the privilege of advice working out: Nonlinear Function
Created: February 23, 2020
Modified: March 06, 2020

the privilege of advice working out

This page is from my personal notes, and has not been specifically reviewed for public consumption. It might be incomplete, wrong, outdated, or stupid. Caveat lector.
  • In every field, there is a store of 'standard' advice that is handed down from mentors to ambitious youngsters. In computer science grad school, this includes maxims like:
    • Ignore your classes, focus on research.
    • Don't choose a grad school, choose an advisor.
    • Don't spend too much effort on teaching.
    • ???
  • Everyone hears versions of this advice from their own mentors. If it is good advice, it is correlated with good outcomes, but it cannot guarantee them. So let's consider the people who arrive at the desired outcomes (in grad school, these might include: good research, an intellectually satisfying job, in academia or elsewhere, mental health), and those who don't.
    • The ones who did well can authentically pass on the advice they received. At some point it becomes not just hearsay, it is a genuine reflection of their own personal experience: following the given advice worked well.
    • But those who didn't do well can't honestly pass on that advice. speech becomes less free. At best they can qualify it: 'this works for most people, but it didn't work for me'.
  • There's an even deeper form of this: the privilege of model-based reasoning working out. When we're young, all we have are models, and we draw conclusions from those models. "Working hard brings rewards". "Tech magnifies human effort and can change the world." Much simpler things: "Giving a name to a group of people helps it persist.". "Spaced repetition is a useful way to learn." Probably a ton I can't think of right now.
  • If we end up succeeding in life, we credit the conclusions we drew and feel justified in sharing our model-based advice with others.
  • But if we don't, we won't. We might have drawn many well-supported, interesting conclusions and developed useful concepts, and just not succeeded due to noise in the system. (unlucky lack of opportunity, lack of mentorship, relationship constraints, family trouble, physical or mental health problems, etc.)
  • This is kind of just life: the only way to validate models is empirically, and empirical feedback on long-term high-level policies is sparse and noisy. But it's also not inherently necessarily: ideally we would be able to 'explain away' our failures by attributing them to particular causes or sets of causes, so that we retain confidence in our other models.