Is it preferable to have equations in LaTeX format instead of images, and if so, is it a desideratum to convert images to LaTeX when possible? I have been going through some older questions and converting equation images to LaTeX, but realized I should probably get a consensus on whether this is desired.

  • 2
    $\begingroup$ I believe there is a consensus: see glen_b's post at stats.meta.stackexchange.com/a/2813/919. $\endgroup$
    – whuber Mod
    Jan 4, 2022 at 19:42
  • 1
    $\begingroup$ Wouldn't there be some tools like 'google lens' that could do this conversion? (Not criticising your admirable effort, but wondering whether it could be done more efficiently. It's amazing that this type of stuff still needs to be done manually. There should be some point that computers can do this work without much human effort in setting up and managing the computer) $\endgroup$ Jan 10, 2022 at 10:07
  • 2
    $\begingroup$ @SextusEmpiricus: There may well be, but I'd expect a human who understands the content to be more accurate, especially for the low quality images. $\endgroup$ Jan 10, 2022 at 12:31
  • 1
    $\begingroup$ @Scortchi maybe that background/content is the last step that separates AI from humans. We can make algorithms perform a specific task but it is not yet running automatically because the learning is too specific for a task and we do not exactly copy human capacities, let alone exceed them. $\endgroup$ Jan 10, 2022 at 12:36
  • 1
    $\begingroup$ My comment was more some rant/frustration. I wonder why, by now, we are not there yet that we can order a computer in simple language. $\endgroup$ Jan 10, 2022 at 12:37
  • 2
    $\begingroup$ A petty consideration: in posts where OP posts a long complicated equation, I often go to "edit" mode to copy swaths of latex code when supplying an answer. There's only so many times I can write \sqrt{n} \left( \bar{x} - \mu \right) \rightarrow_d \mathcal{N} \left(0, \sigma^2 \right) for $\sqrt{n} \left( \bar{x} - \mu \right) \rightarrow_d \mathcal{N} \left(0, \sigma^2 \right)$ $\endgroup$
    – AdamO
    Jan 10, 2022 at 22:24

1 Answer 1


LaTeX can be read by screen-reader software used by people with impaired vision, & is often more legible in any case. It can also be searched, copied, & modifed. So it's a good thing you're doing, but make sure the posts are worth it—there'd be no point bumping a dozen five-year-old posts with no up-votes or answers to the site's front page.

  • 11
    $\begingroup$ Another reason not to bump old posts is that rendering $\LaTeX$ for old posts might introduce errors, either by transcription error or because the original image/notation is simply unclear. For a user who is active, it's plausible that they could correct the error or make a comment, but for an old post authored by an inactive user, that feedback is unlikely to appear. $\endgroup$
    – Sycorax Mod
    Jan 4, 2022 at 19:52
  • 4
    $\begingroup$ @Sycorax: I wouldn't advocate anyone's LaTeXing an equation they don't understand, which ought to mitigate the risk of introducing errors. $\endgroup$ Jan 6, 2022 at 10:50
  • 6
    $\begingroup$ I think whether to edit & bump an old Q is a judgement call. If the OP uses their judgment, it's likely to be OK. $\endgroup$ Jan 7, 2022 at 12:22
  • $\begingroup$ I think editing any existing Q or A to render in LATEX/MathJax what was originally presented as an image or plaintext is more or less always desirable. Also, I wonder if there is a poorly conceived notion of 'realness' of content bumps at play here, as though "people are polluting the algorithmic measurement of 'real' interest with their 'ersatz' interest!". Algorithmic bumping of Qs is ephemera: if you do not share whatever algorithmic interest is currently active, wait 30 minutes, right? $\endgroup$
    – Alexis
    Feb 15, 2022 at 23:09

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .