Hacker News Reader

Would you rather be told you're not going to get a stable job as a researcher when you're 21 and first applying to grad school or at 28-35 when you're finally told you didn't get any interviews for a tenure track job?

Exactly. It's far better to get rejected early from the PhD track, for not being a high achiever, than to muddle through unable to secure a job. You would still be able to prove yourself in a masters degree.

Kinda, some people I know who did poorly at undergraduate/masters did a lot better at the PhD level though and vice versa as research is very different to being taught.

There is such a lack of permanent research positions compared to postdocs and PhD students that I'm not even sure it's selecting for the best people.

Society needs way more than just a few "high achievers" ie people willing to work themselves to the death for vague rewards, doing science.

right, this is the clear improvement. on the other hand, now it means science-oriented high schoolers lives get an extra helping of misery and stress.

on top of this you also filter out anyone who doesn't perfectly stay on this high-stakes pipeline from age 14 to tenure, which is probably not good overall for science.

economics is actually pretty close to working this way already. indeed, the job market for economics PhDs is much less dystopian than say life sciences. but there are tradeoffs.

A different possibility is that we stop trying to find "the best" candidates (which, by the way we're measuring with metrics that have been Goodharted well beyond relevance). What if this (and maybe everything else) worked via a kind of lottery:

1. The institution in question evaluates candidates as either a fit or not a fit. 2. The candidates who fit are the pool from which the positions are filled at random.

Obviously it's gameable as stated, but I think if I spent a day or two designing a better mechanism with the same premise, we'd end up with great candidates but without the hyperoptimization trap.

> there’s increasingly an expectation that students should come into a PhD program with one or more publications already under their belt. For top programs ideally these should be first author (not yet an absolute requirement but it’s more important each year).

This is something I discovered pretty late in my life. The web pages maintained by profs barely have any info.

I have come to hate the whole "everybody looks for different things" type of answer when it comes to PhD apps. I wish profs would just spell it out on their site in concrete terms what they personally look for.

at least in computer science, this document is from CMU and tended to get circulated pretty widely, although probably not widely enough. every field ought to have something like it.

https://www.cs.cmu.edu/~harchol/gradschooltalk.pdf

I have seen that, and it was helpful. But at least in machine learning and AI, if you want an admit into any top 10 US PhD program, you need a publication or 2 in a top tier venue, preferably first author.

In other fields and lower ranked programs, again I wish profs would be explicit (since publications might not be expected).

Like this guy:http://math.stanford.edu/~vakil/potentialstudents.html

"So before I take you on as a new student, you should be comfortable with the foundations of the subject, which means having done the majority of the exercises in Hartshorne or my course notes, and being able to explain them on demand."

That is very explicit and sets clear expectations.

These are regular market forces. More people fighting for a few slots. If people weren’t so desperate to be professors this wouldn’t happen. Pretty sure actors and models go through something parallel.