The curious case of co-pilot and editing

The curious case of co-pilot and editing

Artificial intelligence has certainly caused a stir in the world of language editing and translating. And there is no denying that it does our job fairly well. Fairly well.

Sometimes of course, fairly well is enough. But in my field of editing and translating, i.e., research, academic papers, doctoral theses, health-related and medical texts, fairly good English is not enough. Things simply can’t be expressed in an approximately correct way. Facts must be 100% correct, and there can be no risk that a writer’s meaning or assertion changes or is misunderstood due to a nuance caused by a slightly “off” choice of word or phrasing.

But can AI help someone write good academic English? A client of mine recently approached me with a task that would involve delving into this issue, and I was more than willing to do a bit of research into it!

“Write this more concisely in academic English”

This Finnish-speaking client had written parts of her doctoral thesis in English, and had then asked Co-pilot to “write this more concisely in academic English”. She wanted me to check what AI had done against her original version, which she felt needed a good deal of editing.

But before I begin to go into this, I’d like to share a few personal thoughts about language editing, as this task made me really think about it.

What exactly is language editing?

So, what do I actually do with a text when I edit its language? I’ve listed the main tasks below:

• I condense verbose sentences

One of the things I do most often is chop irrelevant bits out of the text to stop them obscuring the real message.

(In fact, I have another blog post on this very topic! So much to say, so few words… Condensing texts to comply with word limits –)

• I remove redundancies

Sometimes things are said in a way that uses extra words that add nothing to the sentence

• I check the appropriacy of vocabulary, terminology, style and register

The words and tone of a text must be fit for purpose. An advert can’t use the same formal language as a report, and a report can’t use the snappy, colloquial language of an advert. The writing must always cater to the target reader. This may sometimes require a quick fix, such as:

But sometimes this task can be time consuming, and can involve looking up synonyms, or finding accurate and appropriate terminology.

(Another topic I’ve written about! Sounding academic but not pretentious: a balancing act –)

• I try to improve readability

This means ensuring that the text flows: ensuring that the reader doesn’t have to go back and reread a sentence to understand it. It also means checking that the sentences run on smoothly from each other stylistically and content-wise, and substituting any overly-complex words (again, bearing in mind the target readership).

To me, this is the essence of my work – this is where I can help the writer the most. It is also the part of my work that I love the most. It’s where I get down to the nitty gritty. And I believe it may potentially be the part of my work that is the hardest for AI to replace.

• I correct basic errors

Most laypeople think this is what language editing is all about, and yes, it is the nuts and bolts. A lot of my time does go to correcting grammatical inaccuracies. But this is perhaps more mechanical than thought-involving or creative.

The Voice

But in all of this, I always strive to keep in mind the writer’s voice. It is the writer’s text, not mine, and I must avoid making unnecessary changes to things that I may personally prefer to express another way. If something is not written in my personal style or according to my personal preference (and all editors have these!), but it’s correct in all the ways that matter, then I mustn’t touch it, no matter how tempting it may be!

So how well did Co-pilot navigate this task?

And now to the point of this blog post. The real issue. How well did Co-pilot actually do?

Let’s start with the pluses. Co-pilot did the following rather well:

And now the minuses. Some things Co-pilot did not fare so well with. Most of the examples below are ones in which I personally would have intervened:

Human vs machine – already an age-old issue

One of the seminars that I went to early on in the advent of AI was called “AI, master or servant?” an expression I’ve heard very often since. It made an impression on me then, and I still feel very much the same about it now.

There is a great fear that AI will become a master… and this risk is real to some extent.

However, as a servant, or a TOOL, Co-pilot, ChatGPT and their equivalents can be extremely useful for us editors. They can help us with mechanical issues, just as spell checkers and software such as PerfectIt have already been doing for some time. We can ask them for help when we can’t think of a synonym, or when that word is just out of reach of our memory.

But for nuances, things that just don’t “feel” right in a text, and words that can have different meanings depending on the context, and for retaining the writer’s personal voice – our subjectivity, our brains, our life knowledge and our sense of appropriacy are essential.

We language professionals know this in our hearts. AI can’t replace the human side of our work that requires sensitivity, common sense, a “feel” for language and our ability to recognise what is unbefitting of a particular context.

The verdict

There is no getting away from the fact that Co-pilot succeeded to some extent in doing what my client asked it to do, i.e. it made the tone of their text more academic. It corrected some basic grammatical errors. It made some lengthy expressions more concise. It made the register of the text more formal by changing some words and expressions. It reordered some clumsy structures to make the text more readable.

But notice my repeated use of the word “some” here. By no means did it do an excellent job. It actually LOST some text. It failed to pick up on certain things that required tweaking. It also did some things it shouldn’t have, and missed some aspects that I, as a human being with a living brain, would have picked up on. It made assumptions about issues that I personally would have asked the author about.

Altexta offers proofreading and translation services. I’m a native Brit, I grew up in Wales, but spent my childhood summers in Finland, at my grandmother’s house. My mother is Finnish, my father is Irish. Thanks to my mother’s perseverance, I’m bilingual. I have lived in Finland for 30 years and worked as a language editor and translator for 20 years. My specialties are factors related to working life, occupational health and safety.

I love my job!

In my blog, I give advice to Finns on how to write a good English text, using as examples the problem areas that I encounter the most in my language editing work.

Altexta tarjoaa kielentarkistus- ja käännöspalveluita. Olen syntyperäinen britti, kasvanut Walesissa, mutta viettänyt lapsuuteni kesät Suomessa, mummolassa. Äitini on suomalainen, isäni irlantilainen. Kiitos äitini sinnikkyyden, olen kaksikielinen. Olen asunut Suomessa 30 vuotta, ja työskennellyt kielentarkistajana ja kääntäjänä 20 vuotta. Erikoisalani ovat työelämään liittyvät tekijät, työterveys ja työturvallisuus.

Rakastan työtäni!

Blogissani annan suomalaisille neuvoja hyvän englanninkielisen tekstin kirjoittamiseen, käyttäen esimerkkeinä ongelmakohtia, joita eniten kohtaan kielenhuoltotyössäni.