data

Rearrange Us: Finding What I Lost With Apple Music

One of the ways that I like to remind myself that I'm living in the future is to think about what technology lets me do today that would have completely blown my mind in high school or college. I've always been really into discovering music, and when I was on a high school budget, this basically meant listening to the weird indie rock radio station that played Counting Crows six months before anyone else, instead of listening to the same 20 songs over and over again on Z100. The idea of being able to just decide that I want to listen to a particular album, especially if I didn't already own it, would have been mind blowing to me at the time. As it was, there was a local record store chain called Compact Disc World that would let you both listen to any CD they had in stock before you bought it, and return it if you didn't like it, which was unheard of in the late 90s. I ended up buying a lot of albums from them with my supermarket cashier income, and I only returned one or two over the years they existed, for fear that they would put me on a list if I were to try to take advantage of their way too generous return policy. (There were stories, you see.)

So it's no exaggeration when I say that, when I got my first iPod in 2005, it was like the skies opened and the sun started shining down. All of a sudden I had access to all of these albums I'd collected in one place. What's more, I had more than a few albums that I had bought for That One Song, and now I could listen to that song without having to find the CD and seek to that track. Why, I might never have to listen to the radio again! iTunes in particular appealed to the cross-section of my tendencies as a music nerd and a data person. What I realized, after poking around with Smart Playlists for a couple of days, was that iTunes was basically just a database that held my music, and I could query it to tune it exactly the way that I wanted. This ended up being one of my first "big data" projects; the 15,000 tracks I ultimately had in my library might not be big data in the truest sense, the data was big to me, especially at the time.

It took me several iterations, but eventually I ended up with an extremely elaborate system of smart playlists that worked off of metadata like last played date, date added, genre, and even hand keyed tags that I entered into the comments field of the tracks. I underwent a several month project where I rated every single song in my library on a five star scale to facilitate those smart playlists. (There was one dark day where I started researching hacks to enable half star ratings, but cooler heads ultimately prevailed.) I had all of my and my wife's music in the same library, but we never had to listen to each other's music if we didn't want to. I even had a smart playlist to let me re-rank anything that I hadn't listened to in a while, to make sure that I still liked them as much as I had originally. (If this sounds borderline obsessive, by the way, you're not wrong; this is both the positive and the negative of hyperfocus.)

Ah, memories.

Ah, memories.

The only problem I eventually ran into with this system was storage space on my devices, especially once it stopped being reasonable to carry an iPod alongside my phone. My music library, even if I just limited to the smart playlist that had only the cream of the crop, was just too big to fit on an iPhone. Eventually, I caved: I signed up for iTunes Match to alleviate the problem by streaming most of my music instead of storing it locally. That resolved the issue with storage space, but it ultimately created other problems with my system, because iTunes Match was generally unreliable at updating metadata like last played date. For most people, this would be a minor inconvenience at most, but for me, this ended up making my smart playlists less consistent, which, in turn, made them less useful. The smart playlists still worked, but ultimately I ended up using them less and less as time went on.

All this background is to say that I care about my music library. A lot. So it was kind of a big deal that, when Apple Music first came out, I decided that I would stop worrying and learn to love the algorithm. What I'd hoped was that Apple was going to use all of the listening data that they had been collecting from me for nearly a decade (between Genius and iTunes Match) and be able to both surface music that I would want to listen to from my library, as well as find me new artists I might not have found otherwise. That didn't happen. Apple Music asked me if I like to listen to a bunch of artists that I probably have never listened to intentionally in my life, and then I was on my own, and the recommendations were generally irrelevant to me. For You improved somewhat with iOS 10, but though the first few days seemed like an improvement, it eventually began to cycle through the same set of album recommendations, no matter how much I tried to use the love and dislike buttons to correct it. Though their playlists are generally well curated, I found that there were only a handful that ever really appealed to me, and they didn't seem to get updated over time. Even the nature of the recommendations could seem suspect at times; while I'm sure Chance the Rapper is a wonderful artist, it doesn't make sense to me why I would want to listen to him since I'd been listening to The Shins.

Even though I knew that For You wasn't doing a great job, it worked well enough that I'd learned to live with it. It wasn't until MacStories wrote a review of a new app called Picky, an alternative Apple Music browser, that I'd started thinking about this again. What Picky does is very simple: It provides you all of your albums, artists, songs, etc. in a long scrolling list; what makes it different is that you can filter that list by how many songs are in each list item. I'd basically given up on using Apple Music's browse function to find something to listen to, since, unlike my smart playlists, it had every artist who appeared in my library even once, regardless of whether they were one of my all time faves or happened to be the last song on a random sampler CD I'd gotten years ago. Using Picky, on the other hand, immediately led me to find Mates of State again; Rearrange Us is one of my favorite albums, and I realized that I hadn't listened to it in at least a year, simply because I just couldn't find it in my Apple Music library by browsing, and For You's algorithm never surfaced it to recommend it. This led to a negative feedback loop; since I hadn't listened to Mates of State, Apple Music didn't know that I wanted to listen to them, so it didn't suggest them, so I didn't listen to them. All of a sudden, Mates of State was gone from my listening rotation. Presumably, had Apple Music pulled in my listening history from before the service launched, this might not have happened, though without knowing how the algorithm finds its suggestions, this could have happened anyway, or with a different artist.

Screenshot 2017-02-19 19.37.38.jpeg

That's ultimately what ended up concerning me after I had this realization. This isn't going to make me switch to Spotify or cancel my Apple Music subscription, to be clear. There are reasons beyond just music discovery that keep me with Apple Music as my primary streaming service; the convenience of being able to use the built-in music player outweighs whatever bad music discovery bugs there are, and I'd much rather pay a flat monthly fee than go back to agonizing over every album I might want to add to my library. What this does, however, is underscore how careful we need to be about letting AI into every aspect of our day to day lives. This is a case where I knew that the For You algorithm wasn't great, I trusted it anyway, and I ended up completely forgetting about an artist who I really enjoy listening to, simply because it wasn't offered to me. This episode isn't affecting my life materially, but it's very easy to look at recommendations in all sorts of areas of our lives (news, perhaps?) and wonder what we should be seeing that's not being presented. This isn't even to suggest that there's an ulterior motive here. I'm pretty sure Tim Cook isn't trying to keep me from listening to Mates of State, but it's just easy to lose track of things if we're relying on smart assistants and algorithms to offer them to us.

There's a happy ending to this story, though. I haven't listened to Mates of State in so long that they released a new album that I've never gotten to listen to, so it's new to me. Maybe if I binge listen to their back catalog enough, Apple Music will get the hint. Here's hoping.