How well do user preferences align with music algorithms?

18 Ağu 10:40

As an old record collector, I also come across Spotify's "retro-tinged" recommendations, and I'm always amazed at how consistent they are. I find compilations ranging from 70s progressive rock to 90s grunge tracks—stuff that's close to my collection. Yet, sometimes when it suggests an unexpected synthwave track from the 80s post-punk I listen to day and night, I can't help but think the algorithm has "jumped eras." How closely it aligns with user preferences varies from person to person; while it's about 80% accurate for me, my neighbor's algorithm can randomly fill their queue with metal tracks. So, it both adapts to user tastes and, to some extent, plays an "educational" role by subtly shaping our listening habits—just like a DJ selecting and mixing records—but sometimes they can miss a small "crossfade."

18 Ağu 11:32

To test how accurate algorithms are in their recommendations, I ran an experiment based on my own listening habits. In my opinion, the clearest takeaway is to evaluate the tracks they suggest **not by our expectations, but by our immediate reactions to them**. For example, when I listen to bachata or salsa, the algorithms usually steer me right—they align with the rhythm and harmonies of those genres, which I naturally respond to. But when I tried exploring a new wave of Latin pop, most of the songs they recommended didn’t click with me. What I noticed here was that algorithms focus on **the tone and tempo of what we listen to**, whereas for me, capturing the *feeling* is what really matters. In the long run, I do think these systems change our habits, though not always for the better. Lately, after repeatedly listening to songs like "Danza Kuduro," the algorithm started pushing similar tracks my way—some of which are among my favorite Latin tropical tunes, but after a while, they all started tasting the same. **My solution?** Use algorithms to my advantage—meaning I only listen to about **70% of their recommendations** and pick the remaining 30% myself. This way, I still get to explore new music without falling into a rut. Of course, this isn’t for everyone, but it’s worth remembering not to let algorithms dictate our musical tastes entirely!