Yes, I believe a primary use case would be new music discovery. It boils down to, if my music profile matches closely to that of another user, and they interacted with 9/10 same songs as me, it's a good bet ill like that 1 other song on the list. Even stronger if that user rates it highly. Now extrapolate this across a network of interconnected users.

There's a couple ways this can be implemented.

The key in any scenario is to generate individual profiles by npub. You could then use the tagging system for genres/styles, artists, songs, etc. For each one, points are scored for positive interactions (boosts, zaps, likes, etc) and thus builds individual scores, per user, per category.

If this system is build by a third party service, any app can API into this and utilize the same scoring engine. Multiple engines would decentralized the process as each could have a different weighting model. Apps could allow users to select different models.