It’s not always easy to live in an Anglophone scientific world when English isn’t your first language. When careers are built upon the ability to communicate ideas clearly and eloquently, struggling to find the right words can be a real hindrance to explain your science in a way that is taken seriously. Contrary to popular belief, it’s not something you can simply “work” on. Often, it doesn’t matter how many books you’ve read, how many years of education you have, or how articulate you are in your original language — your brain will refuse to summon the right expression, or get stuck in a construction that a native speaker would never use. Struggling with a second language is very much a biological phenomenon.

The standard recommendation for ESL (English as a Second Language) speakers has long been to ask a native colleague to read through any text that needs to be published or submitted somewhere (such as an article or a grant application). Well-intentioned as this advice may be, there are multiple problems with it. Lingua franca or not, only 15% of the world population speaks English, of which only 5% are native speakers — meaning that for most scientists not working in Anglophone countries, the option is rarely available. Even when available, it is unreasonable to expect these colleagues to add charitable proof-reading to their workload simply because they happened to be born speaking a different language. But, most importantly, I have always felt — and I want to emphasize that I truly believe most people who issue this kind of advice to be well-intentioned — that the underliying message sounds too much like “you need vetting by a member of our select linguistic club if you want your ideas to be taken seriously“.

In practice, for all of us who learned English after the critical period, there were really not very many options open. One of the options was to pay for a proof-reader, a native speaker who would read the text and make it more “pallatable”, a service which unfortunately comes with several disadvantages. For one, it takes several days to be completed, making it difficult to meet deadlines. It is also expensive — effectively a tax on ESL speakers. More importantly, while I have never hired a proof-reader myself, most accounts I’m aware of in the scientific world are quite negative, with examples where the proof-reader’s corrections made passages of the text harder or impossible to understand. Unfortunately, as bad as professional proof-reading was, there really was no alternative…

Until now.

1. Proofread your text

Rather than paying a stranger to read over your text, you can ask a large language model like GPT-3 to give you feedback on text. With the right prompt, you can get some light corrections alongside an explanation of why each change was suggested. I have found that the following prompt works best, applied paragraph to paragraph:

Depending on what you are working on, you may also want to use stronger prompts, such as “rewrite the following paragraph in an academic style“. My experience with this style of prompting has been that ChatGPT tends to do a full rewrite. The text changes so much that it no longer sounds like “my voice”. While there are instances where a full rewrite can be helpful, such as when it shows how to rewrite a sentence in a different structure that is easier to read, in most cases the resulting text appears so different and alien that I tend to ignore it.

2. Find in-context synonyms

Using an extensive and varied vocabulary is a cornerstone of writing with flair. This sounds great — except it’s difficult to come up with good alternative words when you are not a native speaker. While you may be tempted to check a thesaurus, it is possible that the word you find doesn’t quite fit in the sentence, for whatever reason. Well, that is not a problem with the magic of Large Language Models:

Alternatively, if you are looking for the right word to say something, you can just describe its meaning to ChatGPT and ask for a list of suggestions.

3. Discover expressions

Good writing, including academic writing, conveys meaning through colourful expressions that illustrate what is going on. Because we are talking about English, I am including the feared “phrasal verbs” that also have a significant place in formal use. These expressions come naturally to the native speaker (and, to a lesser extent, to those of us that have spent many years living in an Anglophone country), but may not be something you can readily use in for your writing. Fortunately for you, ChatGPT can help with that:

4. Check you are using the right style

Language is capricious: parts of it tend to be acceptable in some register, and completely out of place in a different situation. You don’t use “whilst” or “albeit” in the pub (if you do — you may want to look up the word “pedantic” too!). To make sure that a sentence is adequate for the kind of text you are producing, you can ask ChatGPT to check it:

Much has been written about ChatGPT being used for academic dishonesty. And yet, little attention has been given to the potential of this technology to help individuals with valuable insights but who struggle to express them. Think about it: how many valuable ideas remain unheard due to poor presentation? How many brilliant papers have been rejected because the writing didn’t conform to the academic standard, and the reviewers, hidden behind anonymity, decided to reject them without a second chance?

Technologies like LLMs are in their infancy and are yet to meet all necessary requirements in terms of bias, ethical use, and even safety. They are already creating inequality and their economical impact is bound to change things in unexpected, potentially negative ways. And yet, the potential of these applications to bring forward equity in the form of increased accessibility and cheaper education is truly astounding. We are living through the first days of a new printing press and it is up to us to shape it into a tool that improves the world.