Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The culture of rejection in computer science publications

sigbed.org

Post reply on HN
Print view View original

141–150 of 172 posts

Re: The culture of rejection in computer science publications

#141

Post by blacksmithgu »

grayclhn wrote: post #90
FabHK wrote: post #86

One side effect I didn't see mentioned in the article: One professor of mine spoke of the LPU, the least publishable unit. So, if you're lucky enough to have some novel ideas, and build something nice out of it, don't put it all into one coherent and easily digested journal paper! The number of publications counts. Instead, chop it into little pieces that are just "novel" or noteworthy enough (LPUs), and publish them…

Honestly, I think this is often said cynically but is a good practice overall. Would you rather have to read and understand one giant commit reflecting 2 years of work, or 10 well-documented and logically complete individual commits?

The issue is that novel paper ideas will be split across multiple years (and even multiple conferences), making it much harder to actually see the whole picture for a reader. Each little piece of the paper will often also be bloated with unnecessary extra detail in order to reach the threshold for "minimum publishable paper".

Re: The culture of rejection in computer science publications

#142

Post by mizzao »

LunaSea wrote: post #16
mizzao wrote: post #8

Earlier quoted context omitted.

Basically already happening in many Chinese departments and conferences, just without the AI yet, but of equivalent quality. (I'm ethnic Chinese and this is not meant to be a racially charged comment)

Could you give a bit more context about this? I'm not familiar with the subject.

I think the performative, results-based culture of academia in general (particularly in Asia and not just Chinese) has seriously succumbed to Goodhart's law: https://en.wikipedia.org/wiki/Goodhart%27s_law

Because you are judged on citations, publication count in prestigious journals, and research dollars spent... you have people citing themselves and their friends in citation rings, gaming the referee process to pick reviewers, and doing unethical things for money.

My first experience in this is when I was asked (as a postdoc) to give a keynote for a no-name conference in Vancouver. They were paying so why not? I later realized that my pedigree was being used to lend credence to a conference that was totally cargo cult academia. There were conference sessions, and people presenting, and bunches of people in several rooms, but nothing was actually being said and nothing was being done. They then offered me $3,000 cash in an envelope which I declined to take most of because of ethical reasons (I only wanted to cover my expenses). A very unique, maybe even out-of-body experience to stay the least.

If you want to see a crazy instantiation of cargo cult academia, google "extreme learning machines", which is basically a whole field built off of 2-layer feedforward neural networks with 1 randomly initialized layer. The other keynote at this conference was the guy who "started" this field.

Re: The culture of rejection in computer science publications

#143

Post by yodsanklai »

I don't see a "culture of rejection". There's rejection whenever there's competition and a selection. There are tons of computer science conferences and not all of them are elitist. Even mediocre papers get published. Should we have all the submissions in the world accepted at the top conferences so that no researcher feels left behind?

Re: The culture of rejection in computer science publications

#144

Post by blacksmithgu »

bo1024 wrote: post #61
magpi3 wrote: post #57

Academia as a whole is such an interesting place. I value it highly because I value learning. I love school, I loved the conversations and deep dives into interesting waters that nourished me in both college and graduate school, and I loved the opportunity to be fully immersed, without needing to get a job, in an atmosphere of learning. But the rest of it? The petty hierarchies, the papers born of countless hours of…

If the last paragraph is your perspective on being a pawn in an academic structure, I would love to hear what you think of being a pawn in a corporation.

Not OP, but at at least most corporate pawns are cynically aware of how artificial the system is and that you need to play games to get ahead (or just coast along if you don't care). Academia has a luster of meritocracy when in reality you need to game things just as hard to become successful.

Re: The culture of rejection in computer science publications

#145

Post by kleiba »

blagie wrote: post #62
ath0 wrote: post #32

I'm long past my academia phase, but recently led the PC for an industry conference (accept rate: ~15%). 1. Curation is important both for the physical limits (venues only fit a certain number of people), attention limits (attendees will usually retain only a handful of "nuggets" no matter how packed the agenda is) and interaction limits (you can't meet everyone at a large conference). 2. If the goal of a conference…

I'm well-known in a research community. I'm positioned such that I don't need more academic points. I've mostly stopped publishing in branded prestige academic venues, in part due to rejection rates. My goal in doing work and writing papers is to see them disseminated. The acceptance/rejection process is asinine -- studies show it's basically random. I've had one paper in my whole career where the reviewers did a pro…

studies show it's basically random

The "basically" is important though, because there are some nuances to it.

However, the point I've actually come here to make is that since publications are a strong factor for your career progress in academia, a corollary of the above is that making it in academia is basically random, too. Which is also true for other reasons, though: for every open professor position in a certain field, there are usually a number of candidates that are all equally highly qualified. But only one of them can get the gig. If the selection is not random, then it's typically based on other factors, such as, how well you are connected, your gender, whether some other professor at the faculty fears competition from you, etc. -- which may not be random, but is equally out of your control in all but a few cases.

Re: The culture of rejection in computer science publications

#146

Post by magpi3 »

bo1024 wrote: post #61
magpi3 wrote: post #57

Academia as a whole is such an interesting place. I value it highly because I value learning. I love school, I loved the conversations and deep dives into interesting waters that nourished me in both college and graduate school, and I loved the opportunity to be fully immersed, without needing to get a job, in an atmosphere of learning. But the rest of it? The petty hierarchies, the papers born of countless hours of…

If the last paragraph is your perspective on being a pawn in an academic structure, I would love to hear what you think of being a pawn in a corporation.

I suppose a big enough corporation can feel the same, at least in terms of internal politics, but the difference for me is the tenure (well-paid, almost impossible to fire) and adjunct (close to minimum wage, no health insurance, completely disposable) dynamics. Indefensible. And I think the people with tenure don't object to this model simply because they enjoy their privilege.

Re: The culture of rejection in computer science publications

#147

Post by holidaygoose »

blacksmithgu wrote: post #141
grayclhn wrote: post #90

Earlier quoted context omitted.

Honestly, I think this is often said cynically but is a good practice overall. Would you rather have to read and understand one giant commit reflecting 2 years of work, or 10 well-documented and logically complete individual commits?

The issue is that novel paper ideas will be split across multiple years (and even multiple conferences), making it much harder to actually see the whole picture for a reader. Each little piece of the paper will often also be bloated with unnecessary extra detail in order to reach the threshold for "minimum publishable paper".

Using the same code analogy as the parent, this is like code with unnecessary extra commenting. Seems okay to me.

Re: The culture of rejection in computer science publications

#148

Post by gautamdivgi »

BeetleB wrote: post #89
blagie wrote: post #62

Earlier quoted context omitted.

I'm well-known in a research community. I'm positioned such that I don't need more academic points. I've mostly stopped publishing in branded prestige academic venues, in part due to rejection rates. My goal in doing work and writing papers is to see them disseminated. The acceptance/rejection process is asinine -- studies show it's basically random. I've had one paper in my whole career where the reviewers did a pro…

> Comments often show the reviewers never read the paper. This. I was not in computer science, but in a different technical field, and this is sadly common. We would often have to appeal to the editor with "The topic the reviewer said we didn't address? It's in Section X. Get us another reviewer."

I think this happens in all fields. It’s probably a professor on a PC dumping the review on an unsuspecting and overworked PhD student or MS student who really doesn’t care and just wants to get some sleep.

And yes - say what one may - PhD students are overworked and underpaid at least in most of the US.

Re: The culture of rejection in computer science publications

#149

Post by impossiblefork »

TheOtherHobbes wrote: post #71
qzx_pierri wrote: post #19

It doesn't matter what color you are. It doesn't matter where you're from. If you create something like what Dijkstra created back in 56' (undeniable innovation), then no door will remain shut. That's the beauty of Math, CompSci, or STEM as a whole. Not to be too blunt, but I'd rather we keep this soft, "everyone should be accepted" nonsense out of CompSci. The cream will rise to the top. Anyone complaining just need…

The cream isn't rising to the top. I doubt many people on HN - never mind in the STEM community a a whole - can name anyone born after 1960 doing research at the level of Dijkstra, Knuth, Hoare, and maybe Wirth. This either means there's no one that smart or original in the last few generations - possible, but something of a stretch - or the system isn't successfully selecting, motivating, and rewarding people with t…

Cook's theorem was 1971, and I think that's a foundational result.

Then there's Håstad's work from the 1980s-2010s on lower bounds and optimal non-approximability results.

There's Sanjeev Arora, who was one of the guys who worked on the PCP theorem and who has now been writing interesting things about ML.

There's Razborov and Rudich, with natural proofs, which was in 1994.

There's Khot, who would be up with these guys provided that the UGC is true.

Re: The culture of rejection in computer science publications

#150

Post by ghaff »

xdavidliu wrote: post #135
ghaff wrote: post #21

Earlier quoted context omitted.

As someone who has been on the conference committee for (non-academic) conferences, I'm not a particular fan of blind selections. Yes, I get the desire to not just give slots to "the usual suspects" which at least historically has often happened at a lot of conferences I attended. But the reality is that there are people who you know based on overwhelming experience will give great talks that attendees will appreciat…

> there are people who you know based on overwhelming experience will give great talks that attendees will appreciate and learn from ... do you really want to reject them? If I'm reading correctly, you're saying that sometimes a work deserves to get in based (at least partly) on the author's name recognition (for historically having given good talks) and not solely on the merits of the work itself. I see the point yo…

An abstract isn't a talk. It's a limited description of a topic (which is often written pretty quickly by experienced presenters). So, yes, someone who consistently delivers is probably going to do a pretty good job unless the abstract just seems uninteresting. i.e. a good presenter can pick a topic I just don't think attendees would be interested in or, more commonly, just isn't a good fit for the program.

More broadly, yes, you need to balance having people you know will do a solid job whatever their abstract against welcoming new speakers.

Imagine a resume is just so-so. But you've worked with a person and know they're great. Do you judge them based on their resume?

Post reply on HN