What worries me the most isn't Word doing it with English - it's multiple applications and sites doing it in a consistent-ish way with multiple varieties, including marginalised ones. And it all boils down to designers violating a rather simple principle - "don't bloody assume":
- Don't assume that the orthography of a language has a single standard, instead of multiple or zero.
- Don't assume that the user is even trying to follow a standard.
- Don't assume that your software/site has a good grasp of that standard.
- Don't assume that your input is welcome. (Clippy, I'm looking at you. Google Search, too.)
- Don't assume that you know what the user "means".
But the cherry of the cake isn't mentioned in the text: it's the sort of muppet who assumes language based on country.
Based on a quick, albeit arbitrary, experiment, if Harper Lee had used Word to write To Kill a Mockingbird, the software's clarity refinement would have suggested changing: "I never loved to read. One does not love breathing," to "I never loved to read. Breathing is necessary." Does this remove the poetry and depth of the original?
Yes. And it also removes clarity, even if the software is "trying" to increase it.