Default by Design: How Science Went Wrong by Testing Only White Men

(HH Original — Science, Society, and the Bias of Sampling)

Introduction – The Myth of the “Average Human” For decades, the so-called “average human” in biomedical research has been a white man. Not a population average, but a demographic default. Nearly every pill, medical device, and dosage guideline was once calibrated to fit that singular model—an assumption so deep it went unnoticed for most of modern medical history. The irony is bitter: by studying a narrow sliver of humanity, science claimed universality and produced misinformation that has literally cost lives.

Crash-test dummies, for instance, were based on the proportions of a 5’9”, 170-pound male. As a result, women involved in car crashes are 73% more likely to be seriously injured than men. In cardiology, clinical trials long ignored women, leading to generations of physicians trained to recognize male heart-attack symptoms—chest pressure, radiating arm pain—but not the subtler signals more common in women, like shortness of breath or nausea. The bias didn’t just exclude; it misinformed, seeding error into every level of care.

The Literary Reflection – The Absence That Speaks When the first studies of aspirin’s preventive effects on heart disease were published in the 1980s, they were hailed as triumphs of modern epidemiology. The trials were massive, carefully controlled, and entirely male. Researchers concluded that aspirin prevented heart attacks—but when women were finally included years later, it turned out aspirin’s protective effects looked very different—benefiting men mainly by preventing heart attacks and women mainly by reducing stroke risk, with little heart-attack protection for women. The science hadn’t been wrong; it had been incomplete. Yet incomplete science, when treated as universal, becomes falsehood.

The same pattern repeats across fields: in psychiatry (where ADHD in girls is underdiagnosed because early models were built on hyperactive boys), in pharmacology (where drug metabolism varies by sex and melanin-linked enzyme expression), and in ergonomics (where “one size fits all” masks structural inequity). The result is a cascade of distortions: bad data begets bad policy, and bad policy codifies injustice.

Systems-Level Context: How Bias Became Institutionalized

The dominance of white male subjects in research wasn’t a conspiracy—it was a convenience that hardened into dogma. In the mid-20th century, when medical testing ramped up in universities and government labs, most test populations came from the easiest source: nearby institutions. That meant college students, military personnel, and hospital patients—all groups overwhelmingly white and male. Their availability and “controlled” environments were seen as scientific virtues, not limitations. What began as logistical shorthand became a structural blind spot.

In pharmacology, this bias was amplified by fear of legal liability. After the thalidomide disaster of the 1950s, which caused birth defects in thousands of infants, women of childbearing age were routinely excluded from trials. The intention was to protect women; the effect was to erase them. Researchers generalized from male-only data, assuming metabolism, hormone cycles, and immune responses would be “close enough.” They weren’t.

Funding patterns reinforced the problem. Early National Institutes of Health (NIH) grants rewarded studies that could produce clean, easily replicated data. Variability—such as menstrual cycles or differences across ethnic groups—was treated as statistical “noise.” But human variation is the signal. By smoothing away the texture of humanity, researchers created models that fit the lab, not the world.

The pattern extended beyond medicine. Equipment designed for military and industrial use was modeled on the “average male body.” Even facial-recognition technology today, built on datasets dominated by lighter-skinned faces, inherits the same epistemological flaw: assuming that narrow data can speak for everyone. In this way, the white-male research paradigm became not just a medical problem but a systems problem—a network of institutions, incentives, and assumptions perpetuating itself through inertia.

The Consequences: When Universality Becomes Falsehood

The most insidious effect of exclusionary science isn’t omission—it’s distortion. When an entire field treats one body type, one skin tone, or one hormonal profile as universal, the resulting “truths” mislead both patients and practitioners.

Consider cardiovascular disease, long described as a “man’s illness.” For decades, doctors were trained to look for crushing chest pain as the telltale symptom. Yet women often present differently—with fatigue, back pain, or nausea. Because those signs didn’t match the “classic” pattern, women were—and still are—more likely to be misdiagnosed or sent home untreated. Lives have been lost not to a lack of knowledge, but to the false certainty that came from partial data.

The problem runs deeper than any single case. When dose-response curves are drawn from homogenous samples, drug safety thresholds skew. A 70-kilogram male body metabolizes antidepressants and anesthetics differently than a smaller or differently pigmented body might. Yet until the 1990s, these distinctions were rarely accounted for. The same was true for occupational safety equipment and military gear—masks, helmets, flight suits—all sized and calibrated for that same “standard male.” A woman soldier or a worker of smaller stature was left literally unprotected by design.

Even psychology absorbed the bias. Diagnostic frameworks like the DSM evolved largely from studies of white male patients in Western contexts. The consequences reverberate across cultures and genders: conditions like depression, autism, or ADHD manifest differently in women, in Black patients, in Indigenous communities. When those presentations diverge from the “norm,” they’re often dismissed, delayed, or misread.

What emerges is a feedback loop of misinformation. Flawed sampling yields biased data; biased data shapes medical education and device standards; those, in turn, define “normal” physiology. Once canonized in textbooks, that false normal becomes nearly impossible to dislodge.

The Correction Efforts: Science’s Long Overdue Course Adjustment

In 1993, after decades of pressure from women’s health advocates, the U.S. Congress passed the NIH Revitalization Act. It was a bureaucratic correction to a moral and scientific failure: the act required federally funded studies to include women and people of color. It should have been revolutionary. Instead, it became a slow crawl toward adequacy.

Even with legal mandates, participation gaps persist. As of 2020, women still made up less than 40% of clinical trial participants in cardiology and oncology, and data on racial and ethnic representation remain incomplete for roughly one in five studies. Progress has been uneven because the bias isn’t just in recruitment—it’s baked into the frameworks of analysis. Researchers often treat race or sex as variables to be “controlled for,” as though eliminating difference were the goal of science. But understanding difference is the goal.

The new frontier—precision medicine—aims to correct this error by designing treatment around individual genomes, microbiomes, and social determinants of health. It’s a step in the right direction, though even genomic datasets are still disproportionately Eurocentric. Similarly, artificial intelligence models used to predict disease risk or recommend treatment inherit the same blind spots as the data they’re trained on. Without conscious correction, tomorrow’s algorithms will replicate yesterday’s prejudice at scale.

There are bright spots. The All of Us Research Program, launched by the NIH in 2018, explicitly seeks to enroll one million participants reflecting the full diversity of the United States. New statistical methods now treat variability not as noise but as a source of insight—mapping the ways environment, ancestry, and gender identity interact to shape biology. These efforts signal a quiet revolution: the recognition that there is no “average human,” only patterns within plurality.

Beyond Medicine: When Bias Becomes Blueprint

The story of exclusion doesn’t end at the clinic door. Once normalized, the white-male default propagated through every field that claimed to measure, optimize, or predict human behavior. What began as a scientific shortcut became a cultural blueprint.

Technology and AI Early facial-recognition systems were trained primarily on images of light-skinned men. The result was predictable: error rates for darker-skinned women reached as high as thirty-five times those for white men. Similar problems plagued voice-recognition software, which initially failed to understand women’s higher-pitched voices or regional accents. These digital biases are the grandchildren of twentieth-century lab bias—data chosen for convenience mistaken for objectivity.

Ergonomics and Engineering The industrial world inherited the same narrow model. From hard hats to safety harnesses, tools and protective gear have long been built for the average male body. Female firefighters still report ill-fitting gloves and breathing masks. Even the “female” crash-test dummy is a scaled-down male, not an anatomically accurate model. Design bias quite literally shapes which bodies survive accidents.

Economics and Policy Bias seeps into spreadsheets as well. The dominant economic model—the rational, self-interested actor—mirrors the assumptions of the male labor market of the mid-twentieth century. It overlooks caregiving, unpaid domestic work, and the invisible labor that sustains entire economies. When policymakers rely on such models, they inadvertently codify inequity under the banner of neutrality.

Architecture and Urban Design Cities, too, have their blind spots. Transit schedules, lighting plans, and zoning laws were historically designed around the movements of male workers. Women’s travel patterns—shorter, multi-stop, often after dark—were left out of the data, making safety an afterthought. The built environment, like the body of science that inspired it, reflects who was counted and who was not.

These examples reveal the same structural truth: exclusion is contagious. When a system learns from a narrow sample, it reproduces that narrowness in every decision it makes. Bias, once encoded, becomes invisible until we ask the simplest Humboldtian question—Who was missing when this was made?

Sidebar: Invisible Bodies: A Timeline of Exclusion and Reform

1940s–1960s: The Age of Convenience Clinical trials in the postwar era rely heavily on “available populations”: soldiers, medical students, and hospital interns. White men are overrepresented, creating a false scientific baseline that soon spreads across disciplines.

1957–1962: Thalidomide and the Protective Exclusion The thalidomide tragedy—thousands of birth defects caused by an inadequately tested drug—prompts researchers and regulators to exclude women of childbearing age from clinical trials altogether. Intended as protection, this policy silences half of humanity’s physiology for three decades.

1977: The FDA’s Ban on Women of Childbearing Potential The U.S. Food and Drug Administration formalizes the exclusion of women from early-phase drug trials. The ban remains in place until 1993, ensuring that almost every drug dosage and side-effect profile approved in those years is calibrated to the male body.

1985: The Women’s Health Movement Pushes Back Grassroots advocacy—sparked by figures like Dr. Bernadine Healy and organizations such as the Society for Women’s Health Research—challenges the “one-size-fits-all” model. They highlight how this bias endangers women’s lives in heart disease, autoimmune disorders, and cancer.

1993: The NIH Revitalization Act Congress mandates the inclusion of women and minorities in federally funded medical research. It’s the first formal acknowledgment that “neutral” science had, in practice, been exclusionary.

2000s: From Diversity to Data Correction Awareness expands beyond gender: trials and surveys begin to track race, age, and socioeconomic status as essential, not optional, variables. Still, many studies remain demographically narrow, and reporting inconsistencies hide the problem.

2010s: Algorithmic Bias and AI Medicine As machine learning enters healthcare, new forms of exclusion appear. Algorithms trained on biased datasets produce flawed predictions—reinforcing racial disparities in disease risk assessments, insurance approvals, and diagnostic tools.

2020s: Toward Precision and Plurality Initiatives like the All of Us Research Program, the WHO Gender, Equity and Human Rights initiative, and open data movements aim to redefine “representative science.” The emerging ideal: not uniformity, but fidelity—capturing the true diversity of human biology and experience.

Systems Reflection – Bias as Infrastructure The bias toward white male test subjects isn’t just a matter of who was studied—it’s a mirror of how systems perpetuate themselves. Once built, infrastructures defend their own assumptions. In science, that means data models, funding incentives, and peer-review norms all reward “consistency” over variability. The error is systemic, not individual. Every textbook and guideline derived from those biased studies becomes a node in a vast feedback loop of misinformation.

Correcting it requires more than inclusion; it requires humility. When researchers treat difference as disruption rather than information, they misunderstand the very nature of biological systems, which thrive on diversity. A forest with one species is fragile; so is a science built on one kind of body.

The larger lesson is deeply Humboldtian: everything connects, and bias anywhere distorts knowledge everywhere. Medicine, education, design, and policy all depend on accurate models of the human experience. If those models ignore half the population—or entire populations—they cease to be models of humanity at all.

Classroom Prompts

• How did convenience and caution combine to create exclusionary norms in medical research?
• What are examples of “universal” truths or standards in science that may actually be biased by narrow data?
• In what ways can inclusion make research more accurate rather than just more equitable?
• Discuss a field outside medicine (e.g., AI, economics, architecture) where bias in sampling or modeling has produced systemic misinformation.
• How might the Humboldt’s Home concept of interconnectedness apply to correcting bias in science?

Closing Reflection – Seeing the Full Picture For most of history, science sought certainty by narrowing its focus. It mistook control for truth. By studying the easiest bodies to recruit—the white, male, institutionalized—researchers simplified the complexity of humanity into something measurable but misleading. The consequences have rippled across medicine, public health, and trust itself.

Reform begins with a shift in perspective: to understand complexity not as chaos but as pattern. Inclusion isn’t charity; it’s accuracy. In the language of Humboldt’s Home, every system—biological, social, or scientific—depends on the full network of its parts. To leave any out is to misunderstand the whole.

(Click on links for verified sources)

Sources (Annotated)

• NIH Revitalization Act of 1993 – The landmark U.S. legislation requiring inclusion of women and minorities in federally funded clinical research. A pivotal turning point after decades of exclusion.
• Caroline Criado Perez, Invisible Women: Data Bias in a World Designed for Men (2019) – Groundbreaking book documenting how male-centered data skews everything from urban planning to medicine.
• United States Government Accountability Office (GAO) Report (2022) – “Better Oversight Needed for Inclusion in Clinical Trials” – Reveals ongoing underrepresentation in NIH studies.
• Aspirin and Heart Disease Studies (1989–2005) – Comparative trials illustrating sex-based differences in aspirin’s protective effects; demonstrates how omission leads to misinformation.
• The All of Us Research Program (NIH, 2018–present) – An ongoing effort to build a national health database representing all U.S. populations, redefining what “average” means in science.
• Joy Buolamwini and Timnit Gebru, “Gender Shades” (2018) – Study exposing racial and gender bias in facial-recognition AI; a modern echo of the same structural error in digital form.

© 2025 Michael A. Pink