Is your music taste more or less your own in the age of algorithms?
When Spotify is placed upon the scales of justice, the one enormous weight acting in its defence is the experience it provides for users. Almost everyone would agree that there are ethical question marks that need addressing, but that doesn’t stop more than 600million of us – around 8% of the world’s entire population – from subscribing… and, when prying eyes are out of sight, describing the maligned music repository as a beautiful Godsend. It has changed the way we listen to music, with pretty much every song ever written – (and, troublingly, a few that haven’t even been written at all) – a click away.
To what extent has it changed our music tastes in the process? It is undoubted that thanks to streaming services, we listen to more music than ever before. Personally, I listened to around 9,000 artists last year. It is hard to imagine that a radio station with a broad stroke of listeners to satisfy could offer such a breadth. Moreover, if I’d been buying all of that music physically, then I’d be so deep in debt I’d make Mike Tyson’s finances look healthy. But is this scope helping to find the real ‘me’, or am I being fed the same mildly satiating setlist as my neighbour, Ian the Bastard, with whom I hopefully have little in common?
This presentiment arose recently when I first uncovered ‘Groovin’’ by The Young Rascals—a gift from the algorithm gods. It arrived as a delightful ray of 1960s sunshine, prematurely heralding spring. It seemed like it was meant just for me, as though the band themselves thought, ‘Tom will like this’, decades before I was born. In actuality, it was delivered by an unthinking machine to millions of others that very same week, evidenced by the fact that my friend recommended it to me a few days before I had uncovered it myself in a text message I hadn’t bothered to read. So, when I recommended it back to him, it was like Neil Armstrong stepping onto the moon and having Uri Gagarin shake his hand.
This strange modern happenstance plunged me into a rabbit hole. Did it matter that the recommendation wasn’t as personal as it once seemed? Why was it that I was now recalibrating what was a sure-fire victory for the algorithms into something far more sinister? In order to clear up my plight, I sought the help of an anthropologist, Nick Seaver, the music algorithm expert behind the book Computing Taste, and I asked him the simple question: are our music tastes more or less our own in the age of algorithms?
My thinking on the predicament could be summarised as thus: did grooving zap me because it broke through the stream of satisfying sameness evidencing the fact that algorithms cloister us into being pleased by what we know but rarely invigorated by something new? Or was it merely further proof of the power of a personalised curation method that is always good and sometimes great?
In Seaver’s view, while this might seem like a question related to algorithms, there is a preface that stretches way beyond Spotify. “In any context where we have taste, we have taste in a world with other people in it,” he suggests. “No matter what, we’re always learning about music from outside of ourselves. The music is made by other people. Then, we decide what kind of person we want to be in a world full of other people. There’s a lot of sociology work about how taste maps to social status. I don’t think it makes sense to think of taste as something that is uniquely our own.”
“Taste being our own is a myth,” he continues. To frame this in a world seemingly less subjective than music for a second: do I even like ramen inherently, or do I just like it because I think of myself as the type of guy who would like it? In truth, it’s a bit of both: the first time I had ramen, it was drunkenly foisted upon me by my friend when my guard was down. In this instance, he was the algorithmic enforcer of my future taste. Alas, was this recommendation conditioned by my friend saying, ‘You will like this’, or were my taste buds already tingling to the tune of ramen long before I had it?
The same can be said for music in the age of generated recommendations. In the case of ramen, maybe I had been putting it off because of some nebulous prejudice that my friend helped to expel. In the case of music, that might relate to a band that I hadn’t bothered with because of any number of petty preconceptions but that were now waning because a system was urging me to give them a chance. Raising the point that the only difference is whether I trusted my friend – flesh and blood – who knows me well and whether his recommendations have repercussions – more than the faceless void of binary? Well, how can you not trust an algorithm when all it knows is everything you’ve ever told it?
The headache doesn’t stop with that rhetoric, though. To use the horrid parlance of our times, if musical discovery is a ‘journey’, then maybe algorithms knowing us only send us down a one-way road. Eternal variations of ramen. As Seaver explains: “There is this question about whether systems are more actually personalised in a way that lets you change your taste or are they pigeonholing you?”
He continues: “This question is the huge tension in music recommender systems. The premise of them is that there’s music out there in the world that you might like but that you don’t know about yet. However, the assumption is that you won’t like everything. So, you might like anything, but you won’t like everything. In order to figure that out, they’re going to have to profile you. So, simultaneously, we want to help you change what you are, but we’re going to do that on the basis of making a model of what you are now.”
This might sound like a hurdle that precludes taste expansion, but it is impossible to get around. For instance, I could reasonably recommend you – the reader of this article – an album you possibly haven’t heard, based on the fact that you’re on Far Out, interested in this subject, and have reached this point of the feature but I’d be fucked if I had to recommend you what to have for dinner. So, in order for Spotify to offer you something new, it needs to have a pool of old data to make assumptions.
“The criticism is that these systems are fundamentally conservative because they are based on data that has existed in the past, so they are always going to be repeating the past to you. But it’s not obvious where the new stuff comes from,” Seaver explains. “In theory, we are technically conservative as humans—when we’re looking for new stuff, we’re not usually looking for something that isn’t like anything we’ve ever seen before.”
So, in truth, after a thousand words and a lot of head-scratching. The sorry conclusion is that the question, ‘Is your music taste more or less your own in the age of algorithms?’ is neither nothing nor something. It is underpinned by what dictates our taste in general. The psychologist Dr Concetta Tomaino tells me that on that front, “How we take in the world around us is really dependent on who we are, what our previous experiences have been, and what we value. So, it’s really a combination of both. It’s what you’re exposed to and how you are affected by it. There is a give and take when it comes to influences and whether we reject them or accept them.”
So, after all of that, maybe our tastes are just the same, and any crisis we’re experiencing over music algorithms is simply cyber fretting. After all, is Discover Weekly different from our chosen radio station or culture publication’s playlist and content? Thus, the bigger question is perhaps whether algorithms are trustworthy and independent.
