If it’s your first paper, co-authors can be extremely useful to help guide the author through the publication process. It’s also helpful for marketing purposes. Hell, this is basically the graduate school route.
It seems like having all those arbitrary requirements is good for nothing but gatekeeping higher education. Why not have those requirements publicly known and not unnecessarily complicated so that everyone can submit info?
The problem with amateurs publishing papers is emphatically not that the requirements are secretive and arbitrary. The problem is that virtually no one is able to write a publishable paper without basically going to graduate school where you have years to become an expert in the minutia of your chosen field of study. There are a lot of amateur researchers who have the ability to contribute to cutting edge research, but the vast majority of them lack the background to know exactly where the cutting edge lives and how to design, conduct, and then describe a convincing experiment proving to others that the idea stands up to scrutiny. And those aren't things that anyone can or does secret away so they can form some artificial gatekeeper. They're just things that usually require a lot of concentrated effort to study and learn.
They are publicly known. But I would say most paper requirements are "necessarily complicated", simply because science is hard.
Writing a scientific paper requires both solid science (both good results and good scientific practices) and solid writing skills to convey an important idea in a clear and precise way. Mastering both aspects requires time and effort.
An academic author is a person who made a job out of it. Most people don't like doing that, and that's fair, but I don't think it's fair to say that "they are gatekeeping education" when that same uninterested people suddenly find out that the pro level is hard.
(Disclaimer: I'm working in academia, although not in AI)
>Writing a scientific paper requires both solid science (both good results and good scientific practices) and solid writing skills to convey an important idea in a clear and precise way.
The only person who would read a journal paper and say that the the author has good writing skills is another academic.
No one outside of the research community would look at a typical journal paper and claim that the writing skills demonstrated are "solid".
Asimov once wrote about his own dilemma when he was ready to write up his PhD thesis: He had spent over a decade up to that point honing his writing skills, that he truly did not know if he had it in him to write as poorly as is needed for academia.
From my experience in academia:
1. Far too many people still prefer the passive voice over the active voice.
2. Far too many prefer to refer to themselves in the 3rd person.
3. Trying to explain how I arrived at an expression that took me weeks to derive was discounted and I was told to remove it by my advisor: As long as I wrote the starting point "any competent researcher should be able to derive it" (hint, probably half cannot).
4. Writing any background so that someone who is not already an expert can understand was strictly forbidden - always guaranteed to get a complaint by some reviewer or another. If it appears in a textbook or in another paper, do not think about including it in your paper. As a result, the only people who can understand your paper fully are those who have happened to read the same textbooks and papers you have. A poor new graduate student may need to look through several papers and a book or two to get the background needed for one section of your paper, even though you could have explained it all in a page or two. But nope - they may have to go scan over a hundred pages of material to have an idea.
Just a few I remember off the top of my head. Results may vary with discipline.
> The only person who would read a journal paper and say that the author has good writing skills is another academic.
You lack an understanding of the purpose and target audience of journal papers. Papers are written by experts, for experts, to be the most concise presentation of new, field-advancing facts as possible. Prior knowledge of basic, and even intermediate-level knowledge has to be assumed, otherwise it places too much burden on the authors.
You sound like a grad student who's sick of reading papers. I'm sorry, that part sucks. But eventually you don't have to anymore, and then you'll understand.
I understand that fully. You are merely restating the original complaint, which was "gatekeeping higher education"
Essentially, it's: "We write only for insiders".
And it's not for any expert, but an expert in a subfield of a subfield. Someone who is merely an expert in the field will understand it broadly, but often not well enough to reproduce, and often not well enough to even gauge the legitimacy of the techniques.
This becomes quite clear when you see some of the inane stuff reviewers write, which often indicates they did not understand your paper - yet they were picked as experts who were asked to review.
>otherwise it places too much burden on the authors.
There are multiple reasons I do not believe this is the reason:
I can understand a lot of researchers not wanting to bother, but the reality goes deeper than that. They actively do not want others to put in more explanations. If an author wants to put in the time, why are they getting in the way? It's not unusual for a reviewer to ask to excise material that is explanatory.
And frankly, in many research teams in universities, we have grad students who are just starting out and are not to the point of being productive yet. It is very beneficial to have them write the more intermediate stuff. It's not at all time wasted for them.
I'm not saying we need to include standard text book material in all papers. Often an additional 1-5 pages (depends on the scope of the work) will suffice. Any researcher who complains about writing an extra page or two for some project that they worked months on cannot say they care about propagating knowledge with a straight face. The additional time it will take you to write those few pages is vastly offset by the savings everyone (including experts) who reads the paper. That equation that took me days to derive will likely take most experts days as well. Whereas a few pages of derivation would save them all the effort (and would help a reviewer find errors).
I've heard mathematicians be proud if they've read N pages a day (where N is in {1,2,3}) of a typical paper in their field. Yet when I've asked, they've all admitted that had the author put in more effort, that N would be much larger.
They're written for experts, as you say, but even experts have trouble reading them. However, since they mostly only deal with experts, their baseline is very low compared to what the rest of the world would consider "readable". If I wrote a report in my industry that would take some expert a day to read and understand 6 pages, I would be in trouble.
Again, this may vary from discipline to discipline. I definitely have read papers that don't suffer from the above. Ultimately it's a cultural issue. Some are more welcoming of it, others are not.
(1) and (2) greatly depend on field. A lot of people are discouraged using the active voice in, for example, an academic thesis because it allows the nasty question of "So how much of this did you do?" and writing "I did this" comes off as a bit weird. In journal papers the use of the active voice is slowly improving. I still find it a bit odd that people use we for single author work, but that's probably just ingrained.
To address point (3) a common route is to add an appendix or supplementary material. This is commonly done in deep learning for derivatives of novel functions so that others can implement the backward pass. Whether you can get it past a supervisor is another issue entirely! A blog post or accompanying website is an easy option.
(4) is another tricky one. How many times have you read "Deep learning has revolutionised <field>" followed by the usual citing of Hinton, Le Cun, Krizhevsky et al.? The spiel is identical in every paper, how many ways can you describe this stuff after all? At some point you just have to assume that the reader is familiar with the background, but you still have to cite everyone to keep the reviewers happy.
Again, if you need exposition, write a blog post about it.
(This isn't just for deep learning - every field I've worked in suffers from the same copy-pasta introduction.)
This also depends again on field anyway - some journals are much more approachable than others. Nature and Science, controversy aside, generally publish very readable papers.
To be fair, only a few of these sound like actual problems.
>The only person who would read a journal paper and say that the the author has good writing skills is another academic.
No one outside of the research community would look at a typical journal paper and claim that the writing skills demonstrated are "solid".
Scientific papers are written for people in the research community.
>3. Trying to explain how I arrived at an expression that took me weeks to derive was discounted and I was told to remove it by my advisor: As long as I wrote the starting point "any competent researcher should be able to derive it" (hint, probably half cannot).
This seems very context dependent.
>4. Writing any background so that someone who is not already an expert can understand was strictly forbidden - always guaranteed to get a complaint by some reviewer or another. If it appears in a textbook or in another paper, do not think about including it in your paper. As a result, the only people who can understand your paper fully are those who have happened to read the same textbooks and papers you have. A poor new graduate student may need to look through several papers and a book or two to get the background needed for one section of your paper, even though you could have explained it all in a page or two. But nope - they may have to go scan over a hundred pages of material to have an idea.
The poor new grad student should pour over multiple papers to get a solid foundation of the field. Most researchers do not want to read an epic tome to figure out how your research is novel.
I agree with your first two writing points though.
The requirements are generally clearly communicated. The strictness is about publishing quality science. Someone who has not been through the peer review process before might be quite shocked by the stringent requirements to get a paper published. Again, it is NOT arbitrary; it is about making sure the journal is publishing real science and not quackery. It is entirely possible for someone outside of academia to publish, but having guidance to get through the peer review process is invaluable.
Normative publication behavior varies by field, but is generally well known and openly discussed, critiqued, modified, etc. inside each field. It's not exactly kept quiet.
the 'requirements' that are being referred to go well beyond objective rules (e.g., no more than 4 journal formatted pages) and include things like lingo, citation standards, ordering of sections, and general academic writing voice.
Written communication of science is hard, especially when you are new to it. I think a good 50% of what I have learned as a PhD student was how to more effectively communicate through writing. And that builds on me (honestly) being in the top quartile of writers in my program when I walked in the door. I have worked with many co-authors who have had great work rejected by journals for writing reasons. It isn't about scholarly snobbery - a lot of it just isn't understandable to anyone but the author. Writing advertising copy / business plans / anything related to startups I worked with was orders of magnitude easier.
In a strict sense its gate keeping, in a more realistic sense its about norming. These standards/expectations often develop organically and are sometimes hard to write down in the sense you are asking for.
In my own experience, I once wrote a paper in a much more casual and approachable style. I had what I felt was the same scientific rigor around the experiments and data, but tried to make the text more generally readable. I had Simpson's references sprinkled here and there, that sort of thing. The response was generally negative. The paper was accepted for the mid-tier conference I submitted it to, but on the condition that I revise to take care of reviewer concerns, many of which boiled down to "you should write with a more appropriate tone". No one complained that the paper was imprecise as a result. They just didn't like that it didn't read in the standard passive and bland academic voice. I found that somewhat depressing.
> I had Simpson's references sprinkled here and there, that sort of thing.
I should offer to read your paper before criticising it, but I can't even begin to understand why you would think that references to a TV programme would help make a paper more readable. What if you don't happen to watch the Simpsons? Is the paper then even more opaque because none of the references mean anything?
I quite like the bland academic voice - because it lets me focus on what I came for which is the science. I don't want to be amused, and I don't want academics wasting time trying to be witty or fit in references to whatever TV programme is fashionable.
I want just the science, please! Delivered as clearly and simply as possible.
There are parts of the standard academic paper that aren't really necessary for understanding it. You typically have background on the problem, for example. I can't say for certain I didn't lose a reader, but none of the references detracted from meaning in my view.
I totally TOTALLY agree. I was trying to stay away from offering value judgement and be a little more descriptive. it is not necessarily effective, but, it is.
I have done similar things, with similar reactions. I think such approaches in fact negatively impact science...especially through the public's engagement with it. Perhaps my 'favorite' was trying to do a conference session about active learning using...active learning. Got told that it was unprofessional and that those listening were 'peers not students'.
There was a really wonderful paper on academia I read not long ago[1] that basically made the argument that a darwinian biology framework basically explained all this. It basically asked the question what is a defense that arose to solve a problem, and has now transformed into what looks like a defect...and how do we separate those (or should we) from actual defects. Today I stumbled across a similar article which discussed
[1] Lohmann, S. (2004). Darwinian medicine for the university. Governing Academia: Who Is in Charge at the Modern University, 71-90.
[2] ...and of course I can't re-find it. I'll update if I do.
Overlooking the social or communal aspect of research is a mistake. People want to publish their research for a number of reasons (prestige, priority, profits, simple enthusiasm), and people want to read published research for other reasons (to further their own research, to understand the current state of the art, simple curiosity).
Publishing a paper no one reads is pointless. To make people want to read a paper, you need to make sure it is novel (so they don't waste their time reading about something you only rediscovered) and that it is integrated into the existing understanding of the field, with citations of related research, using common terminology, etc (so they can more easily understand your work and what it means).
This is what working with an existing researcher gets you. If you don't work within this framework, why would other researchers waste their time working out that when you say "frazzles" you mean "frizzles", and once you work that out the whole thing is equivalent to a conference paper from '06?
Tottaly. While ACM, IEEE, and similar organizations provide some of this information and do substantial outreach, more progress on this type of work would benefit everyone.
