Using clones of my last.fm library, three virtual agents with divergent listening habits
explore their recommended artists. Weekly playlists of the music they have discovered are created
in my Deezer account. Hopefully, over time they will discover music that, given my own listening habits,
I would not have found.
Below are links to the Deezer playlists and the last.fm tags used to define the listening habits
for each of the possibles. They are unlikely to be perfect, but that is part of the point.
experimental || avant-garde
jazz || nu jazz || neo soul || soul
minimal || downtempo || ambient
The agents operate using some extremely basic huristics based on last.fm tags. Every week, they survey their respective last.fm recommendations for the tags
they have been assigned. Whatever they find is scrobbled back into their account - specifically, they grab that artist's top tracks from Deezer and scrobble those - thus generating more recommendations.
The scrobbled tracks are also dumped into a Deezer playlist, which I can then listen to.
Starting with a clone of my last.fm account ensures that I don't get too many redundant recommendations.
Over time they should move further and further away from this baseline, generating recommendations that I would not have given my own
Whether this approach, based on generating a path between an existing music taste and more exotic recomendations, is useful remains to be seen.
It is hoped that by branching out from an existing music taste - as opposed to exploring a new genre by focusing its top artists/anthems - will generate more unique recomendations.
On a good day, I would say that the agents are exploring alternate possible listening futures - hence the name rdm_possible (
my initials are RDM). What is more likely is that they are injecting
some healthy randomness into my music exploration.