This is really cool. Imagine a map of this across a city, being able to see what different areas tend to listen to. I imagine you'd find some surprising and not-so-surprising things.
Then you'd get profiling to potentially pick out who in particular moved across the city and the exact time of path of their movement.
While this is a nice idea on a local scale, when scaled up it has horrendous privacy implications.
There was a site posted to this place a year or so ago, which looked at work frequencies to find alt-accounts.
I don't hide the fact that I use a different account on different computers, so I have a personal account and work accounts and end up changing accounts each time I change jobs.
This site correlated all my accounts, using a very basic fingerprinting technique of looking for words which a user uses uncommonly often.
It found them all with a good degree of confidence.
I haven't seen reference to that site since, I suspect it got taken down.
Musical fingerprinting would be accurate to a similar degree. You wouldn't look for the music someone listens to most, you'd look for uncommon combinations.
A combination a just a few songs that someone listens to unusually more than other people is probably enough for a good enough correlation for fingerprinting.
Although it's more difficult, it's also possible to be too "middle of the road": very few individuals are very close to the population average in all dimensions.
(Heinrich Böll's At the Bridge is a great short story; Böll had worked in a statistics department so he was probably well aware of the weakness in his protagonist's reasoning)
About the best I'd ask for is that custodes should ipsos be as correlatable as we all are: the amphiopticon?
Lagniappe: https://www.youtube.com/watch?v=HQ7skMnxly0
I'm in two minds about the fact it's down.
1. It's probably a good thing that it isn't super-easy to quickly find everyone's alternate accounts.
2. The capability is clearly there and the technology is out there, but now in the hands of the few people who bother to re-implement it.
It was a useful tool for highlighting the naivety of believing that throwaway accounts were a real possibility when stylometry analysis is so relatively cheap to do.
I had never used it, and wanted to leave a comment on a site (long ago -can't remember where or when).
I started to sign up for Disqus, and it helpfully asked me "We found all these comments from around the Web. Should we associate these with this account?"
It included some old, dead-and-gone-I-would-have-sworn-it troll postings that I had pooped out, back in the last century.
I immediately deleted my signup, and went and had a lie-down.
These days, I deliberately make it obvious who I am, and post as if I had to stand behind my words.
Only if someone can move across the city in three minutes.
Apple and Google could do this if you use their music services, but they already know where you are.
I suppose if I have very unique taste in music and someone else knew about it, they could track me, but this is easily foiled by wearing headphones.