by kcolford on 11/17/21, 6:28 PM with 60 comments
by spekcular on 11/17/21, 7:50 PM
Regarding "grandmothering": I agree with the criticism of the first example. Explaining basic points of the field in a vague way is obviously not helpful. The second example is not as compelling. The key point is that the "..." after "In recent years, the study of preconditioners for iterative methods for solving large linear systems of equations, arising from discretizations of stationary boundary value problems of mathematical physics, has become a major focus of numerical analysts and engineers" usually contains a string of citations. These citations serve to point the reader to the recent works mentioned in the sentence, which may not be readily accessible to someone who doesn't actively do research in that area but is otherwise knowledgeable about numerical computing.
In particular, the author reasons such introductions are bad because "the bulk of the paper is accessible only to those sufficiently expert in the field to know everything in the first two paragraphs of the introduction cold." But this is just wrong. There are plenty of math/physics papers where I can follow the arguments line-by-line, but I don't know the state of the art in the field or why the problem under consideration might be important. I don't think I am alone.
Regarding, "A table of contents in a paragraph": I think the author is partially correct. For short papers, it's perfectly fine to fold this part into the introduction (e.g. in the outline of the proof). But for longer works where the proof is decomposed into multiple lemmas and sub-lemmas, these can be very useful. If one writes the proof in a very clear and structured way, then maybe such "shotgun summaries" can be avoided. But this is not always possible.
Regarding conclusions that only repeat the introduction: I agree here.
by porcoda on 11/17/21, 7:31 PM
I’m more bothered by people who use bizarre notation, don’t provide sufficient definitions/background, and don’t give enough information to reproduce the results myself (and I’m not talking about a git repo). I could care less if they have useless sentences and repetition if the content and methodology is complete and understandable.
by truly on 11/17/21, 7:48 PM
Hence many papers contain exaggerated claims with respect to practicality, importance and so on.
Another sin is that the results need to be "difficult" and "surprising" in order to publish. Hence, if you present your story in a simple-to-understand fashion, you run a high risk of rejection. Better not simplify your results before publishing -- keep all original notation, even if you figured out you do not need that many indices.
This has become a dogma and there is little chance of all this nonsense stopping anytime soon.
It is refreshing to read old papers that merely get to the point and are significant while being nice to read.
by rrobukef on 11/17/21, 8:34 PM
-----
* The grandmothering: Too often have I read an abstract, not understood it (since an abstract is allowed to be dense), and quit on these first lines orienting the paper in the field . If it's your field these lines don't cost, if you're a newcomer, or from another field, you get the keywords you need to know before starting. And often you know this paper isn't what you were looking for. As for the near-meaninglessness of this sentence: look up the first sentence of any book. You can't put 100 pages in one sentence.
* The table of contents: writers can't actually insert a table of contents, yet a paper needs it. True, nobody cares what's in Section 5, yet without this sentence you don't know when it will end, you don't know what you get. You care about how the content escalates. Also note that each of the sentences is more than just the title of the section. The actual title of Section 6 is just 'Time complexity'.
* Conclusions that don't: His solution is literally the opposite of what is taught. Yes to a perspective, no to new information. Also his example is incomplete, three more sentences follow that are not summarizing.
by ModernMech on 11/17/21, 8:35 PM
Moreover, it’s really not telling me what I know — it’s telling me what the author knows. If the author has a strange perspective on things I think I know, then it’s going to make understanding the rest of the paper much harder; how can I trust this author about unknown subjects when we can’t agree on topics of which we ostensibly share a common understanding? Such context is especially important when we are talking about the far edges of knowledge.
by pfortuny on 11/17/21, 7:38 PM
by joe__f on 11/17/21, 10:59 PM
I disagree with most of their suggested ways of doing this. * Grandmothering: I didn't understand the point here. I think writing an overview sentence in your abstract can be helpful. * Table of contents in a paragraph: Whatever it's just one paragraph, if you don't like it don't read it * Conclusions that don't: I think the having first paragraph of your conclusion as a reintroduction is very helpful. I sometimes read only the conclusion, and I often find the summary paragraph at the beginning of the conclusion to be more direct than other parts of the paper. Clearly the example of changing the tenses of the verbs in the introduction and putting this again as the conclusion is inexcusable, but I'd be surprised if that happened other than in the one place the author came across it
by kazinator on 11/17/21, 8:32 PM
But that just follows from the time-honored recipe for essay writing:
1. First tell 'em whatcha gonna tell em.
2. Then tell 'em.
3. Then tell 'em whatcha told 'em.
To conclude doesn't mean to draw some new logical inference, but just to bring the paper or talk to an end.
You don't introduce anything new in a conclusion.
Not any kind of conclusion.
E.g. a musical symphony will rarely introduce entirely new themes in the last bars. Instead various ending devices occur, like condensed re-statements of themes that occurred previously.
by Robotbeat on 11/17/21, 7:36 PM
by aimor on 11/17/21, 8:10 PM
by scrubs on 11/18/21, 2:05 AM
One sin that ought to be included replete in CS, math, and physics papers my English writing instructor made me stop: parenthetical phrases. Excessive footnotes are another variation.
Readers are not interested in the author's breadth-first association of one fact to 87 other facts.
Fixing this requires two things: knowing associations is certainly good. But know what's for you and what's for the reader. Second, know the point you're trying to make and nail it without bringing in half baked facts or edge cases.
In my papers I cut all that out. I either connect and explain clearly or I omit.
Re: grandmothering: I totally agree with op: if I'm a specialist in the area it's not value add. Like too many Ted talks it's people telling the in crowd what they probably already agree with. If not a specialist it's inaccessible. Either way it's bad, it's waste.
by MathMonkeyMan on 11/18/21, 2:56 AM
Writing an academic paper is the existential obligation of its authors. There's something in there that you want to get across to somebody, but the rest of it be damned. I'd expect the quality of the prose to be a low priority, not unlike in much developer documentation, school book reports, speeches, and other writing that is mostly filler to meet an obligation.
by kbenson on 11/17/21, 7:27 PM
by onhn on 11/17/21, 7:47 PM
Nothing like personal attacks to get your point across. Well done author.
by jmmcd on 11/18/21, 8:38 AM
We need a better way to establish consensus for things that can be standardised and we need more taste and tolerance for things where opinions can differ.
by evouga on 11/17/21, 9:21 PM
As for the comment about bad project names and acronyms: Jonathan is most famous for developing the “Triangle” code for Delaunay triangulation, and “Triangle” is of course impossible to Google. Should have chosen a more distinct acronym! :)
by graycat on 11/18/21, 3:03 AM
Here would be my list of the worst sins or how to avoid them:
(1) A term is a word used with other than a standard dictionary definition. Technical fields are awash in terms. Make sure to try hard to define all the terms you use; with some judgment applied, can omit definitions of some terms certainly well understood by any reader with enough background to get anything from the paper. Poorly or undefined terms can be one of the best ways to lose readers. Here, hoping to get more readers, maybe inviting them into the field, bend over backwards in defining terms, that is, maybe give some term definitions for readers who, really, have little hope of getting much from the paper -- at least, when such a reader gives up they won't blame the author!
In addition, for any term new or relatively new to the field, with some judgment, might also include for the term motivation of its importance and examples of its usage -- i.e., make clear the importance, relevance, value, usage, etc. of the term.
(2) For nearly all acronyms, standard in the field or not, for the first use of the acronym include what it abbreviates. Here to be sure are doing well on this issue, strain and bend over backwards and include, say, even TCP (transmission control protocol), SMTP (simple mail transfer protocol), CPU (central processing unit) -- sure, I'm suggesting bending over backwards. Several slaps on the wrist, an hour in the corner with wearing a dunce cap, and a coating of tar and feathers for each acronym used but not defined with what it abbreviates.
(3) For how to write math, in two parts, (A) and (B):
(A) Take some theorem proving courses from some of the most precisely written texts, say, P. Halmos, Finite Dimensional Vector Spaces, W. Rudin, Principles of Mathematical Analysis, Neveu, Mathematical Foundations of the Calculus of Probability, E. Nering, Linear Algebra and Matrix Theory, Royden, Real Analysis where do a lot of the exercises as homework and where the professor DOES read and remark on your writing. Neveu was a Loeve student at Berkeley. Nering was an Artin student at Princeton. As I recall, Rudin's background was in Austria, although I don't know who his professors were. Of course, Royden was long at Stanford. For more, if have time and insist on some really good examples, read some of Bourbaki.
By the way, on use of we, that is standard. So we might have,
"Given topological spaces X and Y and a function f: X --> Y, we say that function f is continuous provided for each set B open in Y f^(-1)(B) is open in X.
So, that's a vanilla example of using we in mathematical writing.
(B) Do some things none of those texts do: Include some intuitive views, some helpful pictures, motivation via applications in math and also outside math, and outlines of research directions.
There is a lot of question about how appropriate is suggestion (B); in math my guess is that there is no question about the relevance, wisdom, importance, value, etc. of suggestion (A).
For the goal of writing math, suggestion (A) is important: Tough to expect good success without the texts I listed or other texts written with similar care. That level of care is extreme, tough to find and learn, and much tougher to do. Writing math with the care of those texts seems to have been understood and practiced significantly often only after about 1950 or 1940. And for at least one course from at least one such text, DO have the good homework grading of a good math professor.
Personal experience and lesson: While I'm not much interested in being a professor, I have published some papers, and from getting those papers reviewed my guess would be that about half of the reason the papers passed review, and they always did with no significant revision or difficulty, is that I wrote the math with nearly the care and precision of, say, W. Rudin. Point: To critical readers, any lapse in that level of care and precision can be like a worm in a baked apple, perhaps otherwise terrific from brown sugar, butter, etc. Or, might guess that, in nearly any field, the good work is less than 10% of the total, maybe less than 1%, with the rest flawed, maybe as bad as that apple with a worm. In writing math, it is way too easy to be in the bottom 99% or 90% just from the care and precision of the writing, and suggestion (A) is IMHO (in my humble opinion) a good way to have at least the writing quality keep you in the top 90+%.
For math used in computer science, for how to write that math carefully, there are of course examples from D. Knuth.
by paulpauper on 11/17/21, 10:46 PM
by swayvil on 11/17/21, 7:42 PM
They say that the google crawler won't take you seriously unless you include your life story as prelude to that recipe for pb&j. Is this making everything less clear?
by toxik on 11/17/21, 7:49 PM