When Russian and Hong Kong coders, supported by Microsoft and Nvidia held a beauty contest judged entirely by an algorithm, 600 000 people from around the world sent in selfies and waited for judgement day. Oh boy. It did not go quite as planned. Of the 44 people judged to be the most beautiful, all were white, except for 6 Asians. Only one woman had visibly dark skin, although in terms of skin color, it’s still pretty light.
Both MotherBoard and the Guardian are scrambling to explain how this might have happened, arguing that even though people did send in selfies from Africa, India, China and all over the US, the sample sizes skewed white, and because the AI was using a deep learning technique, this affected the outcomes. Another term for deep learning is reinforcement learning, so this explanation does kind of make sense. If the largest set of base images the algorithm is using features white faces, then whiteness will become a criteria for ‘beauty’, in addition to things like facial symmetry and skin wrinkling.
When you consider a totally separate study that showed a machine learning to associate Black sounding names (Jamal, Ebony) with unpleasantness and white sounding names (Matt, Emily) with pleasantry, the case seems quite solid that seemingly neutral machines do indeed pick up and replicate our prejudices. Parents who give their children absurd, pretentious, or ridiculous names have always been a pet peeve of mine. I sneer heartily at moms and dads who think giving their kid a ‘cool’ name will somehow make the parents themselves look cool. To me, those parents look insecure as hell and are already making fundamental parenting mistakes. Your child is a completely separate human being, and not some sort of fashion accessory to enhance your own prestige. Angus, Rufus and Pearl are not status symbols. They’re human beings.
I’m not the first person to notice that Black parents in particular are prone to endowing their children with unfortunate monikers. The Black blogger behind Stuff Black People Hate (according to Salon – the link no longer works FYI) broke the names down into subcategories:
“Swahili Bastardizations” (Shaquan)
“Luxury Latch-ons” (Prada)
“Megalomaniacal Descriptors” (Heaven)
“The Unfathomably Ridiculous” (Anfernee)
I understand there are some very good reasons behind how and why Black Americans in particular end up with names that sound jarring and strange, but the end result is that the names are jarring and strange. They are unexpected. And humans really don’t like unexpected things. My own name happens to be a man’s name in the Greek and Italian language, and a woman’s name in the German language. To complicate matters, it’s pronounced the same as the Greek and Italian, while the feminine English form has a different pronunciation. I am reconciled to snickers from Greek and Italian speakers when they first hear my name. It would be akin to meeting an Italian girl named Hans. Ridiculous.
So part of the issue with naming is our collective expectation of what names should be, and our skepticism and often derision aimed at names that break with onomastic traditions. This is very much visible in terms of class, too. “Stripper” or “redneck” names earn as much scorn as Black names, again, largely for the same reason. Chardonnay is a type of grape, not a name you give a child. Bodean is just terrible. Don’t. These are jarring and unexpected. Cultures can mess up names quite nicely, too. Rong is a popular boy’s name in China. Wei is a common surname. No one batted an eye at Rong Wei in China, but in Canada, Toronto based paralegal Rong-Wei Yu wisely goes by Alec. Mohammed is pretty much a terrorist at this point. And if your last name is Stroker, William may not be the best choice you could make for your son.
Do machines learn these prejudices and preferences? Yep. They do. Does that make the machine racist?
I’m not so sure it does. Emily is a distinctly American name that is open for anyone to use. Emily can be a Wong, an Esperanto, a Smith, a Gorski or an Abu Bakhari. Every Emily makes sense – Emily is hallmark of the melting pot. Anyone can be Jonathan. Some names are universal, and it is their very universality that causes machines to mark them as pleasant. They do not jar or provoke us. Emily is pleasant, kind and any race or class at all.
When it comes to the issue of physical beauty, is it similarly possible that the universal features we tend to consider beautiful are expressed more easily in white faces? If we look at the Golden Ratio, for example, no matter what the race of the person, their features tend to ‘look white’.
The ratio between nose width and face width, for example, indicates that smaller noses will appear more beautiful.
Black faces can indeed be stunning, but only if the smaller nose is present.
Larger noses are considered unattractive, no matter what color the skin.
Is the Golden Ratio racist? If it’s a universal expression of beauty, how can it be? Consider, too, that markers of youth and fertility, such as blushing or lip color are easier to detect the lighter the skin. Beauty is a rough synonym for fertility and health. The desire for health and fertility is hardly unique to any one race.
These may be uncomfortable conversations to have, precisely because we do indeed live in a world where many Black people are treated unfairly and discriminated against because they are Black. Racism is a thing, but that doesn’t mean everything is racist. It might be superficially satisfying to scream racism when algorithms prefer white names and white faces, but the issues are far more complex than that and quite frankly, interesting.
Truth must always be our guide. If we stop caring about, seeking out and valuing truth above all else, we cannot progress. Personally, I would rather live in a world where people feel uncomfortable handling truth but still do it, and I’ll spend my time advocating that we pay attention to truths that cast our own groups in poor light, and not just others.
Sure, ‘white’ features might be more beautiful, but why are our kids so dumb compared to Asians? Why are Black people so astonishingly innovative in music? And given that musical ability is so strongly correlated with mathematical ability, why aren’t the majority of PhD holders in mathematics and Field’s Medal winners Black? These are interesting questions, worth exploring.
The truth isn’t always going to be pleasant. When was it ever? That should not stop us from seeking it out.
All truth matters.
Lots of love,
Out of curiosity, I submitted my own image at anaface.com. Here’s what I got: