‘Computer says no…’ how AI can fool you, and why a human editor is your best bet

Robot toy picture

Here’s a scenario.

An author just wants to make sure their text is tip-top, error-free and grammatically sound. They get Grammarly and Hemingway, and run their final text through it after getting it back from their editor.

Yikes! The checkers are having indigestion over what the editor’s done – or appears to have missed.

This is a real problem… or is it?

If we break down the situation, there are a couple of things going on here:

  • Writing checker programs run on artificial intelligence (AI)
  • Editors are human

And this is key: what the computer’s throwing up doesn’t necessarily mean that decisions taken during the edit were wrong, or important things overlooked.

Unfortunately, there are reports in the writing world of precisely that: panicked messages (and even disgruntlement) to editors about ‘mistakes’, because AI appears to have identified them.

To be fair, let’s credit such tools with the ability to run simple checks. They can help to weed out obviously glaring errors, and do help people write basic documents and emails effectively.  There is room for them in the writing space.

But fundamentally, AI is just a programme. It still requires a human to make good judgement calls from what it brings up – and writing checkers are well known in the publishing industry for delivering odd suggestions and false positives.

(An editor colleague tells an amusing story of having put actual passages from Ernest Hemingway through Hemingway. The results were – how can I put this? – interesting…)

In recent times, the spelling and grammar checker in Word has also started to butt in with inappropriate suggestions around what the AI sees as redundant words, or phrasing that would alter tone, meaning and change the author’s intention.

Spellcheck in Word is well known for delivering false positives – even downright incorrect, bizarre suggestions. Again, it is only as good as its user’s ability to understand and recognise what actually might be wrong – and this is where it can present challenge, especially to authors whose first language isn’t English.

What Word is doing isn’t grammar correction, it’s interfering – and badly. The author’s wording is a creative, stylistic choice, which editors respect. Good writing means knowing grammar, and knowing when those rules can be tweaked or broken for the desired effect.

This is why, as a language professional, I don’t use writing checkers – and encourage authors not to consider AI absolute gospel either.

The difference between editorial knowledge and AI is considerable:

  • Editors look stuff up – we access extremely good reference libraries.
  • We know what we don’t know, and always consult reputable, established sources for answers.
  • We do know language by virtue of the fact that we’re human beings who use it every day.
  • We also know terms of art, expression and forms from our subject-specialist degrees and professional training.

The natural language programming that writing checkers use to parse and analyse text is only that: a mechanical program.

Until it has the nuanced analytical judgement and expertise to deal with every possible language anomaly, grammar break, stylistic quirk or creative choice that real writing involves, you’re in far better and safer hands with a qualified human being who does.

So far, that AI doesn’t exist. You can relax and trust your editor to put your writing first, and deliver well.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.