A study performed by researchers in Europe found that a widely used music recommendation algorithm is more likely to choose music by male artists to the detriment of female artists.
Study authors tested a popular music recommendation algorithm across two song datasets. That process revealed that in both instances the algorithm ended up “reproducing” existing biases in the datasets. Results reveal both datasets displayed biases against women, with female artists only accounting for 25 percent of the songs.
In addition, the algorithm constructs a ranked list of songs for a user to sample. Across the board, the highest a female artist ever ranked on such lists was number six or seven.
The problem with this ranking is that researchers found most people usually end up listening to songs their playlists suggest to them. So, as an individual listens to more suggested songs, that reinforces the idea that the algorithm is doing something right, creating a “feedback loop” of gender bias.
The authors of the study propose a new approach that would allow greater exposure of female artists by manually reordering the recommendations, to ensure more female songs are suggested earlier.
In a simulation, the authors studied how classified recommendations would affect user behaviour in the long run. The results showed that, with the help of the reclassified algorithm, users would begin to change their behaviour and thus listen to more female artists than with other music recommendation algorithms and, moreover, the new algorithm, based on machine learning, would consolidate this change in behaviour.
It seems like a good suggestion to me. Algorithms are meant to be tweaked until you think you’ve got it right. So it will be interesting to see how the algorithm does in real life.
I checked in with Spotify and found what is referred to as Discover Weekly. Every Monday morning, Spotify listeners are greeted—some might say gifted—with a new Discover Weekly playlist to help set their soundtrack for the next seven days.
These algorithms look for how those songs are played and ordered in other Spotify users’ playlists. If it turns out that, when people play those songs together in their playlists, there’s another song sandwiched between them that someone has never heard before, that song will show up in your Discover Weekly.
So I did a bit of investigation and took a look at my latest Discover Weekly playlist. Of the 30 songs recommended for me, there were NO women artists.
Now I know that this is partly a function of what music I listen to, which is heavily dominated by male artists, but it’s certainly not void of women artists. But as the researchers pointed out, such a recommendation algorithm just perpetuates the bias.
How is a person to get exposed to new music, if these algorithms just recommend music and artists that are similar to what you already listen to?
So I can see how the manual reordering of the ranked list, with some female artists specifically included near the top of he Discover Weekly, would be helpful.
I would expect that these algorithms keep getting better and better, and so perhaps after a few months of manual reordering there may be an opportunity to let the algorithm work on its own.
In the meantime, I plan to play Gaslighter by the Chicks a few times every day over the next week to see what impact that will have on next Monday’s Discovery Weekly.
And if you haven’t listened to the song or seen the video, well, I think they are both marvelous (I have shared this before):