Upenn professor introduces his mentorship philosophy here with the following my version of summary:

  1. Having a variety of qualitatively different projects is helpful for student-centered mentoring, not project-centered.
  2. Spending time on problem formulation (model justification, relationship to existing literature, changing assumptions, or connections to more distant fields) is important as much as a novel solution.
  3. Create an intellectual environment where students can teach for themselves and from other peers.

2 is what I learn a lot from Garud whenever he asks me what the overall problem that I am solving is.