Reed Rosenbacher: How does shuffle actually work?

Monday, February 18, 2019 - 3:55pm

You have probably found yourself in the same position a number of times: dying to listen to music, but unsure of exactly what you want to listen to, so you pull out a favorite playlist and hit shuffle. But what does the shuffle button actually do? What does it mean to shuffle through songs? How do your imaginations of shuffle actually align with the reality of how it functions? What can this tell us about psychology’s infusion with consumer technology?

Originally, I thought shuffling was a randomizer that gave you as random a song as possible. For a while, this was how most shuffling programs operated, but streaming services like Spotify started receiving complaints that the shuffle button was not random enough. For the coders at Spotify, this complaint was perplexing, as their code was based on the Fisher-Yates Shuffle, which is often considered to be one of the best algorithms for randomization.

The people over at Spotify had a meeting and tried to dissect what the problem was. Their answer came down to an issue of human psychology. Imagine that someone flips a coin: If the person got heads five times in a row, on the next flip the person would expect to get tails because they have already flipped heads so many times. Of course, the rules of probability tell us that on the sixth flip, the chances of getting heads or tails is the same, as it is not influenced by the previous flips. But how does this relate to a Spotify playlist? Imagine that you are using a “truly random” shuffler and you hear “Gold Digger” by Kanye West. When you hit shuffle again, your expectation is that you won’t hear “Gold Digger,” but of course the probability of hearing “Gold Digger” is the same as any other song in the playlist.

Spotify realized that people didn’t want “true randomness,” but rather the human expectation of what should be random. So they recreated their algorithm based around clustering. The essential idea of the new algorithm is that when people shuffle music, what they really want is an equal spread of artists and genres over the course of the listening experience. If someone has a playlist with Kanye West, Prince, Lou Reed (who knows why you would), that person would not want to listen to 10 songs from the same artist in a row. Instead, they would want about three songs from each artist over the course of the ten songs.

For lovers of Apple Music, don’t worry — they have also adopted a similar mechanism of somewhat non-totally random shuffling. Yet, despite this intensive manipulation of the algorithm and how it aligns with the human psyche, I can’t help but find myself frustrated with the shuffle button. It often seems like songs deeper into the playlist don’t get played, and for some long playlists, I do want to hear the same song twice in a listening session. What's interesting about this is that it would be easy for Spotify to offer users the ability to select from different types of shuffling in the settings. Yet there is no customization feature, on any platform, for how the shuffling works.

While a certain part of me wants to reject the idea that tech companies like Spotify have the ability to understand how the human mind perceives randomness, I must admit that I would not want to hear the same song twice in a row on a shuffle playlist. Yet I still find myself feeling like the current algorithm is not random enough. It feels too calculated and cautious.

All of this manipulation of how shuffling works is indicative of a much larger trend going on in the tech world. Often times, tech companies are viewed as providing us with products that serve a function that seems to exist in a semi-objective way. Technology often hides under science’s guise of objectivity. For example, the order of Google search results, how your newsfeed (on any social media platform) is created, and even the infrastructure necessary to make the internet function is all hidden, obscured and made to seem a basic necessity rather than a privilege.

Tech companies no longer design their products to solve a problem, such as finding a piece of information. Their aims have extended much further. Now the information must be presented in the prettiest and most convenient way possible. Much of the consumer tech world is dominated by design decisions that attempt to profit off the functions of the human psyche. Shuffle play is a “fake random” button that manipulates how our brains function, which underscores a more interesting story about what makes consumer tech succeed. 

Reed Rosenbacher can be reached at rrosenb@umich.edu.