I really don't think it is necessary to bring in deliberate obfuscation as an explanation. The fact is that writing clearly is hard. It takes work. And tenure/promotion/etc. decisions are often made on output, not on clarity.
So, say I'm a researcher. I've written a paper. It's a mess. I can put some more time into it, rewriting it to be more understandable, or I can use that time to do research & writing for another paper. In the former case, I get one clear paper; in the latter, I get two unclear ones. The people on my tenure committee are going to read my publications list, but not the papers themselves. Which scenario makes me look better to them?
My post is somewhat tongue-in-cheek :) The main point is that the incentive is there to wrap uninteresting research in terminology and complexity - so if the incentive is there, you wonder if anyone does it deliberately. I've read a lot of papers which were difficult to grok but ended up being remarkably simple.
Another factor is that reviewers are sometimes inclined to reject papers that are too straightforward: if the final result is so simple, is the work really novel or interesting?
I've seen the same a lot of times, and I wonder if it's how many textbook companies work. I've read a couple wonderful textbooks, but they are definitely the exception rather than the rule.
Typically, the more complicated the language, the less is being said. Ever read legal documents?
Also, having read a few (first year grad student here), I'm starting to think it might sometimes be a simple matter of lack of space. Explaining algorithm and method correctly requires examples, precisions etc. which might take away space to pack in more results and experiments. (I know some conferences have submission guidelines with page limits)
There might be "shadier" reasons, but if that explanation makes sense, I for one would like it a lot if some researchers got in the habit of having longer explanations somewhere, say on their website. I don't know if that'd somehow break implicit rules in the academic field, but it'd lead to less head-banging-in-the-wall-trying-to-understand-what-that-damn-symbol-means.
The top physics journal is Physical Review Letters, which restricts submissions to 4 pages. That presents a number of difficulties in presenting results, especially if you need to include a lot of plots/graphics/tables. If you are publishing the result of research that took years and dozens or hundreds of people, things must be greatly condensed. It's not Scientific American.
Another contributing factor is that most papers target an expert in the field, so a lot of information is missing that is critical for people outside the field. When you try to read a paper outside your field, the missing bits of context and information make it really difficult to understand.
Extending that, is the fact that clarity is relative to your audience. Specialized jargon that is opaque to the layperson is also (ideally) far less ambiguous than everyday language. If you're familiar enough with the jargon, technical writing isn't just more concise, it's also an actual increase in clarity due to the precision.
Another problem I find is that important concepts are often given without motivation of why they are important. A proof obviously starts with definitions of terms, but it's worth knowing why these concepts are important and why they are defined a certain way.
the SCIgen example he cited at the end of that post is an extreme straw-man ... it was accepted into a fraudulent conference without receiving any peer reviews. the more reputable the conference or journal, the harder it is to BS the reviewers with obfuscation.
I disagree. I am a late-stage PhD candidate in CS, and I cannot stress the importance given to quality writing in academic circles. All of us know that clear, concise writing plays a large part in getting your paper accepted, and we're repeatedly told that if a reviewer did not understand something in a paper, the fault should be assumed to be with the writing. Strunk & White's "The Elements of Style" is a permanent fixture on most grad student desks.
My point is: if you're reading a reputable CS paper published at a top conference, it is less likely to be hard to understand due to writing flaws than due to the complexity of subject matter. Journals are a different beast, but most good CS papers get published at conferences anyway.
I agree with this. I've read so many 10-15 page papers that could have been explained more succinctly and clearly in an average length colloquial blog post. Researchers come up with some dumb unoriginal idea and dress it up so it looks complicated and new. Getting your paper accepted is not always about quality.
I think it's important to recognize that the truly interesting and useful research papers are hard to understand, too.
It's not supposed to be Harry Potter, and if you don't actually sit down and work through the steps on page two, you have no right to complain that page three is hard to understand. Feynman had a nice anecdote to this effect in 'Surely you're joking'...
Some of the most useful research papers I've come across have taken months for me to actually understand.
Nope. Here's the anecdote, cut'n'pasted since I can't direct-link to it:
> During the conference I was staying with my sister in Syracuse. I
brought the paper home and said to her, "I can't understand these things
that Lee and Yang are saying. It's all so complicated."
> "No," she' said, "what you mean is not that you can't understand it,
but that you didn't invent it. You didn't figure it out your own way, from
hearing the clue. What you should do is imagine you're a student again, and
take this paper upstairs, read every line of it, and check the equations.
Then you'll understand it very easily."
> I took her advice, and checked through the whole thing, and found it to
be very obvious and simple. I had been afraid to read it, thinking it was
too difficult.
Agreed. Most papers, even good ones, are hard to understand because the subject matter is hard. They are written with the assumption of a certain amount of prior knowledge of the field, and depending on the intended audience, very little background might be explained.
Writing clearly is also remarkably difficult and not everyone can do it. I have often found that the initial exposition of a scientific concept is not the easiest. After it been been digested by the community for a while, clearer ways of explaining it might be found.
Now, there are controversies like the Bogdanov papers, and even successful hoaxes like the famous Sokal paper, but I think these are exceptional, and usually reviewers _will_ say that a paper is "flat out bad" despite, or even because of obfuscation.
In Biology, most of the paper seems to be spent explaining why the new result is relevant and how it differs from previous work. It's usually dry as a bone but I wouldn't call it obfuscated. Readers should probably have Wikipedia handy for defining terms.
Long papers tend to have a heck of a lot of research. A short paper could be about a single experiment, but often a writer will publish a larger body of work in one paper. Value added, I suppose.
19 comments
[ 2.9 ms ] story [ 58.2 ms ] threadSo, say I'm a researcher. I've written a paper. It's a mess. I can put some more time into it, rewriting it to be more understandable, or I can use that time to do research & writing for another paper. In the former case, I get one clear paper; in the latter, I get two unclear ones. The people on my tenure committee are going to read my publications list, but not the papers themselves. Which scenario makes me look better to them?
Typically, the more complicated the language, the less is being said. Ever read legal documents?
There might be "shadier" reasons, but if that explanation makes sense, I for one would like it a lot if some researchers got in the habit of having longer explanations somewhere, say on their website. I don't know if that'd somehow break implicit rules in the academic field, but it'd lead to less head-banging-in-the-wall-trying-to-understand-what-that-damn-symbol-means.
My point is: if you're reading a reputable CS paper published at a top conference, it is less likely to be hard to understand due to writing flaws than due to the complexity of subject matter. Journals are a different beast, but most good CS papers get published at conferences anyway.
It's not supposed to be Harry Potter, and if you don't actually sit down and work through the steps on page two, you have no right to complain that page three is hard to understand. Feynman had a nice anecdote to this effect in 'Surely you're joking'...
Some of the most useful research papers I've come across have taken months for me to actually understand.
> During the conference I was staying with my sister in Syracuse. I brought the paper home and said to her, "I can't understand these things that Lee and Yang are saying. It's all so complicated."
> "No," she' said, "what you mean is not that you can't understand it, but that you didn't invent it. You didn't figure it out your own way, from hearing the clue. What you should do is imagine you're a student again, and take this paper upstairs, read every line of it, and check the equations. Then you'll understand it very easily."
> I took her advice, and checked through the whole thing, and found it to be very obvious and simple. I had been afraid to read it, thinking it was too difficult.
Writing clearly is also remarkably difficult and not everyone can do it. I have often found that the initial exposition of a scientific concept is not the easiest. After it been been digested by the community for a while, clearer ways of explaining it might be found.
Now, there are controversies like the Bogdanov papers, and even successful hoaxes like the famous Sokal paper, but I think these are exceptional, and usually reviewers _will_ say that a paper is "flat out bad" despite, or even because of obfuscation.
Long papers tend to have a heck of a lot of research. A short paper could be about a single experiment, but often a writer will publish a larger body of work in one paper. Value added, I suppose.