Everybody loves numbers—they give us the feeling that we’re objective, rational. But a quick look at the history of the BMI shows us exactly why we should always take data-driven claims with an element of scepticism.
The BMI, or “Body Mass Index,” is a simple metric. Take a person’s weight in kilograms, divide it by the square of their height in meters, and you have their BMI. My BMI, for example, is 26.5, placing me in the “overweight” class. Some of my fellow women might be surprised by that, given my dress size is an Australian 10 to 12 (an American 6 to 8).
I lift weights about twice a week. Muscle weighs more than fat, so if I ever start getting to competition level, I’ll probably move up a class—into obese.
Most designers have a certain degree of suspicion when it comes to anything “data-driven” or even “data-led.” It can be difficult to articulate, especially in tech-oriented organizations that value measurable outcomes. In such cultures, a collection of findings from qualitative research, however robust, can’t compete against the reassuring objectivity of quantitative data.
But that doesn’t mean the suspicion is unfounded. A culture that values data is just that—a culture. And all cultures have histories.
All technologies have histories, too, including those used in data collection. Many of the first units of measurement were based on human bodies. A cubit was the length of a forearm, a span was the distance between thumb-tip and little finger, and a foot speaks for itself. Collecting measurement data from human bodies, on the other hand, requires a different kind of tool: standardized measurement. The lengths of feet can differ, the length of “a foot” cannot.
Measurement systems are hard to shift, and we often need seismic cultural events to make the change. The metric system, for example, is a direct result of the French Revolution. In 1790, the French National Assembly commissioned the creation of a universal, rational set of constants, and the metric system is the result.
A meter, they decided, is one ten-millionth of the distance from the North Pole to the equator, and a kilogram is the weight of one liter of water at freezing point.
French National Assembly. First order of business—end feudalism. Second—can we stop measuring things in feet? | Image courtesy of worldhistory.org.
Introducing base ten measurements allowed all sorts of previously impossible calculations, especially in statistics. One enthusiast was Belgian mathematician, astronomer, criminologist, and statistician Adolphe Quetelet, who, in 1832, came up with “the Quetelet Index”—the exact same height-to-weight ratio that we call the BMI today.
Like his intellectual forebears in the French Revolution, Quetelet believed that rational, mathematical methods could make sense of the messiness of human life. Math had already done a great deal to find almost magical solutions to complex problems.
Carl Gauss, for example, proved that the probability of events could be drawn on a “normal curve.” In mathematics, a normal curve is any distribution of data that shows symmetry around the average value. If a hundred people all flip a coin a hundred times, a few will flip mostly heads, a similar number will flip mostly tails, but the majority will flip a more-or-less equal number of heads and tails.
Events closer to the center of the curve are more representative of the true nature of the thing being measured, and those further away are less so. Gauss’ probability curve is easy to recognize: today, we call it “the bell curve.”
Imagine Quetelet’s surprise when he discovered that his metric measurements of people’s weights and heights fit perfectly with Gauss’ bell curve. The conclusion seemed obvious: if the center of Gauss’ probability curve represented something’s most accurate form, then the center of Quetelet’s curve represented the perfect person. More than that, the further away from the center a person’s measurement was, the less perfect they must be.
Adolphe Jacques Quetelet | image courtesy of creative commons.
“If the average man were completely determined, we might, as I have already observed, consider him a type of perfection: and everything differing from his proportions or condition, would constitute deformity and disease.”—Adolphe Quetelet, A treatise on man and the development of his faculties.
Emboldened by his success, Quetelet searched for—and found—bell curves in many aspects of the human state. He collected data for identifying the perfect face, intelligence, character. Unsurprisingly, his work would later be foundational to the rise of eugenics and so-called scientific racism.
It was not until the 1970s that Quetelet’s Index made its way into health care. Ancel Keys, an American physiologist and dietician, conducted a study that showed the Index—which he renamed “Body Mass Index”—was a good way to screen people for obesity quickly. Good if you were a man, at least: while Keys’ research cohort included men from Finland, Italy, Japan, America, and Bantu men from South Africa, it contained no women.
Which is to say, it contained no one from the 50 percent of the human population who have proportionately higher body fat than the half that was measured.
For his part, Quetelet would have been shocked by this outcome. His purpose was to find patterns and solve problems at the population scale, not at the level of individual people.
He would have been less shocked at the exclusion of women, as his own measurements for the original Index were all on French and Scottish soldiers—all of whom, at that time, were men.
The BMI has many downsides for people at all levels of health. It has a lot of upsides, however, for doctors. Health decisions can be made easily with only a scale based on kilograms, a tape measure based on centimeters, and a quick calculation. Based on a person’s BMI, eligibility criteria for certain medicines are met or not met, fitness for some surgeries approved or not approved, access to health services granted or denied.
That 67 percent is an objective, reliable fact. What it isn’t is a fact with objective, reliable foundations.
This isn’t the only instance of a metric providing information that isn’t fit for purpose. Still, for those of us who do human-centered work, especially in tech-oriented organizational cultures, it’s evidence that our suspicions are, at least sometimes, correct.