Visit your dentist twice a year. Sleep eight hours in an unbroken block. Floss every night. These instructions arrive wrapped in the authority of science, the reassurance of consensus, the patina of timelessness. They feel like facts discovered rather than decisions made – as though someone in a laboratory, bent over instruments, unearthed them from nature’s substrate. What if the opposite were true? What if these “best practices” emerged not from laboratories but from advertising agencies, not from controlled trials but from the practical needs of insurance actuaries and factory owners? This question matters beyond intellectual curiosity. Millions of people structure their lives around recommendations whose origins they’ve never examined. Night-shift workers absorb warnings that their schedules are slowly killing them, warnings they cannot act upon without abandoning the jobs that keep them fed. Parents enforce bedtimes based on sleep hygiene principles manufactured for industrial convenience. Patients submit to procedures whose frequency was determined by a toothpaste jingle. Understanding how knowledge gets constructed doesn’t require abandoning expertise or embracing conspiracy. It requires something harder: holding findings provisionally while remaining curious about the conditions of their production. The distinction between skepticism and cynicism matters here. Cynicism dismisses everything as corrupt, which takes no intellectual effort and produces no useful insight. Skepticism asks questions: Who funded this research? What alternative explanations weren’t explored? Whose interests does this recommendation serve? Cynicism is a posture; skepticism is a practice. One closes inquiry, the other opens it.

A toothpaste company in the 1940s needed to sell more product. Their advertising executive – a man renowned for techniques that appeared scientific but weren’t – crafted a slogan: “Use Pepsodent every day – see your dentist twice a year.” The frequency wasn’t derived from any study. It was invented, whole cloth, because it sounded authoritative and created a behavioral rhythm that kept the product top-of-mind. Repetition made it feel inevitable. Three decades later, when dental insurance emerged as an employee benefit, companies faced a practical problem: how many cleanings should they cover? The answer, conveniently, already existed in public consciousness. Twice yearly. The interval fit nicely into benefits administration. It controlled costs while appearing generous. What began as advertising copy became insurance policy, and insurance policy became settled medical wisdom. Today, dentists recommend biannual visits with the solemnity of people announcing natural law. What does the evidence actually show? When researchers conduct rigorous reviews of the literature – the kind that pool multiple studies and correct for bias – they find insufficient evidence to conclude anything definitive about optimal recall intervals. Computer modeling suggests ideal frequencies might range from just over a year to a decade, depending on individual risk factors. Yet the six-month interval persists, zombie-like, because it serves everyone’s convenience: insurers who want predictable costs, dentists who want reliable revenue, patients who want simple rules.

The flossing story goes further. When journalists filed requests asking federal health agencies to produce the scientific basis for flossing recommendations, the government quietly acknowledged that the practice had never been adequately researched. The evidence base consisted largely of short-term studies, often funded by companies that manufacture floss, with sample sizes too small and durations too brief to support the confident claims printed on packaging. The recommendation was removed from dietary guidelines without announcement – a silent admission that decades of advice had rested on surprisingly weak foundations. “So flossing is useless?” asks someone hearing this for the first time. Not necessarily. The absence of strong evidence for a practice differs from evidence that the practice doesn’t work. Maybe flossing helps; maybe it doesn’t. The point isn’t that dental professionals lie or that oral hygiene is a scam. The point is that the confidence with which recommendations are delivered often exceeds the confidence the evidence warrants. The authoritative tone obscures the provisionality of the underlying knowledge.

Consider how we sleep – or rather, how we’re told to sleep. Eight hours, consolidated, at night. This pattern feels so natural that imagining alternatives requires effort. Surely humans have always slept this way, bodies attuned to circadian rhythms that demand unbroken darkness? Historical research tells a different story. Before industrialization, across cultures and centuries, people commonly slept in two distinct phases separated by an hour or more of quiet wakefulness. References to “first sleep” and “second sleep” appear throughout the documentary record, from ancient texts to medieval manuscripts to early modern diaries. This wasn’t a disorder to be corrected; it was the pattern. People used the interval between sleeps for prayer, conversation, intimacy, or simply lying in peaceful awareness. When researchers placed subjects in extended darkness – fourteen hours without artificial light – they spontaneously reverted to this biphasic pattern within weeks. Their bodies, freed from electric illumination and alarm clocks, remembered something their minds had forgotten.

What changed? The consolidated eight-hour block wasn’t discovered; it was manufactured. A labor reformer in the early nineteenth century coined a slogan dividing the day into thirds: eight hours for work, eight hours for leisure, eight hours for rest. This was a political response to factory conditions that demanded sixteen-hour shifts, not a biological insight. The slogan became policy, then assumption, then “natural.” Artificial lighting played its part. When cities first illuminated their streets – gas lamps flickering where darkness once reigned – the night became available for commerce and entertainment. The quiet hours between sleeps, once protected by impenetrable dark, now competed with coffeehouses and theaters. Mass production of alarm clocks created millions of devices whose sole purpose was wrenching people from sleep at predetermined times. Compulsory schooling required children to appear at specific hours, training bodies from childhood to conform to industrial rhythms. A fascination with productivity made nighttime wakefulness seem lazy rather than natural. None of this means the eight-hour block is bad or that everyone should adopt polyphasic sleep. It means that what feels like biological imperative emerged from historical contingency. The body is more flexible than sleep hygiene recommendations suggest.

Roughly fifteen million people in a single country work overnight shifts. They staff hospitals and stock shelves, drive trucks through empty highways and watch security monitors in silent buildings. Society cannot function without them. The bread on your breakfast table, the nurse who took your 3am emergency call, the power that kept your refrigerator running while you slept – all depend on people whose schedules invert the recommended pattern. These workers are told constantly that their lives are dangerous. Night shifts probably cause cancer; an international health agency classifies them as “probably carcinogenic.” Elevated risks of heart disease, diabetes, depression, cognitive decline – the warnings accumulate into a drumbeat of doom. Yet, look at the people delivering these warnings. Researchers who sleep soundly through the night, in homes they can afford because they work prestigious daytime jobs. The gap between those who study night work and those who perform it reflects a broader pattern: recommendations designed for people with choices, delivered to people without them. A night-shift worker cannot simply decide to sleep at recommended times. The job requires otherwise, and the job pays the bills. Follow the health advice and you lose the income that makes healthcare accessible. Ignore the advice and you’re reminded constantly of the damage you’re doing. This catch-22 structures millions of lives: the work that funds survival is also, allegedly, the work that shortens it.

Zoom out and the paradox sharpens. Throughout history, many of the most productive and creative people maintained schedules that would horrify a sleep scientist. They worked through the night, slept at irregular hours, and lived long enough to change the world. The successful night owl is dismissed as survivorship bias – we see those who thrived while ignoring those who burned out. Fair enough. But survivorship bias cuts both directions. We also don’t see all the people who followed every health recommendation and still developed the diseases they were promised protection from. Burnout, heart attacks, depression – these afflict nine-to-fivers too. Controlling for socioeconomic factors is harder than it sounds. Night-shift workers are disproportionately poorer, less educated, more stressed, with worse healthcare access. When studies find that night shifts harm health, how much is the schedule and how much is the circumstances of people forced into such schedules? Long-term employment studies consistently find that social rank predicts health outcomes even controlling for lifestyle – that where you sit in a hierarchy affects your body through mechanisms we’re only beginning to understand. The honest answer is: we don’t fully know. The research is real, the risks may be real, but the certainty with which warnings are issued outpaces the confidence the evidence warrants. A little humility would serve everyone better than the current approach, which gives night-shift workers guilt without agency.

The broader problem is that evidence changes its mind more often than anyone likes to admit. In 2015, a collaborative effort attempted to replicate one hundred psychology studies published in top journals. These weren’t obscure findings but the kind that appear in textbooks and TED talks. Just over a third produced statistically significant results the second time around. Average effect sizes – the magnitude of the phenomena being measured – dropped by half. The findings that seemed solid melted into vapor upon closer inspection. The problem isn’t limited to psychology. When pharmaceutical companies tried to replicate landmark preclinical cancer studies – the kind that justify moving potential drugs into human trials – success rates hovered around ten to twenty percent. The foundation upon which billions of research dollars rest turns out to be remarkably unstable. Some collapses are dramatic. The idea that willpower functions like a muscle that fatigues with use – that resisting one temptation depletes your capacity to resist the next – seemed to explain so much about human behavior. It appeared in hundreds of papers, spawned practical applications, earned its originator professional acclaim. Then multiple laboratories, involving thousands of participants, tested it again. The effect vanished. It was essentially zero, indistinguishable from noise. A decade of research had elaborated an illusion. The phenomenon where adopting expansive physical postures supposedly increases testosterone and decreases stress hormones – making you feel more powerful through body positioning – captured public imagination in one of the most-viewed presentations of the video-lecture era. Millions of people practiced these poses before job interviews and dates. Then the key claims failed replication. One of the original researchers publicly stated she no longer believed the effects were real.

Why does this happen? The answer is structural rather than conspiratorial. The incentives governing research actively produce unreliable results. When you conduct a study, you analyze data. Multiple analytical approaches are usually possible – different ways of slicing the numbers, different covariates to include or exclude, different subgroups to examine. If the first analysis doesn’t yield a significant result, you try another. And another. Until something crosses the threshold of statistical significance and becomes publishable. This practice, sometimes called p-hacking, can inflate false positive rates dramatically. What looks like a 5% chance of random error becomes 60% or higher when researchers analyze data repeatedly until they find what they’re looking for. Journals want interesting results. “We tried this and nothing happened” doesn’t generate citations or attention. So negative findings languish in desk drawers while positive findings (even if inflated or false) reach publication. This skews the literature toward effects that may not exist. Small sample sizes compound the problem. With few participants, random variation masquerades as real effects. Early studies, conducted with insufficient power to detect true effects reliably, often produce exaggerated estimates that shrink as evidence accumulates. Researchers call this the “decline effect” – the phenomenon where early flashy findings fade toward nothing as more careful work is done. Surveys consistently find that majorities of researchers admit to practices that compromise reliability: selectively reporting only the results that “worked,” continuing to collect data until significance appeared, or not disclosing all experimental conditions. These aren’t aberrations but norms. The system incentivizes unreliable science, then expresses surprise when unreliable science is what it gets.

Major health recommendations have undergone reversals that would be comical if they hadn’t affected millions of lives. For decades, dietary fat was the enemy. Saturated fat clogged arteries and caused heart disease – everyone knew this. The recommendation to reduce fat intake shaped national guidelines, food industry reformulations, and personal guilt. Butter became poison; margarine became virtue. But the evidence accumulated in a different direction. More than twenty independent reviews have concluded that saturated fats show no significant effect on major cardiovascular outcomes. The foundations of the low-fat diet crumbled even as its recommendations persisted. Eggs, once limited to a few per week because of cholesterol concerns, were rehabilitated when guidelines finally acknowledged that dietary cholesterol has little relationship to blood cholesterol in most people. For decades, people avoided one of nature’s most complete foods based on a mechanistic assumption that bypassed the complexity of human metabolism.

Hormone replacement therapy for postmenopausal women went from routine prescription to dangerous treatment and back toward rehabilitation – a trajectory driven by studies whose design flaws became apparent only later. The research that condemned the treatment used older formulations in women a decade past menopause, distorting risks while obscuring benefits for the population that might actually benefit. The pattern repeats: confident recommendations based on the best available evidence, followed by reversal when better evidence arrives. This isn’t a failure of science – it’s science working as designed, correcting itself over time. But the self-correction happens on timescales of decades while recommendations affect lives immediately. People who followed the advice faithfully discover, years later, that the advice was wrong. “So we should ignore medical recommendations entirely?” No. But we should hold them with appropriate uncertainty. The gap between what research actually shows (provisional findings, effect sizes that may shrink, mechanisms incompletely understood) and how recommendations are communicated (confident directives delivered as settled fact) constitutes a kind of epistemic malpractice. Honest communication would acknowledge uncertainty. Instead, we get certainty that later evaporates, breeding precisely the cynicism that honest uncertainty could prevent.

Some distortions are less innocent than methodological artifact. In the mid-1960s, an industry trade group paid scientists to produce a literature review on heart disease. The group set the review’s objectives, contributed articles for consideration, and received drafts before publication. The resulting paper, published in a prestigious medical journal without disclosing the funding, downplayed sugar’s role while implicating fat – shaping dietary guidelines for generations. The scientists weren’t frauds; they may have genuinely believed their conclusions. But the money came with expectations, and the expectations shaped the work. This is the machinery of influence: not crude bribery but subtle shaping of what questions get asked, what comparisons get made, what endpoints get measured. A company funding research on its product has many tools short of falsification. Compare your product to placebo rather than to competitors. Measure outcomes at timepoints when effects peak rather than when they fade. Report the subgroups that showed benefits while burying those that didn’t. These are judgment calls, technically defensible, that systematically tilt conclusions in the funder’s favor. Consider pharmaceutical research. Studies sponsored by drug companies are over four times more likely to favor the sponsor’s product than independent studies. The drugs aren’t necessarily worse – but the research is designed to make them look better. When clinical trial results threaten profits, data has been known to disappear. During lawsuits over one major pain medication, documents revealed that evidence of heart attacks had been deleted from datasets before publication. Internal drafts of scientific papers listed company employees as lead authors; published versions credited outside academics recruited later, with the words “external author?” literally written in the manuscripts where their names would go.

The tobacco industry pioneered these techniques, refining them over decades into a playbook still deployed today. Internal documents, released through litigation, show explicit strategy: “Doubt is our product,” one memo declared, “since it is the best means of competing with the ‘body of fact’ that exists in the mind of the general public.” Fund research that produces uncertainty. Attack scientists who reach inconvenient conclusions. Manufacture controversy where consensus exists. The goal isn’t to win the argument but to muddy it – to make the public believe that experts disagree when they largely don’t. The playbook spread. Petrochemical companies knew about climate change in the late 1970s – their internal research accurately predicted warming decades before public awareness coalesced. Rather than act on this knowledge, they funded groups that misrepresented the science, sowing confusion that delayed action for a generation. Lead manufacturers, when confronted with evidence that their products poisoned children, attacked the researchers rather than reformulating the product. For over forty years, all research on leaded gasoline’s health effects was funded by the companies that made it – a fox-guarding-henhouse arrangement that slowed recognition of lead’s dangers by decades. None of this means all industry-funded research is fraudulent. Much of it is competent, useful, and honestly conducted. But the pattern is clear: when profits conflict with findings, the integrity of findings cannot be taken for granted. Transparency about funding sources matters. Independent replication matters. Attention to conflicts of interest matters. The question isn’t whether researchers are good people (usually they are) but whether incentive structures reliably produce reliable knowledge (sometimes they don’t).

Technology companies offer a contemporary illustration, one perhaps more familiar to anyone who’s tried to do something simple online and found themselves blocked by an artificial limitation. A cloud storage company that once offered unlimited device synchronization now limits free accounts to three devices. A video conferencing platform caps free meetings at forty minutes – not because technology demands it but because forty minutes falls just short of the average meeting length, creating maximum pressure to upgrade. A note-taking application that once stored essentially unlimited notes now restricts free users to fifty. These aren’t engineering constraints but business decisions, presented as though they were natural features of the landscape. The pattern has a name now – enshittification – a term that won Word of the Year recognition for capturing something everyone recognized but hadn’t articulated. First, platforms attract users by offering generous value: useful features, low friction, reasonable terms. Once users are locked in, the platform degrades their experience to extract more value, serving advertisers and business customers at users’ expense. Finally, the platform extracts maximum value from everyone while maintaining just enough functionality to prevent exodus.

The parallels to research institutions illuminate something about how knowledge gets produced under commercial pressure. Both platforms and research operate within incentive structures that can diverge from user or public interest. Both present outputs – features or findings – as neutral while obscuring the decisions shaping them. The cloud storage company doesn’t advertise that its limitation exists to coerce upgrades; it presents the limit as a fact of the product. The funded study doesn’t announce that its design was optimized to favor the sponsor; it presents conclusions as objective findings. Interface designers have developed a taxonomy for the darker techniques: patterns that make unsubscribing difficult, that pre-check boxes users would reject, that hide costs until the final purchase step, that guilt users into choices they’d otherwise refuse. Studies find these patterns on the vast majority of major websites. They work not because users are foolish but because designers are sophisticated, exploiting the gap between how people think they behave and how they actually do.

How, then, should anyone navigate a landscape where recommendations may be influenced by interests that supersede truth? The first step is abandoning the fantasy that expertise is either wholly reliable or wholly corrupt. Neither extreme fits reality. Most researchers are honest people doing their best within flawed systems. Most findings contain useful signal even when wrapped in noise. The task isn’t to accept or reject wholesale but to calibrate confidence appropriately – to hold findings provisionally, update when new evidence arrives, and attend to the conditions under which knowledge was produced. Ask who funded the research. Not because funding automatically invalidates findings, but because it demands greater scrutiny. Ask whether the study has been replicated. Not because single studies are worthless, but because replicated findings are more trustworthy. Ask whether effect sizes are large enough to matter. Not because small effects are meaningless, but because statistical significance and practical significance are different things. Ask what alternative explanations weren’t tested. Not because researchers are hiding something, but because any study illuminates some possibilities while leaving others in shadow. Spin multiple hypotheses before settling on one. Seek independent confirmation rather than relying on single sources. Remember that arguments from authority carry limited weight – credentials don’t guarantee correctness. Quantify claims where possible; vague assertions resist evaluation. Recognize that every link in an argument must hold; one weak connection undermines the chain. This is skepticism as practice, not posture. It requires more effort than either credulous acceptance or blanket dismissal. It means reading beyond headlines, considering incentives, and tolerating uncertainty when certainty isn’t warranted. It means distinguishing between the claim that “we don’t fully know” and the claim that “we can’t know anything” – the first is honest, the second is evasion.

Something solid remains beneath the noise. The self-correcting nature of inquiry does work, just slowly and imperfectly. The decline effect – where early studies show large effects that shrink over time – demonstrates the process in action. Initial findings, often exaggerated, gradually converge toward truer values as evidence accumulates. Meta-analyses, despite their limitations, routinely overturn individual studies. Replication failures, painful as they are, drive improvement. The gap between what individual studies show and what accumulated evidence supports is itself instructive. One study is a hypothesis; a hundred studies are evidence. The problem isn’t that research is worthless but that isolated findings are overinterpreted – that the publicity machinery treats every significant result as a breakthrough rather than a preliminary signal. Better science communication would acknowledge uncertainty from the start rather than presenting conclusions with false confidence that later requires embarrassing revision. Night-shift workers don’t need researchers to validate their experience. They know what their bodies tell them, filtered through the practical constraints of their lives. Patients questioning a recommended procedure deserve more than eye rolls; they deserve honest engagement with the evidence base. The entrepreneur working through the night doesn’t need permission from sleep scientists to know that schedules bending toward natural inspiration sometimes produce results that schedules bending toward optimization don’t.

We arrive, finally, at a position neither naive nor cynical – one that takes expertise seriously without treating it as infallible. Research operates within ecosystems of funding, incentives, and social dynamics that systematically shape what gets studied and how findings are framed. The eight-hour sleep block emerged from industrial convenience. The twice-yearly dental visit emerged from advertising. The warnings about night work emerge from studies conducted by people who’ve never had to work nights. None of this invalidates the underlying science; all of it demands we hold conclusions with appropriate humility. What feels like timeless wisdom often turns out to be surprisingly recent, surprisingly contingent, and surprisingly shaped by forces beyond pure inquiry. The most honest response isn’t certainty in either direction. It’s the uncomfortable, productive position of taking findings seriously while remembering they emerged from institutions run by humans with interests, operating within economies that reward particular outcomes. Cynicism would dismiss this complexity, collapsing everything into accusations of corruption. But corruption isn’t the main problem – structure is. Good people operating within bad incentive structures produce systematically distorted outputs. Recognizing this neither absolves bad actors nor condemns the enterprise of research. It just suggests that we bring to knowledge the same critical eye we’d bring to any other human production: appreciating its value while remaining alert to its conditions.

The night-shift worker, returning home at dawn while the rest of the world wakes, doesn’t need to wait for scientific consensus to justify their schedule. They’re doing what circumstances require, contributing to the world’s functioning in ways that comfortable advice-givers rarely acknowledge. The gap between what “research shows” and what lives demand is one that individuals navigate every day, reading the tea leaves as best they can, making decisions without the luxury of waiting for perfect evidence. This is not cynicism. It’s the beginning of genuine understanding – an understanding that holds knowledge lightly enough to update when it changes, firmly enough to act upon when action is required, and humbly enough to recognize that today’s certainties may be tomorrow’s discarded hypotheses. The architecture of what we know rests on foundations that are sturdier than pure social construction but less solid than pure truth. Navigating it well means building that instability into our relationship with knowledge itself.

Featured song:

Caught this during a night shift. The jangling dissonance that almost throws it off balance is what holds it together.

The Six Month Dental Recall – Science or Legend? | Science-Based Medicine, Featured Review: How often should you see your dentist for a check-up? | Cochrane, Routine dental care: does the evidence give us something to smile about? | Cochrane, Should you floss your teeth? The evidence is surprisingly weak | STAT News, Flossing Isn’t Backed by Science? | Snopes, The Modern Origins of the 8-Hour Sleep Cycle | History, How working night shifts furthers health disparities | Becker’s Hospital Review, Replications of replications suggest that prior failures to replicate were not due to failure to replicate well | Center for Open Science, The Collapse of Ego Depletion | Michael Inzlicht, Sugar Papers Reveal Industry Role in Shifting National Heart Disease Focus | UCSF, Sugar Industry and Coronary Heart Disease Research: A Historical Analysis of Internal Industry Documents | PMC, How Tobacco Companies Created the Disinformation Playbook | Union of Concerned Scientists, Tobacco industry playbook | Wikipedia, Merck Manipulated the Science about the Drug Vioxx | Union of Concerned Scientists, Studies allege drugmaker manipulated data on painkiller Vioxx | CBC News, Ghost Management: How Much of the Medical Literature Is Shaped Behind the Scenes by the Pharmaceutical Industry? | PMC, The Secret History of Lead | Type Investigations, Thomas Midgley and the toxic legacy of leaded fuel | Chemistry World, Oil companies discourage climate action, study says | Harvard Gazette, Enshittification | Wikipedia, Dark pattern | Wikipedia, The Meaningfulness of Effect Sizes in Psychological Research | Frontiers, Replication Crisis: Challenges in Research | The Power Moves

Similar Post: Connective Trust
Image Source