For a while I measured my music-listening diversity using three aspects: diversity of artists, albums, and tracks. For a few reasons, I stopped keeping track of album diversity. So for most of the last year I've been measuring artist and track diversity, both by amount of time (per artist, e.g.) and number of plays. Recently I decided that it makes the most sense to only consider artist diversity by amount of time, and track diversity by number of plays.
For example, if I hear 1 hour of Miles Davis, and to balance it out I want to hear some Bach, then it seems most balanced to hear 1 hour of Bach. That's true even if that means hearing 6 Miles Davis tracks and 15 Bach tracks.
But when it comes to specific tracks, what seems balanced is to hear all the tracks the same number of times.
So, I measure diversity these ways, over multiple periods of time, and then calculate two things: for each artist, how will it affect artist-diversity-by-time if I hear (say) 5 more minutes of this artist? and for each track, how will it affect track-diversity-by-number-of-plays to hear this track once? Whatever tracks result in the greatest improvement of both track diversity and artist diversity are what I listen to next.
