April 21, 2021

"Music streaming platforms have sexism wired-in."

Jawad Iqbal writes in the London Times: 

Their algorithms, which recommend things you might like based on your listening habits, are basically sexist, generating playlist after playlist dominated by male musicians.

Isn't it like sexual preference — you really do respond to the sex of the singer? With no machine helping me at all, I can see in the music choices I am making that I prefer a male voice. 

Those damning findings...


... come from research conducted at Utrecht University in the Netherlands and Universitat Pompeu Fabra in Barcelona. Academics analysed the listening patterns of 330,000 users over nine years: only a quarter of the artists they listened to were women, because on average the first algorithm-recommended track was always by a man; listeners had to wait until the seventh or eighth song before hearing from a woman.

It's a problem only if you assume that the outcome should be equal. But isn't the algorithm attuned to what people have responded to?

Some of this bias is a reflection of historical failures in the music industry, which has always been dominated by male acts, save for occasional superstars such as Taylor Swift or Beyoncé.

Maybe we respond to what we respond to because it's familiar, and what is familiar is a consequence of sexist decision-making within the music industry. Why do we like what we like? Is it deep or is it shallow?

There's something called the "mere exposure" effect (Wikipedia):

The mere-exposure effect is a psychological phenomenon by which people tend to develop a preference for things merely because they are familiar with them. In social psychology, this effect is sometimes called the familiarity principle. The effect has been demonstrated with many kinds of things, including words, Chinese characters, paintings, pictures of faces, geometric figures, and sounds. In studies of interpersonal attraction, the more often someone sees a person, the more pleasing and likeable they find that person... 
In the 1960s, a series of Robert Zajonc's laboratory experiments demonstrated that simply exposing subjects to a familiar stimulus led them to rate it more positively than other, similar stimuli that had not been presented before....

In 1980, Zajonc proposed the affective primacy hypothesis: that affective reactions (such as liking) can be "elicited with minimal stimulus input." Through mere-exposure experiments, Zajonc sought to provide evidence for the affective-primacy hypothesis, namely that affective judgments are made without prior cognitive processes....

[One] experiment exposed Chinese characters for short times to two groups of people. They were then told that these symbols represented adjectives and were asked to rate whether the symbols held positive or negative connotations. The symbols the subjects had previously seen were consistently rated more positively than those they had not. In a similar experiment, people were not asked to rate the connotations of the symbols, but to describe their mood after the experiment. Members of the group with repeated exposure to certain characters reported being in better moods than those without....

Why do we like the pop songs we like, and what is Spotify doing to our brains refeeding us what we've been fed before? Is it evil — damnable? The question whether it's sexist is only a small part of the problem. That seems to focus on whether the singers are getting their due or are penalized by pernicious subordination. But the minds of the listeners are much more important. There are billions of us, and we're plugged into this system of endless feeding and refeeding. Within it, we have the feeling of being pleased. But neither the machine nor we understand why we are pleased.


There is no comments section anymore, but you can email me here. Unless you say otherwise, I will presume you'd enjoy an update to this post with a quote from your email.


Hi Ann, 
This is the first time I've written in, as I never registered as a commenter.

A few months ago, I realized I'd stopped listening to Pandora, when I used to listen to it almost every day. I had grown irritated with the service, and it sounds like Spotify has the same problem. It's not what I would call sexism so much as what I think of as the algorithm whirlpool, which is the same reason I dropped my Netflix subscription and get exasperated with other feeds.

The algorithm whirlpool goes like this. I have a song stuck in my head, so I put the name of the artist into a channel. I click like on the song, then move on to other work. I never give other feedback, unless I skip one or two songs, other than continuing to listen while using other applications. Say the song in my head has a tenor in a rock band. The first few songs are more tenors, more rock bands. After a while, I feel quizzical -- why so many tenors, so much guitar, so much of the same beat? It gets monotonous, especially as the playlist moves away from the original artist, into songs that sound like weaker and weaker knockoffs. A DJ who did this kind of playlist would be fired. This soundtrack would never sell for a movie. I found myself increasingly annoyed that I would only hear female artists if I put a female artist first, or only male artists if I went with a song by a male artist, when I'd rather hear a mix of voices and variations on style. The whirlpool, at first a fun ride, has sucked me down into an inescapable stream of music that all sounds too much the same.

I'd rather the playlist built in some diversity while staying in the same era of music. I also like instrumental music, but if I select a piano concerto, I'll only get to hear piano music. No violins, no orchestra, no trumpet. Just ... more piano, all with the same approximate tempo and dynamics.

You might wonder, why not just pick "classic rock" or "classical music" for my station? My answer is that then, I tend to get the more (to me) blah, overplayed stuff that I already got sick of hearing on the radio.

I don't consider this "sexism," but bad programming that doesn't allow for a randomizer escape. On Netflix, it was even worse. I'd watch one comedy special from a favorite comedian of mine. Then I'd be offered even more comedians from the same demographic as the first. I'd try a couple out -- these would also be fairly funny. Then I'd be offered more, more, more, until I wasn't even smiling at their rambling non-jokes. Each time I'd open Netflix, I'd be faced with unrecognizable comedians. I'm not obsessed with stand-up, but the Netflix algorithm had pushed us both into a whirlpool. Their old interface was more indifferent to my prior choices, and therefore less annoying. I dropped my subscription and don't miss it.

Currently, everyone in my Youtube suggestions is from Australia or New Zealand. All after I liked one cupcake video. Go figure.

What ticks me off, besides the annoying algorithm design, is that now if people start hearing the female voices (or the non-comedy specials, or the non-Australians) pop up in their feeds, they'll assume this is a cloying "eat-your-vegetables" directive and resent women vocalists. The programmers created the problem by failing to understand how a lot of us engage with media & enjoy a little bit of surprise or variety. Then their solution will be something that supposedly looks like they are fighting sexism, but is just a distraction from how boring their algorithm-directed feeds are. So the conversation will be about whether female or male voices are more pleasant, instead of how wealthy computer programmers are treating us all like my toddler, bringing me every shoe in the house to try to get me to play outside.

AND: James emails: 

I listen to a lot of songs on Spotify and I had no idea that the recommended playlist algorithms could potentially be sexist. I merely thought that they were lame.

But I did decide to put it to the test and see how long I had to wait to hear something from a female artist. For today's Daily Mix #1 it's not a good start with the first female artist Janis Joplin appearing at number 24! Mix #2 is a little more promising with the first female artist is Sidney Gish at number 4. Back to sexism with Mix #3 that has Lana Del Rey at number 16, but Chrissy Hynde fronting the Pretenders takes the number 1 spot on Mix #4. With an unsurprising ZERO female artists appearing on Mix #5 due to its heavy German techno-metal content, Mix #6 comes back strong with a number 1 Irma Thomas.

My unscientific and completely personal experience verdict: not sexist.

That made me look at my Daily Mixes from Spotify. I don't see a Mix#1, but in Mix #2, I have Nico twice in the top 6 and Lana Del Rey and Yoko Ono.  In Mix #3, 7 or the top 10 are females — Billie Holiday, Ella Fitzgerald, Dinah Washington, Etta James. The only males are Frank Sinatra and — doing "Whiter Shade of Pale" — King Curtis. Mix #4 is mostly male, and Mix #5 is all male. My unscientific and completely personal experience verdict: not sexist.