount rate is not consistent across time but roughly hyperbolic: disproportionately steep for near-term delays, flattening out for delays further in the future. That is not how the textbook rational agent discounts; one applying a consistent rate to future rewards would have the same preference structure regardless of when the comparison window starts. People often treat the present as categorically different from the future, which produces systematic preference reversals, and the closer the immediate option, the more extreme the effect. Offer people a reward now or a larger one next week and many take the reward now; move both options forward in time, to one month versus five weeks, and many of them switch. The payoffs are the same; only their distance from the present has changed.
The usual evolutionary explanation runs through risk. In an environment of real scarcity and uncertain survival, steeply discounting the future was a sound bet. The probability of being alive in a year, in an environment with high rates of predation, disease, violence, and famine, was substantially lower than it is today. Preferring a certain caloric reward now over a larger but uncertain future one was sensible expected-value reasoning, and when the risk itself is uncertain, that reasoning yields the same steep-then-flat curve. On this account, a brain that consistently sacrificed present resources for future gains it might not live to collect would have left fewer descendants.
For most people reading this, that environment is gone, and the mismatch shows up at two scales. At the individual scale, virtually every significant domain of modern life in which long-term outcomes diverge substantially from short-term preferences, from retirement savings and diet to physical fitness, preventive medicine and education, runs directly against present bias. The 25-year-old who cannot sustain motivation to contribute to a retirement account is not making a reasoning error about compound interest. The mechanism that makes future rewards feel abstractly small relative to present costs is working as built, in a world that has stopped rewarding it. Richard Thaler and Shlomo Benartzi’s Save More Tomorrow program showed what happens when you design around this bias rather than try to overcome it with information: employees committed in advance to send part of each future pay raise into their retirement plan, and participants’ saving rates climbed from 3.5 percent to 13.6 percent in 40 months. Automatic enrollment works the same way; at one large US firm it lifted participation among new hires by some 48 percentage points. Financial-education programs improve behavior too, but by smaller margins. Making the beneficial action the default removes the effort of taking it from the present. Telling people about compound interest, which assumes they can reason their way past present bias, helps, but less reliably.
At the civilizational scale, democratic electorates look as if they discount future harms and benefits much as individuals do. Ignorance, corrupt politicians and misinformation play their part, but the electorate is also composed of humans with human time preferences. Infrastructure, climate mitigation, pandemic preparedness, antibiotic resistance and pension solvency all require accepting real present costs for benefits that materialize over decades, exactly the trade present bias handles worst. The present cost is vivid and immediate and registers as a loss; the future benefit is abstract and distant and registers as small.
Calling this a failure of political will, in the sense of weak character, misses the mechanism, and better leadership or clearer communication cannot solve it on their own. An ancestral mechanism is being applied to problems it never evolved to handle. On this reading, collective political decisions discount the future far more steeply than governing across decades requires. Public information campaigns alone are unlikely to close that gap, because they try to change a preference structure using information, and the preference structure responds to information far less than the campaigns assume. The cognitive architecture of the species is not well suited to managing multi-decadal threats, and institutions meant to manage them have to be designed around that limit instead of pretending it isn’t there. The usual next question, how individuals can fix themselves, has filled shelves of self-help books. The more useful one is what to change around them.
Redesigning the Environment
If the mismatches described above are structural, evolved mechanisms operating in environments they were never designed for, then individual behavioral change is the wrong primary lever. It still has its uses, but it is hard to scale, and it asks people to override evolved mechanisms by willpower alone — an unreliable tool against an environment engineered to defeat it. Change across a population comes from redesigning the environment.
The idea already has a proof of concept. Thaler and Sunstein’s nudge program showed that changing the default environment can shift behavior in ways that education and exhortation struggle to match, and while large-scale government trials find most nudges smaller than early academic studies promised, defaults remain among the most reliable levers there are. Auto-enrollment in pension plans does not teach people to want retirement security more than they already did; it redesigns the environment so that the path of least resistance leads toward the better outcome. Organ donation opt-out systems leave people’s values alone and change only the default. Because most people never get around to ticking a box either way, effective consent, the share of people counted as willing donors, runs at very high levels in opt-out countries and stays stubbornly low in opt-in ones, whatever people say they want. How much that raises actual donations depends on hospitals, families and infrastructure, but the default does the heavy lifting on consent.
What does not yet exist is policy ambition that matches the scale of the problem. The attention economy’s business model exploits the negativity bias, the status-monitoring system and the variable reward response at industrial scale. Regulating the engagement-maximizing design patterns that put this exploitation into practice, such as infinite scroll, variable-reward notification timing, algorithmic amplification of threat and outrage content and the architecture of social comparison metrics, would regulate design rather than speech, in the same category of intervention as banning leaded gasoline: removing from the environment a substance that damages a biological system that has no evolved defense against it. Lead was banned on three grounds: the harm was structural, the exposure was involuntary and the mechanism was understood. Nobody argued that individuals should simply have avoided it. The attention economy meets the same tests only in part: the engagement mechanism is understood, though how much harm it does varies widely from user to user, and for a significant portion of the global population the exposure is effectively involuntary, given the social and professional infrastructure now built on these platforms, while the industry goes on deploying the mechanism.
The tobacco industry knew nicotine was addictive long before the regulatory framework caught up: in 1963 Brown & Williamson’s general counsel wrote that the company was “in the business of selling nicotine, an addictive drug”, and the US Surgeon General formally declared nicotine addictive in 1988. The gap between scientific consensus on harm and effective regulatory response was measured in decades, during which the industry continued to optimize its product for addictive effect while funding research designed to muddy the scientific record. The attention economy is roughly where tobacco was by the mid-1960s: the engagement mechanism is documented in the research literature and understood internally by the companies deploying it, the size of the harm is still argued over, and regulators are only now beginning to act. The EU’s Digital Services Act has required the largest platforms to assess and mitigate risks to users’ mental wellbeing since 2023. In July 2025 the European Commission recommended that platforms switch off streaks, autoplay and push notifications by default for minors, and in February 2026 it preliminarily found TikTok’s infinite scroll, autoplay, push notifications and personalized recommender system in breach of the law. Australia barred under-16s from holding accounts on major social media platforms in December 2025, New York’s limits on algorithmic feeds for minors take effect in January 2027, and in March 2026 a Los Angeles jury found Meta and YouTube liable for designing features that hooked a young user, a first test case that both companies dispute. None of this yet amounts to a leaded-gasoline moment: the engagement-maximizing business model itself remains legal almost everywhere. How long the lag will be this time is an open question; the last one ran for decades.
The companies have not been waiting for the science. Meta’s own leaked research shows it measuring the trade-offs and shipping the designs anyway; about one in three teenage girls with body-image problems told Instagram’s researchers that the app made them feel worse.
The harder problem is that the institutions that would need to act on what is known are staffed by humans running the same cognitive architecture. They face the same short-horizon electoral and financial incentives, work inside the same information environments, and are as prone to discounting the future, worrying about rank and reacting to threat as the populations they would need to protect. Policymakers do not stand above the mismatch with clear vision while it plays out in the electorate or the market.
Gen AI Disclaimer
Some contents of this page were generated and/or edited with the help of a Generative AI.
Media
Key Sources and References
Evolution and mismatch
Tooby, J., & Cosmides, L. (1992). The psychological foundations of culture. In J. H. Barkow, L. Cosmides, & J. Tooby (Eds.), The Adapted Mind: Evolutionary Psychology and the Generation of Culture (pp. 19–136). Oxford University Press. https://archive.org/details/adaptedmindevolu0000unse_p8s2
Barkow, J. H., Cosmides, L., & Tooby, J. (Eds.). (1992). The Adapted Mind: Evolutionary Psychology and the Generation of Culture. Oxford University Press. https://archive.org/details/adaptedmindevolu0000unse_p8s2
Center for Evolutionary Psychology, University of California, Santa Barbara. Environments of evolutionary adaptedness (EEA). https://www.cep.ucsb.edu/environments-of-evolutionary-adaptedness-eea/
Lancaster University Psychology Glossary. Environment of evolutionary adaptedness (term introduced by John Bowlby, 1969). https://www.lancaster.ac.uk/fas/psych/glossary/environment_of_evolutionary_adaptedness/
Hawks, J., Wang, E. T., Cochran, G. M., Harpending, H. C., & Moyzis, R. K. (2007). Recent acceleration of human adaptive evolution. Proceedings of the National Academy of Sciences, 104(52), 20753–20758. https://pubmed.ncbi.nlm.nih.gov/18087044/
Stress, threat and the negativity bias
Sapolsky, R. M. (2004). Why Zebras Don’t Get Ulcers (3rd ed.). Owl Books/Henry Holt. (Original work published 1994.) https://openlibrary.org/isbn/9780805073690
Sapolsky, R. M. (2017). Behave: The Biology of Humans at Our Best and Worst. Penguin Press. https://www.penguinrandomhouse.com/books/311787/behave-by-robert-m-sapolsky/9781594205071/
Harvard Health Publishing. Understanding the stress response. https://www.health.harvard.edu/healthy-aging-and-longevity/understanding-the-stress-response
Sapolsky, R. M., Romero, L. M., & Munck, A. U. (2000). How do glucocorticoids influence stress responses? Integrating permissive, suppressive, stimulatory, and preparative actions. Endocrine Reviews, 21(1), 55–89. https://pubmed.ncbi.nlm.nih.gov/10696570/
Sapolsky, R. M. (2000). Glucocorticoids and hippocampal atrophy in neuropsychiatric disorders. Archives of General Psychiatry, 57(10), 925–935. https://doi.org/10.1001/archpsyc.57.10.925
Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001). Bad is stronger than good. Review of General Psychology, 5(4), 323–370. https://doi.org/10.1037/1089-2680.5.4.323
Walker, W. R., & Skowronski, J. J. (2009). The fading affect bias: But what the hell is it for? Applied Cognitive Psychology, 23(8), 1122–1136. https://doi.org/10.1002/acp.1614
Walker, W. R., Yancu, C. N., & Skowronski, J. J. (2014). Trait anxiety reduces affective fading for both positive and negative autobiographical memories. Advances in Cognitive Psychology, 10(3), 81–89. https://pmc.ncbi.nlm.nih.gov/articles/PMC4197641/
Engagement, negativity and platform design
Robertson, C. E., Pröllochs, N., Schwarzenegger, K., Pärnamets, P., Van Bavel, J. J., & Feuerriegel, S. (2023). Negativity drives online news consumption. Nature Human Behaviour, 7(5), 812–822. https://doi.org/10.1038/s41562-023-01538-4
Rathje, S., Van Bavel, J. J., & van der Linden, S. (2021). Out-group animosity drives engagement on social media. Proceedings of the National Academy of Sciences, 118(26), e2024292118. https://doi.org/10.1073/pnas.2024292118 · https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8256037/
Milli, S., Carroll, M., Wang, Y., Pandey, S., Zhao, S., & Dragan, A. D. (2025). Engagement, user satisfaction, and the amplification of divisive content on social media. PNAS Nexus, 4(3), pgaf062. https://pmc.ncbi.nlm.nih.gov/articles/PMC11894805/
Merrill, J. B., & Oremus, W. (2021, October 26). Five points for anger, one for a ‘like’: How Facebook’s formula fostered rage and misinformation. The Washington Post (via Anchorage Daily News). https://www.adn.com/nation-world/2021/10/26/five-points-for-anger-one-for-a-like-how-facebooks-formula-fostered-rage-and-misinformation/
Feiner, L. (2021, September 14). Facebook documents show how toxic Instagram is for teens, Wall Street Journal reports. CNBC (via NBC Bay Area). https://www.nbcbayarea.com/news/business/money-report/facebook-documents-show-how-toxic-instagram-is-for-teens-wall-street-journal-reports/2655293/
Salinas, S. (2017, November 9). Facebook co-founder Sean Parker bashes company, saying it was built to exploit human vulnerability. CNBC. Also: The Register (2017, November 10), Sean Parker: I helped destroy humanity with Facebook. https://www.cnbc.com/2017/11/09/facebooks-sean-parker-on-social-media.html · https://www.theregister.com/offbeat/2017/11/10/sean-parker-i-helped-destroy-humanity-with-facebook/1340447
Orben, A., & Przybylski, A. K. (2019). The association between adolescent well-being and digital technology use. Nature Human Behaviour, 3(2), 173–182. https://doi.org/10.1038/s41562-018-0506-1
Supernormal stimuli
Tinbergen, N. (1951). The Study of Instinct. Clarendon Press. https://archive.org/details/studyofinstinct0000ntin_a1t3
Tinbergen, N., & Perdeck, A. C. (1950). On the stimulus situation releasing the begging response in the newly hatched herring gull chick (Larus argentatus argentatus Pont.). Behaviour, 3(1), 1–39. https://doi.org/10.1163/156853951X00197
The social brain, status and reward
Dunbar, R. I. M. (1992). Neocortex size as a constraint on group size in primates. Journal of Human Evolution, 22(6), 469–493. https://doi.org/10.1016/0047-2484(92)90081-J
Dunbar, R. I. M. (1993). Co-evolution of neocortex size, group size and language in humans. Behavioral and Brain Sciences, 16(4), 681–694. https://doi.org/10.1017/S0140525X00032325 · https://mnytud.arts.unideb.hu/tananyag/dunbar.htm
Lindenfors, P., Wartel, A., & Lind, J. (2021). ‘Dunbar’s number’ deconstructed. Biology Letters, 17(5), 20210158. https://doi.org/10.1098/rsbl.2021.0158
DeCasien, A. R., Williams, S. A., & Higham, J. P. (2017). Primate brain size is predicted by diet but not sociality. Nature Ecology & Evolution, 1, 0112. https://doi.org/10.1038/s41559-017-0112
Sapolsky, R. M. (1992). Cortisol concentrations and the social significance of rank instability among wild baboons. Psychoneuroendocrinology, 17(6), 701–709. https://doi.org/10.1016/0306-4530(92)90029-7
Sapolsky, R. M. (2005). The influence of social hierarchy on primate health. Science, 308(5722), 648–652. https://doi.org/10.1126/science.1106477
Fareri, D. S., Niznikiewicz, M. A., Lee, V. K., & Delgado, M. R. (2012). Social network modulation of reward-related signals. Journal of Neuroscience, 32(26), 9045–9052. https://doi.org/10.1523/JNEUROSCI.0610-12.2012
Ferster, C. B., & Skinner, B. F. (1957). Schedules of Reinforcement. Appleton-Century-Crofts. Summarized in: Schedules of reinforcement, Psychology of Learning, Texas Wesleyan University. https://txwes.pressbooks.pub/psychologyoflearningtxwes/chapter/06-3-reinforcement-schedules/
Discounting, defaults and nudges
Thaler, R. H. (1981). Some empirical evidence on dynamic inconsistency. Economics Letters, 8(3), 201–207. https://doi.org/10.1016/0165-1765(81)90067-7
Ainslie, G. (1975). Specious reward: A behavioral theory of impulsiveness and impulse control. Psychological Bulletin, 82(4), 463–496. https://doi.org/10.1037/h0076860
Laibson, D. (1997). Golden eggs and hyperbolic discounting. Quarterly Journal of Economics, 112(2), 443–478. https://doi.org/10.1162/003355397555253
Frederick, S., Loewenstein, G., & O’Donoghue, T. (2002). Time discounting and time preference: A critical review. Journal of Economic Literature, 40(2), 351–401. https://www.aeaweb.org/articles?id=10.1257/002205102320161311
Sozou, P. D. (1998). On hyperbolic discounting and uncertain hazard rates. Proceedings of the Royal Society B: Biological Sciences, 265(1409), 2015–2020. https://doi.org/10.1098/rspb.1998.0534
Imai, T., Rutter, T. A., & Camerer, C. F. (2021). Meta-analysis of present-bias estimation using convex time budgets. The Economic Journal, 131(636), 1788–1814. https://resolver.caltech.edu/CaltechAUTHORS:20200117-112420357
Thaler, R. H., & Benartzi, S. (2004). Save More Tomorrow™: Using behavioral economics to increase employee saving. Journal of Political Economy, 112(S1), S164–S187. https://doi.org/10.1086/380085
Benartzi, S., & Thaler, R. H. (2013). Behavioral economics and the retirement savings crisis. Science, 339(6124), 1152–1153. https://harrt.ucla.edu/wp-content/uploads/2015/12/Science-2013-Benartzi-1152-3.pdf
Madrian, B. C., & Shea, D. F. (2001). The power of suggestion: Inertia in 401(k) participation and savings behavior. Quarterly Journal of Economics, 116(4), 1149–1187. https://www.nber.org/papers/w7682
Fernandes, D., Lynch, J. G., Jr., & Netemeyer, R. G. (2014). Financial literacy, financial education, and downstream financial behaviors. Management Science, 60(8), 1861–1883. https://doi.org/10.1287/mnsc.2013.1849 · https://www.nefe.org/news/nefe-digest/2014/time-for-financial-education-to-evolve.aspx
Kaiser, T., Lusardi, A., Menkhoff, L., & Urban, C. (2022). Financial education affects financial knowledge and downstream behaviors. Journal of Financial Economics, 145(2), 255–272. https://doi.org/10.1016/j.jfineco.2021.09.022
Thaler, R. H., & Sunstein, C. R. (2008). Nudge: Improving Decisions About Health, Wealth, and Happiness. Yale University Press. (Updated Final Edition: Penguin Books, 2021.) https://yalebooks.yale.edu/book/9780300122237/nudge/
DellaVigna, S., & Linos, E. (2022). RCTs to scale: Comprehensive evidence from two nudge units. Econometrica, 90(1), 81–116. https://www.econometricsociety.org/publications/econometrica/2022/01/01/rcts-scale-comprehensive-evidence-two-nudge-units
Johnson, E. J., & Goldstein, D. (2003). Do defaults save lives? Science, 302(5649), 1338–1339. https://doi.org/10.1126/science.1091721
Shepherd, L., O’Carroll, R. E., & Ferguson, E. (2014). An international comparison of deceased and living organ donation/transplant rates in opt-in and opt-out systems: A panel study. BMC Medicine, 12, 131. https://pmc.ncbi.nlm.nih.gov/articles/PMC4175622/
Willpower
Hagger, M. S., Chatzisarantis, N. L. D., et al. (2016). A multilab preregistered replication of the ego-depletion effect. Perspectives on Psychological Science, 11(4), 546–573. https://doi.org/10.1177/1745691616652873
Vohs, K. D., Schmeichel, B. J., et al. (2021). A multisite preregistered paradigmatic test of the ego-depletion effect. Psychological Science, 32(10), 1566–1581. https://doi.org/10.1177/0956797621989733
Tobacco, lead and regulation
Yeaman, A. (1963, July 17). Brown & Williamson internal memorandum by the company’s general counsel. Reproduced in “Smoking Howitzers”, Inside the Tobacco Deal, PBS Frontline. https://www.pbs.org/wgbh/pages/frontline/shows/settlement/case/yeaman.html
U.S. Department of Health and Human Services (1988). The Health Consequences of Smoking: Nicotine Addiction. A Report of the Surgeon General. Office on Smoking and Health. https://profiles.nlm.nih.gov/101584932X423
U.S. Environmental Protection Agency (1996, January 29). EPA takes final step in phaseout of leaded gasoline. https://www.epa.gov/archive/epa/aboutepa/epa-takes-final-step-phaseout-leaded-gasoline.html
Regulation (EU) 2022/2065 of the European Parliament and of the Council on a Single Market for Digital Services (Digital Services Act). https://eur-lex.europa.eu/eli/reg/2022/2065/oj/eng · https://digital-strategy.ec.europa.eu/en/policies/digital-services-act
Lewis Silkin (2024, February 17). DSA applies to all online platforms in the EU. https://www.lewissilkin.com/insights/2024/02/17/dsa-applies-to-all-online-platforms-in-the-eu-102j0fj
European Commission (2025, July 14). Commission publishes guidelines on the protection of minors. https://digital-strategy.ec.europa.eu/en/library/commission-publishes-guidelines-protection-minors
European Commission (2026, February 6). Preliminary findings that TikTok’s addictive design breaches the Digital Services Act (press release IP/26/312); summarized by Slaughter and May. https://ec.europa.eu/commission/presscorner/detail/en/ip_26_312 · https://thelens.slaughterandmay.com/post/102miei/commission-makes-preliminary-finding-that-tiktoks-addictive-design-breaches-th
McCullough Robertson (2025, December 10). World-first: Australia’s social media ban for children starts today. https://mccullough.com.au/2025/12/10/australias-social-media-ban-under-16/
New York State Attorney General (2026, July 28). Attorney General James and Governor Hochul release final SAFE for Kids Act rules. https://ag.ny.gov/press-release/2026/attorney-general-james-and-governor-hochul-release-final-safe-kids-act-rules
6abc / ABC News (2026, March 26). Jury finds Instagram and YouTube liable in landmark social media addiction trial in Los Angeles. https://6abc.com/18771272/
