I’m not sure if I’m mad or sad. Okay, those are words I would rarely use. More likely, I’m just irritated because some automagick technology is telling me to simplify my writing. Increasingly, if I’m in a Microsoft product, a little Editor bubble will appear to let me know that there’s something non-conforming with my writing. The thing I’ve noticed most is that its suggestions, based on whatever machine learning it is using, are frequently wrong and as often change my voice. It makes me wonder what we are giving up to create this universally boring approach to writing.

I first noticed it on SharePoint. We’re a Microsoft 365 shop and I use the web browser versions of all the apps. Outlook in one tab, Teams in another, my calendar in a third. When I click on a document to edit, I tend to use the Excel or Word apps. The Editor is a new toolbar feature that brings together some of the document review features that I like. And many that I don’t.

A screenshot of Microsoft Word’s web app with the Editor pane open and the writing style drop down selected.

I actually like a bit of guidance on my language. I write a lot but do not consider myself a writer. My partner commented recently on a writing tic that I’ve had for many years that I’m now working on undoing. I’m pretty good natively on spelling, more iffy on grammar. I’ve even had my mother email me about typos in blog posts.

The Microsoft Editor is not limited to the Microsoft 365 universe. You can also add it, by way of a browser extension, to Microsoft Edge or Google Chrome for editing features across other apps, including GMail.

What I’m trying to say is that I’m open to constructive criticism. But one thing I love about reading is hearing the voice of the author. Their word choice, their decisions when to abide by a grammar or rule and when not to.

Fewer Isn’t Better

One thing you notice with the Microsoft Editor is that it seems to like fewer words. But fewer isn’t always better. Sometimes you need a few words to be more specific, to be clearer. Here’s a good example (for those curious, I’m using a part of the Wikipedia entry on “The Rain in Spain” song from “My Fair Lady”):

Microsoft Editor is suggesting that “nearly exhausted” is equivalent to “exhausted”

Microsoft Editor will highlight the potential problems in your text. This is really helpful because it guides you towards things to fix. This is just like it has been in the desktop apps for years – red squiggles under text for typos, green squiggles for grammar (you can choose your own colors too).

The document will still show typos with the standard red underlining squiggle. Now you also have dashed lines under things that trigger the Editor’s conditions. Click on the word or phrase with the dotted underline and see what it suggests.

If Eliza was nearly exhausted, she wasn’t exhausted. The machine learning is uncomfortable with uncertainty, though. If you click on the offending text and then expand the tip (click on the light bulb in the pop up dialog), you can see why it is making its suggestion:

Microsoft Editor prefers “ready” to “basically ready” and “prepared” to “adequately prepared” and “exhausted” to “nearly exhausted”

If I say I was “nearly exhausted,” that’s to convey something that’s different from “exhausted.” If I feel “adequately prepared,” it’s because I don’t feel “overly prepared” or I want to communicate that, while feeling prepared, I may have some lack of confidence about it. Perhaps, from a social aspect, I’m using the word to express to someone that lack of confidence so that I receive some reinforcement or encouragement. I’m not a machine.

I think Miracle Max would agree.

Montoya: “He’s dead.”
Miracle Max: “Ooo hoo hoo, look who knows so much. He’s only MOSTLY dead.”
from The Princess Bride

You get the picture. This is machined language. It is exposed to a very specific, repeatable process and the output would be the same repeatable, shiny output. But that’s not how people communicate. In particular, if you’re writing in a knowledge domain that requires nuance – like law, perhaps – then you want to make sure you’re being clear, not simple. The professional challenge is to be clear while being understandable.

YMMV

The Microsoft Editor allowed me to add YMMV (your mileage might vary) without flagging it on the Formal Writing setting. YMMV is definitely an informal usage. That gives me some concerns about what it’s up to. In the Word web app, the Editor pane is toggled on and off under the Home tab. But you can change the Editor settings under the Review Tab.

A screenshot of the Word web app toolbar (ribbon) showing the Editor Settings under the Review tab.

A bit confusingly, in the Word desktop app, you can find the Editor under both Home and Review tabs too. But both buttons just toggle the Editor pane on and off. You need to go into Word Options > Proofing > Writing Style and choose the settings to get to the underlying list of things that the Editor is checking.

The Grammar Settings dialog appears in front of the Microsoft Word Options dialog after clicking the Settings button beside the Writing Style choose under Proofing.

It’s worth knowing what you want to check and what you don’t. For me, the two Editor settings appeared to be set differently. They both have the same options available. This difference meant that the very same text, when pasted into the Word web app and the Word desktop app, resulted in different scoring.

A screenshot with the Word web app, in light grey and blue on the left, and the Word desktop app, in dark grey and black on the right. Both show the Editor pane for the same text, but with different Editor Scores and finding different problems (spelling, conciseness, and so on)

This consistency probably doesn’t matter if you only use one or the other of the writing tools. Or if you take the time to set the writing styles exactly the same on both apps. But it adds to my uncertainty about the Microsoft Editor value.

Yes, But

Every cloud has a silver lining. One thing that is not on entirely by default but I have turned on now are the bias checkers. If you want to know if your writing is inclusive or not, according to a machine-guided system, this might help.

The Microsoft Editor settings showing the Inclusiveness categories for writing

I’ve already manually dropped the he/she for the universal singular they, both in writing and in speech And I try to be careful in my writing in general about not using words that are exclusionary. But you don’t know what you don’t know and it’s my hope that having this turned on will at least flag for me possible problems.

I tried it with the Wikipedia entry for “Critical Race Theory,” which I thought might have some terminology that could challenge the machine learning. But the writing for that was found to be impeccable.

Screenshot of the Wikipedia entry for Critical Race Theory pasted into the Word web app and with the Editor pane showing 98% approval, below the text.

For the same reason that I leave the spelling and grammar check on in Word, I’ll probably leave Editor on. But lightly. To that end:

  • The Editor Text Predictions will stay off. I find these prompts while I’m typing to be distracting from my thinking as I write. I’ve already thought out my sentence, perhaps paragraph. Having some ghostly text appear at that point throws off my thinking. And I have yet to accept a suggestion.
  • There is a overlay bubble in some Microsoft app. I’ve turned it off and now I can’t find it again. But it is essentially a counter, without the Editor pane, that shows issues. It hovers above your text. I find it irritating. You can turn it off but, having done so, I have no idea where it first appeared.

In other words, I am happy to have machine learning-guided editorial ideas. I love that the readability information (Editor uses Flesch-Kincaid) is built-in. Grammar, spelling, readability are all things that I’ve focused on with my writing in the past. They are also things that I think anyone might think about if they are writing for an unknown reader. There isn’t a judge or legislator or regulator or lawyer in the world who couldn’t learn to write better than they do.

At the same time, I will continue to be somewhat guarded about which suggestions I take. Like the messy human I am, I’ll use the parts of machine learning that seem to work. I will leave the parts that take away my voice, and try to make me conform to some machined perfection that does not sound as alive.