h1

ChatGPT: Some uses for editors

January 21, 2023

After doing a few ChatGPT experiments recently and having a lengthy Zoom discussion yesterday with Adrienne Montgomerie, a freelance Canadian editor, I have some initial thoughts on how editors can use this as another tool in their suite of editing tools. And just like any other tool (e.g. PerfectIt, Editors Toolkit, even Word’s find and replace and spellchecker) the results ChatGPT (and similar AI tools) suggests should be used judiciously and with a human eye and brain involved—keep what fits your purpose and the context, and ignore the rest. And if you don’t want to use it, don’t. Some people eschew looking up dictionaries and adhering to style guides, others refuse to use other automation tools—and that’s fine (there were plenty of people who refused to see any benefit in word processors, and before that the printing presses that replaced pen and ink).

But for those who are interested in how AI tools can help in the editing process, here are some of my suggestions (no particular order) based on just scratching the surface of ChatGPT—no doubt there are more, so feel free to add your suggestions in the comments:

General advice:

  • Phrase your request in various ways—you will get quite different results depending on how you phrase your request, or whether your request is broad or quite specific
  • If what you get the first time doesn’t really work, get it to rephrase in a different way (there’s also a Regenerate Response button you can click to get another variation)
  • Use it as a starting place for ideas when you’re stuck, not as the end point (e.g. if you’re an author writing a report from scratch, you might use it to generate an outline based on the topic—again, you’d use this as a starting point and would need to be aware of what needed to be added or deleted for your circumstances)
  • Try some other tools too (https://cybertext.wordpress.com/2023/01/19/chatgpt-is-not-the-only-ai-writing-tool/)

As with any editing, your human brain is the best tool you have even if you use an AI tool to kick off the process:

  • Be aware of obvious or potential unconscious bias:
    • Who are the people behind the AI tool? What is their demographic? (e.g. are they young, predominantly white, male programmers living in a ‘developed’ world?) You may never know, but be aware that the people behind these tools have their own biases as to what they include or not. For an excellent documentary on bias (in facial recognition software) that reflects the demographic of programmers, watch Coded Bias on Netflix
    • Where have they drawn the information from? (good luck with that!)
    • Is there inherent bias in how the responses are phrased? (e.g. are certain groups ignored, such as women, people of colour, people over or under a certain age, entire regions of the world, people with various disabilities?)
    • Bias could be any one or more of these:
      • Is the response applicable to all, or just those in particular societies/countries/regions/socioeconomic groups?
      • Does the response ignore or discriminate on the basis of race, ethnicity, gender, age, ability/disability, etc.?)
  • Be aware of language limitations—as far as I am aware, ChatGPT only responds to requests in English and various computer programming languages, thus excluding/ignoring much of the world’s population, and there are certain areas (e.g. mathematics)  it has limited knowledge about
  • Be aware of false information—try this: ask it to write a resume for you and check how many things just aren’t true (people who’ve done this have reported that ChatGPT lists the titles of non-existent books they’ve supposedly written!). Don’t trust the information given—verify, fact-check, do what you have to do to confirm the information
  • Be aware of the base date of the information provided—from all reports, ChatGPT’s ‘database’ of information only goes to 2021, so any discoveries etc. after this date are not ‘known’ by the tool
  • Be aware that, at this stage, ChatGPT reportedly doesn’t do internet searches and does NOT cite sources for the information it provides, which means it’s impossible to independently verify its results yourself

For some good summaries of what ChatGPT can and can’t do, check out these resources:

[Links last checked January 2023]

3 comments

  1. This is a really thorough and useful summary, Rhonda, and thank you for covering so much information in such a short time.


  2. About other languages — yes, the bot apparently translates English to French and back very well. On the tests I tried in Hebrew, the results were all miserable.

    This, then, also speaks to the comment about bias.


  3. […] the AI as well. These tools can only build on what they’ve been fed (programmed with) and, as Rhonda Bracey explains, we don’t know what that was but we might assume it was largely white middle-class male […]



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.

%d bloggers like this: