# Retroactive automatic editing of question punctuation affecting correctness

The question formerly titled "Does Stirling's formula give the correct number of digits of n!?" is now titled "Does Stirling's formula give the correct number of digits of n?". This was not due to an edit, and is incorrect. I speculated that this might be the result of a retroactive automatic systemwide editing of punctuation.

Searching meta.SO, I found this answer that seems to show that this was done intentionally. I can understand the motivation to clean up questions with bad punctuation, but I am concerned that it may remove certain instances of correct notation appearing in titles. There are plenty of ways to work around it in the example given above (e.g. by writing ...n factorial? or ...$n!$?). But I wonder if it is worthwhile enough that we should have to work around it, and I wonder if there have been/will be other instances of valid mathematical notation conflicting with automatic punctuation editing that is not as easy to work around, or other cases of old posts like the one above that may confuse readers during the potentially long interval before they are corrected.

I am not making this a feature request, because I would like to hear what others think. I am undecided on whether the benefits outweigh the costs. Admittedly, I may have (luckily) happened upon the only instance of the changes making a title incorrect, but I have no way of knowing. Please share your thoughts if you would like to.

• well, that question could do with a thorough editing anyway, could it not? – Jeff Atwood Aug 10 '11 at 8:21
• @Jeff: That's not the point, is it now? The point is that automated editing can be harmful. – Asaf Karagila Aug 10 '11 at 8:23
• @Jeff: Yeah, the body could use some work, but that is another story. KennyTM did his best, but the OP removed the LaTeX for some reason. – Jonas Meyer Aug 10 '11 at 8:24
• What is especially... inconvenient is that automatic editing leaves no traces in revision history, so one can't even understand what happened... – Grigory M Aug 10 '11 at 9:05
• (Posted last comment as suggestion on Meta.SO.) – Grigory M Aug 10 '11 at 9:27
• Moreover this "feature" makes incorrect many other mathematical questions, e.g. any terminal use of ellipses is deleted, and ellipses occur in many mathematical objects (power series, sequences, ...). So unless they are in TeX they will be silently deleted. – Bill Dubuque Aug 10 '11 at 13:37
• Perhaps this deserves posting where other stackexchangers will see it. – GEdgar Aug 10 '11 at 14:22
• To paraphrase what Theo commented in another post: "I was expecting humans to be editing my posts, not some mindless scripts..." – J. M. isn't a mathematician Aug 11 '11 at 1:57
• ...and most galling is that these auto-edits are untraceable. If the SE overlords will be saddling us with their scripts, it would be nice if they, you know, left a trail for debugging purposes? If (smarter) humans' edits are always logged, why should dumb scripts be exempt? – J. M. isn't a mathematician Aug 11 '11 at 2:01
• @J.M.: You probably noticed, but just in case: That request has been posted by Grigory M in another thread (also linked 5 comments up). – Jonas Meyer Aug 11 '11 at 2:04
• I did @Jonas, but still, thanks. :) – J. M. isn't a mathematician Aug 11 '11 at 2:08
• @J.M. \phantom, hmm? Clever. What is phantom actually for? – davidlowryduda Aug 11 '11 at 2:23
• @mix: This might illuminate \phantom{} a bit... :) – J. M. isn't a mathematician Aug 11 '11 at 2:27
• Cue the old joke: "Professor Smith, I've finished typing your paper. By the way, I noticed several places where you misspelled 'if' with an extra 'f', so I corrected them." – Nate Eldredge Aug 26 '11 at 21:54

To expand on Noah Snyder's answer: If you agree that no automated editing of old questions or answers should be done in the future without being first proposed and agreed upon on this meta, please vote this answer up. If you disagree, please vote this answer down.

• I'm going to have to disagree with this particular idea, simply because of the scale of such a requirement. If we catered to every site that had an opposition to a global feature here and there, and had to get through each one of them, it'd immensely slow the process of implementing global features considering there are 60 sites currently and quite more expected. I certainly can understand the discomfort of what happened here, and something could be done to mitigate this, but I can't say I find that this solution is truly sustainable. – Grace Note Aug 12 '11 at 14:03
• @Grace If you desire to survive in a competitive marketplace then at some point you will be forced to deal with the fact that you cannot treat all SE sites the same. Better to start sooner than later in configuring the software to smoothly deal with such localization demands. It's not rocket science and it need not "immensely slow down" anything if implemented correctly. – Bill Dubuque Aug 12 '11 at 14:32
• @Grace By the way, worth emphasis is the fact that already math.SE has some site specific localizations, namely MathJax support, which is not available on all SE sites. – Bill Dubuque Aug 12 '11 at 19:00
• @Bill A lot of the sites have specific stuff that's only enabled on them if not a small subset of sites, and that's fine. I'm not opposed to being able to screen out sites from global features. – Grace Note Aug 12 '11 at 19:03
• Shorter version of Bill's comment: "one size doesn't fit all..." – J. M. isn't a mathematician Aug 13 '11 at 6:58

While it is hard to form a strong opinion about this without knowing how many posts are affected, a priori this seems like a dubious feature to implement on math.SE, given the particular (and sometimes peculiar) way that punctuation is used in mathematics.

While I appreciate the desire to clean up badly typeset questions, my guess would be that there are few posts that are rendered incomprehensible by virtue of poor punctuation, whereas there is a real possibility of rendering a piece of mathematics incomprehensible by altering its puncutation (and Jonas has given a real, and illustrative, example of this possibility).

Is it possible to get more data on this issue? What is a typical example of a badly typeset post that is being improved by this automated process? And how many such posts are there?

• According to the linked meta.SO question, a little more than 130,000 questions were affected. – Qiaochu Yuan Aug 10 '11 at 15:36
• @Qiaochu Let's speak about Math.SE only here -- which makes it more like 2 errors in 15 (fifteen!) edits (according to Shog9). Doesn't look too effective to me... – Grigory M Aug 10 '11 at 15:54
• @Grigory: I was a bit off... (see my edit) - only looking at question / exclamation marks didn't reveal the vast majority of the modifications came from correcting more subtle problems. – Shog9 Aug 10 '11 at 16:27
• @Shog9, Qiaochu Yes, I got carried away, sorry. (But the real point is, SO sooo huge so it really has some specific problems not relevant to Math.SE. In particular, the question whether automatic edits are effective should be discussed separately for SO and for other sites, I believe.) – Grigory M Aug 10 '11 at 17:06

I think it's pretty clear that the "!?" and "..." changes should not have been automatically implemented on Math.SE. The main issue I see going forward is that this would have been obvious to anyone who writes mathematics on a regular basis, just as the foolishness of getting rid of "+?" would be to a user of SO. What we need is some way to notice that changes are going to have bad affects on particular sites before they're implemented.

Is there an easy way to figure out what changes are about to be implemented short of reading every entry on meta.SO? Would more people from math.SE following certain tags help? Should there be a place where new changes are held for a day for discussion?

• Officially, the elected site moderators are intended to be the "go-betweens"; of course, nothing stops anyone else with the time and interest from watching Meta.SO. [feature-request] is a pretty good tag, as are the recent feature changes lists. We're also trying to get more of this stuff into the newsletters we send out to moderators, although deciding what's important and what's trivial can be a challenge. – Shog9 Aug 10 '11 at 18:20
• Ah good, I hadn't known previously that there was a recent feature changes list. On the other hand, if I'm not mistaken the feature we're discussing here is not on that list. – Noah Snyder Aug 10 '11 at 18:24
• @Shog9 Is there any sort of comprehensive change log for the SE platform? Any way to know when new revs are rolled out? – Bill Dubuque Aug 10 '11 at 18:58
• @Bill: there's an internal one, but it's not exactly usable for this (or any other external) purpose. The MSO feeds / questions are an attempt to get something usable, but not always comprehensive. – Shog9 Aug 10 '11 at 19:02
• @Noah: my understanding is that SE rolls out various small changes frequently enough that it would completely drag down the whole process for all new changes to be subject to discussion. The underlying philosophy is that it's easier to beg forgiveness than ask permission: that way we only end up having the discussions that people actually care enough to have (like we're doing right now). – Qiaochu Yuan Aug 10 '11 at 20:05
• "easier to beg forgiveness than to ask permission" - reminds me of the Emo Philips joke, quotedb.net/…’t-work-that-way-so-i-stole-one-and-asked-him-to-forgive-me-–-emo-philips/ – Gerry Myerson Aug 11 '11 at 0:59
• @Qia The point is not discussion of pending changes but, rather, keeping an eye out for (unintended) consequences of existing changes. The more information the community has the better feedback they can provide. For example, if I know that a recent rev rolled out major changes to the MathJax interface then that is a very valuable piece of information to know when troubleshooting the source of related problems. – Bill Dubuque Aug 11 '11 at 16:53
• @Bill: Noah's suggestion is very clearly for discussion of pending changes, and it is to Noah's suggestion that I was responding. – Qiaochu Yuan Aug 11 '11 at 16:55
• @Qia "The underlying philosophy is that it's easier to beg forgiveness than ask permission" You do understand what the proper game-theoretic move for the other player is? Make begging for forgiveness costlier. – Phira Aug 25 '11 at 16:05

Yeah, that's crap. I've made a small edit to side-step the problem (on this and the one other question with the same issue), but obviously it's something that we'll need to be watchful for. Fortunately, it's not common to end a title in this way.

As for the reason behind it, and whether the benefits outweigh the costs... Here's some data. This site has only 15 questions (as of the last data-dump) that end with ??, ?!, or !? - two of which were serving a useful purpose, and the rest being the usual "I'm incredulous?!" or "extra questiony question???" nonsense.

You could easily clean these up manually...

On the larger sites, or sites that aren't good about cleaning up titles, this can grow into a problem over time, leaving the site looking quite messy.

Edit: I'd missed that this script was also cleaning up titles that end with a space and then a question mark. That adds another 250+ questions to the list - probably a bit more work to get through by hand. There were also 80 or so questions ending in ellipses / repeating periods that would have been modified.

Math.SE may not be among the worst offenders as far as superfluous ending punctuation goes, but looks like it did get some benefit from the script in exchange for a few false positives.

• This is a classic example of a "big city solution" being misapplied on a small site. Mathematicians have peculiar requirements on their punctuation, and "false positives" that completely break meaning are really bad. (Like the death penalty, eh?) – Scott Morrison Aug 10 '11 at 18:03
• The ellipses list includes a number of "false positives." – Noah Snyder Aug 10 '11 at 18:09
• @Scott: for sure. On a site this size, a few conscientious editors could have taken care of the whole bunch without any external assistance. – Shog9 Aug 10 '11 at 18:12
• @Noah: good catch. I've edited them (well, the ones I noticed) to avoid the problem. Note that some of these should have stuck out like a fist full of sore thumbs as lousy titles anyway... – Shog9 Aug 10 '11 at 18:17
• @Shog9, so why run this script here? My point was not that "big city solutions" are unneeded here, but that they are actively harmful. – Scott Morrison Aug 10 '11 at 23:42
• @Scott: It was run everywhere. Bit sites, little sites. So far, the number of false positives isn't high. As for why? Because the poorly-punctuated titles look bad and set a bad example for new users. This is no less a problem on small sites than on large ones. Yes, doing this automatically is less than ideal; in a perfect world, real live knowledgeable humans would be taking care of this, and this is actually much more feasible on smaller sites (we're making some effort to kick-start more)... but some always slip through the cracks. – Shog9 Aug 10 '11 at 23:53

I really think that if this such a problem at SO, then there should be separate scripts run on SO messages only. What's next? Say, leetspeak becomes way too common in some computer related SE. Consequently the geniuses at helm decide to run a script that substitutes all 3s, 1s and 7s with 'e', 'l', 't'...

• The day when somebody comes to post an m.SE question entirely in 1337 is going to be a most interesting day... – J. M. isn't a mathematician Aug 11 '11 at 16:14

To be honest, this is an incredibly small exception to the general idea that it's probably a good idea to get rid of multiple punctuation marks, and is very unlikely to be an issue for more than a handful of questions on math and a few other sites. I'm curious why you think making a few old questions slightly less comprehensible could potentially be more harmful than cleaning up 100,000 other questions.

• Perhaps, because, automagically altering a user's carefully composed text is a very bad idea. It's quite scary that the SE platform is implementing more and more of such Big Brotherish policies as time passes. Are your true opinions stifled now that you are an SE employee? – Bill Dubuque Aug 10 '11 at 13:29
• @Bill: no. I wish people would stop saying that. The content you generate on SE was never intended to be exclusively yours - there's a reason other users can edit your content, and that reason is that SE is supposed to be a community-maintained resource for the long term. Automating part of that maintenance is just a matter of efficiency, and if the automation makes a few mistakes, so what? They are easily fixable. – Qiaochu Yuan Aug 10 '11 at 14:31
• @Bill: if working at SE has changed my opinions in any way, it's helped me think of the SE team as a group of human beings trying to build a strong network rather than as monsters under the bed who are out to get us. I'm not sure I would call this "stifling" so much as correcting for the cognitive bias of tribal thinking. – Qiaochu Yuan Aug 10 '11 at 14:34
• @Bill It seems pretty clear that the mechanism was not put in place to alter "carefully composed text", but rather sloppily thrown out garbage. It is clear that no automatic algorithm will be able to tell the one from the other in 100% of the cases. Which is why Jonas asked the question whether the benefits outweigh the cost. You think that the cost cannot be outweighed, but others may think differently. I think that this is a serious and honest discussion that is worth having. Ideally, one should form his opinion after knowing what the benefits and the costs are precisely, and not before. – Alex B. Aug 10 '11 at 14:51
• @Alex B. Besides being a mathematician, I also have extensive experience in software engineering. My various critiques of the SE platform are based upon such. Many of the SE "features" are ill-conceived "hacks" or "kludges" as one would say in CS lingo. There are far too many subtleties involved in automatic text editing for such to be automated by machine. Such editing should only be done by humans - of which we have many here the last time I counted. – Bill Dubuque Aug 10 '11 at 15:03
• @Bill My point was that presenting your experience as concrete justification of your critique would be more valuable and would further the discussion more than comparisons with 1984 and insinuations that Qiaochu is not expressing his own opinion. – Alex B. Aug 10 '11 at 15:12
• @Qia The issue has nothing to do with other humans editing content. Rather, it has to do with ill-conceived buggy software hacks that forcibly alter our texts. On such matters it is always better to pull in the reins sooner than later. – Bill Dubuque Aug 10 '11 at 15:16
• @Shog9 That said, it is worth emphasis that because mathematics compresses so much information into its highly optimized language, the odds of altering its meaning by automatically changing a single character is much higher than in other fields. That's not to say it doesn't happen elsewhere, e.g. recall the famous joke about the importance of the lowly comma: this text message was sent from a wife to her husband: $$\rm\text{NOT GETTING ANY, BETTER COME HOME AT ONCE.}$$" – Bill Dubuque Aug 10 '11 at 15:58
• @Qiaochu: Thank you. For the record, I am disturbed by Bill's first comment and its upvotes. As for why I think it could potentially be more harmful, I think that Alex B put it well about knowing what the benefits and costs are before forming an opinion. I also think it would be possible to implement the auto-corrections on SO (where most of those 100,000 instances occur) without implementing it on math.SE (where there is higher potential for error). (My concern was also because of Grigory M's point that it is hard to see what is going on, but he has already made a new thread for that.) – Jonas Meyer Aug 10 '11 at 20:12
• Deletion of ellipsis leads to use of hyperbole. Are we going around in circles? – Gerry Myerson Aug 11 '11 at 0:57
• For the record: I upvoted Bill's first comment for the first sentence, but I certainly don't like the last two sentences... – J. M. isn't a mathematician Aug 11 '11 at 2:29
• @Bill: in my experience here and on meta.MO, it seems to me that some of the more vocal criticism comes from a strong "us vs. them" mentality that I find unconstructive. I was calling out this attitude. – Qiaochu Yuan Aug 11 '11 at 16:53
• @Bill: It is hard for me to understand how speaking of "scary" trends toward "Big Brotherish" policies can be constructive here. There is nothing sinister here. Automatic edits were done to help clean up many sloppy posts. However, there are potential downsides to this, at least on this site, due to "the particular (and sometimes peculiar) way that punctuation is used in mathematics," to quote Matt E. Jeff Atwood and others who run these sites cannot please everyone at all times, but they are clearly invested in the success of the sites. – Jonas Meyer Aug 11 '11 at 17:17
• @Bill: I do not presume to read your intentions, and I suppose my being disturbed says as much about me as anything. Excessive hyperbole in the direction of comparing those with whom you have a disagreement with totalitarian thugs is humorous in a way, but in a way that makes it hard for some of us to take your point seriously. Tastes differ. Personally, I prefer Gerry Myerson's brand of humor. – Jonas Meyer Aug 11 '11 at 20:29
• @Bill: "between Gerry and I"??? Where's the automated grammar-correcting software when you need it? – Gerry Myerson Aug 12 '11 at 2:10