Invisible Women: Data Bias in a World Designed for Men

Products are consistently designed to accommodate the male body. Piano keyboards are designed for the average male hand size, a fact which disadvantages female players and results in greater incidences of pain and injury. While a keyboard for people with smaller hands has been designed, the piano world has not embraced it. Smartphones are also designed for the larger hands of men: The average man can use the device one-handed to take a picture, while the average woman cannot. Voice recognition software is “often hopelessly male-biased” (162), failing to respond to the higher pitch of female voices because it is trained on corpora, or databases of voice recordings, which skew heavily male. Text corpora, used to train translation software and web search algorithms, are also male-biased. In image datasets, men also greatly outnumber women.

Women are both under-represented and misrepresented in data sets. As a result, stereotypes are being incorporated into artificial intelligence, as algorithms not only reflect male biases, but amplify them. With algorithms now scanning resumes, these male biases can prevent women from gaining employment. Meanwhile, the increasing use of artificial intelligence in medicine, “so heavily skewed towards the male body” (167), could decrease the accuracy of diagnoses for women. Poor product design disadvantages women in the acquisition of jobs, the ability to do jobs effectively, and in their health and safety. 

Venture capitalists, 93% of whom are men, are more likely to invest in other men than women and dismiss proposals for companies targeting primarily women. For example, despite the fact that one company “cornered the market” (170) in breast pumps with a deficient product, which is painful and slow to use, venture capitalists did not seriously consider a better design for this lucrative market.

There is a data gap about the female body. For instance, although 50% of women over 50 require surgery because of pelvic-floor issues, there is very little study of this problem. There have also been no real advances in birth control since the development of the pill, which causes side effects in many women. Menstruation is not studied at all, so there is little knowledge about normal bleeding patterns for adolescent women. Trying to address this data gap with research or better products, women run into a catch-22: Already disadvantaged because they do not fit venture capitalists’ perception of developers, whom they expect to be white men in hoodies, women are less likely to have data to support their plans, given the lack of previous research.

Dominated by “a sea of dudes,” the tech industry all too often manufactures products that fail to cater to women’s needs. For example, the Apple health tracker initially did not account for women’s periods, Fitbits fail to include activities such as pushing a baby stroller, and some robots are frightening to women. Video gaming has a sexual harassment problem, but game developers were unaware of it until it was brought to their attention—only then were design changes made to prevent virtual assaults. New technology like virtual reality devices does not take into account that women are more likely to experience motion sickness because men and women rely on different mechanisms for depth perception, and 3D environments send “signals that benefit male over female depth perception” (183).

The design of automobiles is also geared toward men. Women are more likely to be seriously injured or die in a car crash because car safety testing uses male crash-test dummies, despite the fact that women have different muscle-mass distribution, lower bone density, and vertebrae spacing. Relying only on male dummies means these differences are not accommodated in safety designs. No seat belt works well for pregnant women, despite car crashes being the “number-one cause” (188) of fetal deaths. Automobiles and other products are created without reference to the needs of women and as a result, fail them.

The medical system discriminates against women from “root to tip” (196). Physicians are trained to focus on the male norm—medical education describes women, when they are mentioned at all, as variation on the standard. Medical textbooks use male bodies three times as often as female ones, according to a 2008 study of prestigious universities in the US, Canada, and Europe (197). Curricula do not adequately teach medical students gender-related issues. This gap matters because sex differences are pervasive, including lung capacity, the workings of the heart, the experience of common diseases, and organ systems.

There is a dearth of sex-specific information because women are routinely excluded from or vastly underrepresented in clinical trials and medical research. Generic drugs are tested less rigorously, typically only on young adult males with concern only for the active ingredient. Yet different inactive ingredients can impact a drug’s effectiveness, and women are more likely to experience a different reaction from generic substitutes. Most drugs are also not tested on women at different stages in their menstrual cycles, which could matter. For women, the second most common adverse reaction to a drug is that it simply does not work. Noting that researchers found that women responded differently to estrogen while fighting off a virus, Perez wonders “how many treatments” women have missed “because they had no effect on the male cells on which they were exclusively tested” (207). Even studies for female-prevalent diseases woefully underrepresent women. The lack of sex-disaggregated data harms women and prevents doctors from giving good advice.

Animal testing similarly relies primarily upon male subjects even though on those occasions when animal sex was considered, differences have been found. For example, one study demonstrated opposite results about the timing of heart attacks and survival in male and female mice.

Types of exercise beneficial for women and men vary as well. Resistance training, for example, is bad for men who have high blood pressure, but good for women. Female diabetics are not helped by high-intensity interval training, while male ones are. Given the significance and impact of sex differences in health, some countries have mandated the inclusion of women in trials, yet there remain loopholes to such requirements. It is scandalous that the medical world is complicit in the unnecessary suffering and death of women.

Named for the movie in which an actress played a Jewish woman who pretended to be a man so as to receive an education, Yentl syndrome is “the phenomenon whereby women are misdiagnosed and poorly treated unless their symptoms or diseases conform to that of men” (217). Heart attacks are a prime example, as men and women present with different symptoms and common preventative medications do not work as well for women. Diagnostic tests, originally developed for men, are problematic for women in other areas as well. For instance, colonoscopies are more likely to be incomplete on women, who on average have longer and narrower colons than men. Among infectious diseases, tuberculosis is the leading killer of women globally, yet women are less likely to be screened for it and are often misdiagnosed when they are screened because their symptoms present unlike those of men.

In preventative efforts, medical practices do not adapt to female socialization. As a result, autism spectrum disorder and attention deficit hyperactivity disorder are underdiagnosed in girls. When women voice their concerns, doctors do not always believe them. If their symptoms do not match a male pattern, doctors tell women that their pain is imagined and are more likely to prescribe anti-depressants than pain killers—the most recent manifestation of a long pattern of declaring complaining women hysterical and placing them in insane asylums for behaviors “out of feminine bounds” (225).

Undiagnosed and untreated pain in women is often a result of exclusively female diseases or symptoms more common in women than men. Very little research has been done on medications for premenstrual syndrome or dysmenorrhea (period pain), despite the fact that the latter impacts 90% of women on a monthly basis! Although the drug sildenafil citrate, commonly known as Viagra, was found in a 2013 study to provide complete relief for dysmenorrhea for four hours with no side effects, pharmaceutical companies, fearing that these findings would impact the lucrative market for men, have failed to conduct more trials. More research is sorely needed on uterine failure, on means to reduce the need for caesarean sections, and on why African American women have high maternal mortality rates. The data gap on women’s health is causing women to be “dismissed, disbelieved and ignored” (234). It is time to start listening to women and doing research on them.