“In Spotify’s view…when users search for and listen to music they are providing a measurable set of inputs from which musical tastes and desires can be extrapolated. This user behavior is recorded, compared and evaluated against that of other users, then sorted into metadata and used to calculate every song and artist’s degree of relevance to each individual.”
– Thomas Hodgson, “Spotify and the Democratization of Music”,
Popular Music 2020, Vol 40/1., p. 7.
“A difference which makes a difference is an idea.
It is a ‘bit,’ a unit of information.” – Gregory Bateson
On Spotify, music recommendations are flying my way daily—a bombardment of you might enjoy this! based on the company’s data of since you and others with similar listening habits already listened to this. Usually though, the algorithm is off. I like to think this is so because I don’t listen guided by style per se, but by Quality, which comprises a more difficult to quantify–in other words, subjective–set of attributes. From standard playlist prods such as “The State Of Music Today” and “Discover Something New” to recommendations narrowly targeted to what I apparently listen to (and therefore am, musically speaking), such as “Neo-Classical” and “Atmospheric Piano” and “Experimental Electronica”, on Spotify my tastes are reflected back to me and this reflection is refined in real time as I click to listen to this track, but not that one. As I listen, the algorithm refines itself to better reflect my tastes now, and maybe anticipate them in the future.
•
What’s the end goal? For Spotify to perfectly predict its users’ listening interests and habits to better keep them inside this streaming universe, paying by the month? Admittedly, sometimes it is nice to be figured out, even by software. I appreciate it when every year or two I get notice of say, a new Autechre recording. (Yes I will listen and yes I will probably find some beautiful moments therein. Speaking of which, have you heard “32a-reflected”?)
But Spotify also generates vast numbers of playlists on which musicians of varying levels of Quality get lumped together via a presumed shared style. This is the reason why what I hear as the cloying piano music of Ludovico Einaudi sits alongside the non-cloying but more meditative/introspective piano of Nils Frahm. By virtue of their shared sonic surfaces and general style (neo- or post-classical?), tracks by each composer could be considered related, and maybe both musicians are, in the end, makers of atmospheric piano music that is roughly similar. But their Quality quotients are different, and Quality comes from details, from bits of information. As the cybernetic anthropologist Gregory Bateson once said, information is “any difference that makes a difference.” One could make the case that Einaudi’s and Frahm’s music are different in substantial, if hard to pin down ways, yet a Spotify playlist that groups them together glosses over such differences. Tiny differences make something (or someone) who they are, and such details can render one thing (or person) slightly annoying, and another thing well-balanced. Whether we’re talking about musical style or musical Quality, differences can make all the difference.
•
Which brings me to non-algorithmic music recommendations. The other day I looked up a Frahm piece and noticed that on his Spotify page he had posted a few of his own music recommendations via a playlist of a collection of dub-influenced tracks. I wasn’t expecting this, so rather than searching for Frahm’s music I spot-checked his playlist instead. One track, a 1996 piece by a German artist named Nonplace Urban Field (Berndt Friedmann) caught my ear with its minimalist rhythmic profile. This little musical discovery, I thought, was worth it. It was worth it because the sounds have something, some Quality. Here was music offered up not by an algorithm, but by another musician, as an influence:
brettworks 