Females discriminated against in images on the Internet, study claims

f someone gave out a prize for the most improbable finding of the year, the claim by researchers from Berkeley, publishing in Nature mag, that women are discriminated against by pictures on the Internet ^[1] would be a shoo-in. The authors write:

The documented underrepresentation of women online is substantially worse in images than in text, public opinion and US census data. . . . Googling for images rather than textual descriptions of occupations amplifies gender bias in participants' beliefs.

Good points

The authors recognize that Google's “recommendation” algorithm makes it scientifically useless, so they try to compensate for it by creating new accounts. They claim to have looked at 840,669 images downloaded from Google, Wikipedia, and IMDb with the help of 6392 human coders from Amazon Mechanical Turk, which is a type of low-budget crowdsourcing for such menial tasks.

Flaws

Image of a woman found on the Internet, said to be an artist and photographer, painted by some guy

Now the flaws. They scored the “gender” of the person (by which they mean sex), as identified by their picture, on an axis of −1 = female to +1 = male and correlated it with the surrounding text describing the person's supposed occupation.

To the utter shock of everybody who lives under a rock, they found that most images of plumbers, devel­op­ers, and blacksmiths were of males and most images of beauty consultants, nutritionists, and cosmetologists were of females. Even more shockingly, most cheer­leaders, ballet dancers, and models were female, while detec­tives, football players, and police chiefs were mostly male.

Weirdly, ‘computer programmer’ was female, while ‘developer’ was male. The authors claim that these differences are evidence of “gender bias”, which they don't explicitly define. That suggests it's another self-reifying construct like “structural racism” where any difference, no matter what the cause, is proof of discrimination.

The authors seem to have overlooked “bureaucrat”, “porn star”, and “birthing person” (three popular occupations dominated by women) and evidently have never looked at science-oriented websites like Quanta, which feature women scientists at every conceivable opportunity. They deduce that women are “underrepresented” in online images compared to online texts from Google News and census data.

Next they administered an Implicit Association Test, a widely discredited measure of subconscious bias, to all the participants and claim that the images they examined increased their level of ‘bias’.

Aside from the obvious problem of correlating data from sources known to be severely biased with a test that is widely regarded as worthless, it raises the question of why anyone should be offended that there is such a thing as a male- or female-dominated profession.

Besides, news stories are far from neutral. Women, for example, are infrequently accused of rape or of invading nearby countries, which might result in their picture being plastered all over the news. And unless I missed it, there's no mention of the total number of images of each sex. This ratio varies dramatically between, say, the UK Telegraph and the Daily Mail. (I'm not familiar with US news sites like Google News, having dropped them all years ago as unreliable sources, so I can't speculate on any sex bias they might have.)

If we follow the authors' advice, we'd also have to drop the policy of showing the perp's photo instead of the vic's photo in rape and sexual harassment cases.

Silly conclusions

The authors conclude that Internet pictures will

affect the well-being of, social status of and economic opportunities for not only women, who are systematically underrepresented in online images, but also men in female-typed categories such as care-oriented occupations.

Okay then. Let's have Bettie Knott-Shoutt and Plastique Dellabomba as your plumber and garbage collector. Hire Luigi Mascarpone, Vespuccio Guancaiale and Antonio Pecorino as babysitters and see how your kid turns out. If the authors' assumption that men and women are identical is right, then women must be committing as many crimes as men, yet they're not being arrested enough. Clearly a correction is needed.

Of course the goal is not really to achieve equality, but to buttress the idea that men's and women's brains are identical—an idea that has been repeatedly disproved in science ^[2,3,4]. Now that intimate connections between the brain and the immune system have been found, there is no longer any doubt that male and female brains are different—not better or worse, just different—which as Jiang et al.^[2] point out highlights the importance of considering sex as a variable in brain research. The differences are so well established that they're becoming scientifically uninteresting.

Even before NIH went ‘woke’, clinical researchers were required to use equal numbers of both sexes whenever possible to be eligible for funding. The question as to whether sex is a variable is routinely considered in every study on the brain, and grant reviewers will nail you for not doing so. (Even NIH usually calls it ‘sex’ and not ‘gender’ because ‘gender’ is now devoid of meaning.)

Nature's crusade is that women should receive Nobel prizes in numbers equal to men. The temptation to manufacture a rationale for equal results while retaining the patina of merit that is still a strong value in the hard sciences must be overwhelming.

This is the dilemma the activists face. If men and women are identical in every way, there's no reason to include women in research studies, so they'll be discriminated against. If they're different, there will be differences in what they do, and that's also discrimination. It's an example of how activists play the game of heads I win, tails you lose.

This is not a left-right cultural battle. It's a battle about whether truth is established by scientific experimentation or by ideology. It's a battle the activists can't win. The most they can accomplish is to drag science into the mud and knock Nature off the list of trusted journals.

feb 21 2024, 7:52 am