Blogging With AI

Is ChatGPT a useful tool for writing engaging blog posts?

This post (written by hand, with no artificial intelligence involved) details my attempts to find out.

The Challenge?

Recently, I wrote what I thought was a considered response to a post on the JISC External Examiners mailing list, although for some reason the post was not approved. The post was captured by the person who initiated the discussion (who replied copying it on the list). I decided that the post would also make a good blog post, but needed more context adding.

My original discussion contribution discussed the launch of ChatGPT and the way it can be misused by students to write their assignments, an area I have been commenting on in the media.

I decided to see if I could use ChatGPT to expand on the ideas, whilst maintaining my own writing style.

Using ChatGPT

I loaded the original post into ChatGPT and asked it to provide a structure for the new blog post. I then asked it to expand the structure into a post.

The initial prompt I gave was:

Read the following and suggest how I can turn it into a better structured and more interesting blog post. The new post should not have any inaccuracies introduced into it.

The problem was that the process produced well-written text, but every time I tried to refine the post further, the text deviated further away from my writing style and my original ideas.

The chatbot was very reluctant to mention ChatGPT at all, moving instead to talk about AI generated text. It removed my specific examples to talk in very general terms. It changed the terminology to move away from “academic integrity” to “academic honesty“. It changed UK English to US English. It repeated ideas and text. And it overinflated the strengths of AI writing, at the expense of many of the challenges that academic had to address.

In short, the blog post became depersonalised and stopped sounding like me.

This may be all well and good for some blog types, where the volume of content is more important that the quality of the content, but not for this blog.

I tried several methods to keep the personalisation, but I just could not get any of them to stick.

The End Result

So, was the final output from ChatGPT of any use?

It was, to some extent. The chatbot came up with a good title. It came up with a useful structure. And it generated some interesting image creation prompts, although in the end, I did not use them.

The final result I ended up with was something of a hybrid post. I cut down the generated version, removed the repetition, changed the language to sound more like me, then added back some of my own ideas and examples.

I suspect I still saved time and am happy with the end result, but it is still a different post to the one I would have written without any external input.

You can see the end result as a separate blog post here.

Hybrid Writing

An interesting ethical question to consider. Is writing and publishing in this way ethically dishonest? There are certainly words and ideas in the linked blog post that I did not directly write, even though I went through levels of prompt engineering, refinement and editing. There are also ideas that could not have been there had I not fed the initial discussion board post into ChatGPT.

Would I write posts like this again? I have a huge quantity of forum posts, Quora contributions, transcripts from talks, lecture recordings and the like, which could easily be polished up into interesting blog posts, all based on my original words and ideas. Is using ChatGPT then just like using a human editor, or does this use need to be formally declared?

I mention this as a parallel question to the one that universities have to be having about student use of AI writing tools in their own institutions. The necessary discussion is both nuanced and complex. My blog is in some ways a promotional vehicle, although I pride myself on sharing interesting and high quality information. I believe my use is acceptable. What would happen if a student generated text build on their own ideas, then edited it? What would happen if they had few ideas of their own and used ChatGPT to kick off the writing process? Where are the boundaries?

There are interesting times ahead.