[pimlottc]

The classic terminology for this in AI/ML is "explore vs exploit", i.e. striking a balance between trying new things (in hopes of finding a new favorite) vs going back to the tried-and-true.