32 ms·
Why are you so surprised that a metric like GPA (or undergrad accomplishment in general), with a hard ceiling, loses its correlation with actual talent at the t
by eli_gottlieb 13y ago
Why are you so surprised that a metric like GPA (or undergrad accomplishment in general), with a hard ceiling, loses its correlation with actual talent at the top end of the scale? That is, most undergrads will simply never get the opportunity to publish one paper in undergrad, let alone two or three, so if you admit them to grad-school based on maybe a publication, good grades, and maybe some internship/REU experience, you should expect to admit some highly talented students and some students who were very talented at optimizing the capped metrics on their resume to the maximum but not at actual research.
Hell, in a way, it's basic reinforcement learning, is it not? What sort of rational agent or system will "go above and beyond" when the reward signal is capped at a maximum value? When you look at it that way, every single step of the selection, sorting, and filtering process we call academic meritocracy is optimizing for irrational devotion to careerist signaling, which has only a weak correlation to real devotion to science.
- hga 13y agoWell, one thing you can do is admit everyone who a colleague you know and trust says can do research (what I've heard, quite logically, is the #1 way to get into a good graduate program, after of course convincing a professor in one to demand your admittance). So the "schmucks" ... and I've known of one MIT professor who didn't get tenure who was in that category (a friend tried to do research in her laboratory and found it impossible, it was so badly run, something subsequent events proved) ... are either false positives from the above metric or from the cohort of department's best guesses after running out of them.
- eli_gottlieb 13y ago>Well, one thing you can do is admit everyone who a colleague you know and trust says can do research I always thought this was simply how you get into grad-school: do research in undergrad, and have your research mentors write you letters of recommendation, trusting their reputation to carry information the admissions committee will understand and care about where grades and awards won't. >So the "schmucks" ... and I've known of one MIT professor who didn't get tenure who was in that category (a friend tried to do research in her laboratory and found it impossible, it was so badly run, something subsequent events proved) ... are either false positives from the above metric or from the cohort of department's best guesses after running out of them. Kinda. Politics become a problem, but it is true that once you actually get to the point of publishing papers, the cap on your achievements disappears, so you do, in some way, get career credit for doing more.