Reader Comments

Post a new comment on this article

Error rates are interesting, speed not so much

Posted by cben on 29 Dec 2014 at 00:43 GMT

The data on error rates is interesting because (1) differences were quite significant (2) it suggests real usability problems, much stronger than a mild difference in speed (3) the price of uncorrected errors in papers is large. Authors are in much better position to correct errors than readers, and most papers are never revised after publishing, so any effort "saved" by sloppy authors will cost much worse in wasted effort by readers. Errors can easily obstruct comprehension to the point that readers will give up entirely (the best case being that reviews give up and reject, sparing readers and making author expend a lot of effort on resubmission).

The dramatic differences in errors were
- in favor of LaTeX in math formulas (IMHO, math errors are especially harmful!)
- in favor of Word in the table task.
It's hard to conclude who wins based on this, or to learn what the usability problems were; I wish the authors included more details on which errors these were (esp. samples of formatting errors).
As the authors mention, it's hard to compare LaTeX usability without comparing specific LaTeX tools. I wish the authors included more details on the tools used.

OTOH, the reported differences in speed don't convince me.
1. It's only meaningful to compare the speed of writing when the resulting error rate is the same (and acceptably low). If one group of writers produced twice as many errors, they should make additional pass(es) to fix these errors!
2. The speed differences were not that large (10%~30%).
3. This study only considered a small one-author case. I'd very interesting to see a study multi-author of *collaboration* efficiency; I expect the potential frictions and wasted efforts of collaborative editing to dwarf the speed differences reported here. (Such a study should include at least Word's track changes, compare & combine facilities vs. LaTeX with version control such as Git vs. LaTeX with real-time sync i.e. sharelatex.com or writelatex.com.)
4. If speed is the goal, it's easy to outperform LaTeX by writing in lighter formats such as reStructuredText or markdown, and converting to latex with tools such as Pandoc. Similarly, writing in LyX or TeXmacs and exporting LaTeX is *probably* more productive than LaTeX for users that prefer Word. I understand that would require a much bigger study; but leaving out those options means "LaTeX is slow to write" less conclusive.

Tool recommendations are also more needed for the collaborative case because collaborators are not free to choose each his own — all co-authors pretty much have to agree on one tool (converting back-and-forth generally results in much more friction than agreeing on a common tool you like/know less...). E.g. if you're using LyX, everyone has to use LyX; if you're using real-time sync, you can't mix it with version control, (almost) can't work offline and have to use the same in-browser editor rather than desktop editors of your choice (exception: Floobits lets you mix several desktop editors — but has no latex integration).

Whereas in the single-author case you're free to use whatever process makes *you* efficient as long as the final output is acceptable.

Competing interests declared: I'm developing a (free) collaborative markdown editing site with in-line LaTeX formulas. So am biased towards some approaches.