Adventures in Technology Enhanced Learning @ UoP

Tag: AI Tools

Guest Blogger: Tom Langston – How ChatGPT Revolutionized My Writing Process – The AI version

(Co-writer: ChatGPT)

This version was generated after feeding in my attempt at writing this blog post and asking ChatGPT to ‘shorten this text into a blog post’.

I recently discovered the incredible potential of ChatGPT, and it completely revolutionized the way I approached my writing. As a student pursuing a Professional Doctorate in Education, I often struggled with crafting cohesive and focused assignments. My writing style was scattered and verbose, making it difficult for readers to follow my train of thought. Enter ChatGPT, a powerful AI language model that became my secret weapon in developing a well-structured research proposal.

Initially, I sought human help from my project supervisor, Stephen Webb, who pointed out that my writing relied too heavily on vague terms like “this” and “it.” While I understood the meaning behind my words, Stephen reminded me that readers might not share the same clarity. Armed with his feedback, I set out to improve my proposal.

When I completed the next draft, I thought I was finished. However, the word count was far too high, so I turned to ChatGPT for assistance in condensing my writing. It did the job, but the result lacked my personal touch and voice as a writer. I realized I would have to manually trim down the content to achieve the desired outcome.

During this editing process, I discovered the true power of ChatGPT. I started using it to check sentences that I struggled to articulate clearly. When I asked for grammar assistance, ChatGPT not only provided corrections but also suggested ways to enhance sentence structure and improve readability. This iterative process allowed me to refine my writing, discarding extraneous thoughts and rearranging sentences for better flow. It became evident that if a particular text no longer fit within the new leaner version, despite my attachment to it, it had to be eliminated. This experience epitomized the essence of editing.

As I continued to utilize ChatGPT, I encountered questions of ethics. While it proved immensely useful in reducing my word count, I couldn’t help but feel uneasy about presenting the AI-generated content as my own. The ethical implications of using AI in writing remain a divisive topic, with differing perspectives on its appropriateness and boundaries. To navigate this issue, some suggest referencing AI tools in an appendix, detailing their specific use and purpose in the writing process. However, this raises questions about the extent to which we should acknowledge AI contributions and how that compares to other forms of assistance, such as spell checkers or language tools like Grammarly.

One important consideration is whether AI tools are always correct. While they excel at constructing language, they may strip away the author’s voice and inadvertently introduce errors. In my experience, Grammarly sometimes made corrections to the text generated by ChatGPT, highlighting the need for human judgment in refining the output.

In conclusion, ChatGPT became a game-changer for me, transforming my writing process in ways I had never imagined. I believe it should be embraced as a valuable resource for students to tighten their drafts before seeking extensive academic support. By incorporating AI tools, we can alleviate the burden on educators, who can then focus on reviewing more polished submissions requiring minimal intervention. However, it is crucial to strike a balance and ensure that AI enhances our own voices rather than overshadowing them. AI is undoubtedly here to stay, and by harnessing its power thoughtfully, we can elevate our writing while preserving our unique identities as authors.

Thoughts from the Proofreader:

I was asked to proofread Tom’s blog post here and have a look at ChatGPT’s version too. Comparing the two, I definitely found Tom’s writing much more personable – I could imagine him speaking as I read through it. ChatGPT’s version presented the information efficiently and effectively, and I found nothing to correct. However, in its response to the prompt, it had taken away everything that made Tom sound like Tom!

This experience also really made me realise how many misconceptions I have about AI tools. I’d assumed that ChatGPT would have a very grammatical approach to its interpretation of language – rather like I’ve been told I have! However, when Tom asked it about the difference between ‘that allowed’ and ‘allowing’, ChatGPT talked with authority about implications and potential meanings. This answer was a long way from my interpretation, which attributed the difference to the grammar of relative clauses (X refers to one thing, Y refers to another). As Tom demonstrated with his irony example, it’s worth being cautious with how far we trust its responses. And I think we can be confident that human input will still be needed for a few years (or at least months) yet. 

Credit Image: Photo by Bram Naus on Unsplash

Guest Blogger: Tom Langston – Last Night ChatGPT Saved My Life…

The classic line from Indeep’s 1982 post-disco hit “Last Night a DJ Saved My Life” serves as the basis of my title. As ChatGPT did – not literally, but figuratively – save my life.

I am currently studying for the university’s Professional Doctorate in Education and, since February, have been completing my final taught assignment entitled “Research Proposal”.

It takes me a while to write, as I don’t find it easy, and I am not that good at it. My approach to writing is best described as little and often, frequently getting distracted and losing focus on the task at hand. If you have ever seen Dave Gorman’s “Googlewhack Adventure”, you will understand my problem: his book and stage show outlined how he was meant to write a fictional novel and ended up writing and performing about his experiences of Googlewhacks (Googlewhacking?) instead. He got distracted by less important but much more fun endeavours. 

The other problem is that I don’t get to the point. I write a verbose mess of unconnected thoughts. So with that in mind, I am going to explain how the rise of the machines (we are not far away from ChatGPT working out it wants to be Skynet) assisted me in writing my “research proposal” assignment. 

First Drafts

To start with, I had human help. Stephen Webb, as my project supervisor, read a draft I had written. He demonstrated that a lot of my writing relied on small words to, as he put it “do the heavy lifting”. Using ‘this…’ and ‘it…’ frequently because I understood what I was trying to say –  he pointed out that the reader might not. The only reason he could even start to understand my work was because he knew me and the context of my subject. 

From his extensive feedback, I redrafted, edited and tried to improve my proposal. 

After completing the next draft, I thought I’m done. However, I was well over the word count, so I put sections into ChatGPT and told it to re-write it within my word limit. It only bloomin’ did it! The problem was it was very sterile; it lost all sense of who I am as a writer. This meant I was going to have to manually get the word count down. 

After another version, I asked Stephen for some clarification on some of his earlier feedback. His reply was:

“In the section you sent through, there are still individual sentences that are hard to parse. The initial word of a sentence or clause primes the reader to think in a certain way; if the sentence or clause doesn’t deliver, then confusion arises. You really should pass this through a grammar checker, Tom.”

Not being a confident writer, I already knew that my writing needed work. What this feedback demonstrated to me was that I lacked cohesion and structure. My train of thought constantly jumping around the entire document is a by-product, I guess, of writing in such short unfocussed bursts. 

I wrote a heading titled “Introduction” but did not actually introduce anything. As Stephen had pointed out, I had poorly communicated my narrative. What this meant was that readers were getting lost in my disconnected thoughts. I could now see that certain sections felt shoe-horned in because I liked what I had written. Realistically, those “gems” made no sense to the reader as they were not connected with the text surrounding them. 

I tried to use ChatGPT to see if it could make sense of it, but there was too much there for it to tell me how to rearrange it. The system did not accept the entirety of the text and could not provide clarity to the sections I had written. ChatGPT could not offer clear guidance for how best to rework and rearrange it.

A Quick Aside

At this point, on re-reading the above section, I think it nicely illustrates the problem I was talking about. I have said the same thing here in three different ways. 

  1. I tried to use ChatGPT to see if it could make sense of it, but there was too much there for it to tell me how to rearrange it. 
  2. The system did not accept the entirety of the text and could not provide clarity to the sections I had written. 
  3. ChatGPT could not offer clear guidance for how best to rework and rearrange it.

I only need to use one. In the past, Stephen had raised this issue with me, but it is a hard thing to change. What I am going to discuss next is my recent experiences of using ChatGPT as a tool to develop the parts in my writing I would not have picked up on myself. Heck, as I go back through this blog, I’m sure I’ll find more. I just won’t mention anything I’ve changed, so you’ll never know… 

Editing with ChatGPT

As I worked my way through the editing process, I started to use ChatGPT to check sentences I wanted to keep but was struggling to make coherent. They were either too long, or the idea I was trying to convey was not coming through clearly enough. 

An example of this would be when I asked ChatGPT to “check the grammar”:

My original question:

My situation should be made clear to the participants, allowing them to have fully informed consent, but to also highlight that if they perceive issues in the current hybrid environments I am hoping to act on their concerns and not silence their voices.

ChatGPTs reply:

The grammar of the provided text is correct. However, there are a few suggestions for improving clarity and readability:

“My situation should be made clear to the participants, enabling them to provide fully informed consent. Moreover, it is important to emphasise that if they perceive any issues in the current hybrid environments, I am committed to addressing their concerns and not silencing their voices.”

In this revised version, I made a few adjustments to enhance the sentence structure and clarify the message.

It was this repeated process that helped me (despite, at times, slowing me down) as I kept re-reading the submission each time, trying to tighten up my writing. Finding new thoughts that were slightly out and attempting to move them to the correct place. It became a good rule of thumb that, if the text I wrote did not fit in the new, leaner version, despite how much I wanted to keep it, it was gone. (Well done Tom – you just discovered the true meaning of editing!)

Below is an example of how I started to use ChatGPT to my advantage so I could improve my writing. 

My question to ChatGPT:

Should it be “that allowed” or “allowing” in the following text – 

Before the pandemic only the most confident teaching staff employed interactive tools (Nearpod, Mentimeter, etc.) within a lecture space that allowed students the ability to creatively engage with the material (Tan et al., 2020).

ChatGPTs reply:

Both “that allowed” and “allowing” are grammatically correct in the given sentence. However, the choice between them depends on the intended meaning and emphasis.

“That allowed”

Using “that allowed” suggests that the interactive tools were specifically designed or configured in a way that granted students the ability to creatively engage with the material. It implies that the tools had a built-in functionality or feature enabling such engagement.

“Allowing”

Using “allowing” suggests that the interactive tools, in a general sense, provided the opportunity or capability for students to creatively engage with the material. It implies that the tools had the potential or capacity to facilitate such engagement.

Both options are valid, but if you want to emphasize that the interactive tools had a specific functionality designed to enable creative engagement, “that allowed” would be more suitable. If you want to convey a more general sense of opportunity or capability, “allowing” would be a better choice.

As you can see, ChatGPT helped to explain how the slight difference in the sentence structure changed the meaning of what I was writing. I was able to consider what it was I’m trying to say and use the most appropriate version. 

Through all my previous doctoral assignments, I had a different tutor who helped me with my drafts. They told me they liked my writing style, which gave me the impression I was good at it. I can’t say what they specifically liked about my writing. I can make an educated guess that, because it was a level 8 assessment, my mistakes were forgiven. The marker was able to interpret the little words “that are doing the heavy lifting” (as Stephen had pointed out) more easily than your ‘average’ reader. Stephen helped me understand that it is one thing to have an interesting voice running through your text but is quite something else to actually be good at writing. 

The ethics of using AI

When I got ChatGPT to reduce my word count, I spoke with Stephen about the ethics of it and (ignoring the sanitisation of the output) it felt like a great way for a writer to do this. However, it felt wrong to take what ChatGPT had created as my own.

There is going to be an ethical divide between those who see using any form of AI as wrong and those who see it as a tool for improvement. The problem (as always) lies in the grey area and the boundaries of where people choose to deploy it, for example how far the tool is shaping their work beyond what would have been possible by them alone. While knowing it might be unethical, some will use it due to other commitments (work, family, etc). This scenario is a foreseeable eventuality, much like those who copied other work or paid essay mills for their work. But perhaps AI may feel slightly more ethical? As I am about to discuss, maybe a strong referencing code is what is required. But people (I was going to put students, but felt that unfair as we all have a personal line we move and adjust depending on the subject) will always push and flex the boundaries of fairness.

Referencing AI

In a recent ALT mailing list post, the question was asked about referencing when AI was used to  support work. The reply pointed to Boston University’s faculty of computing and data science guidance “Using generative AI in coursework”. The post highlighted this text:

“When using AI tools on assignments, add an appendix showing

(a) the entire exchange, highlighting the most relevant sections;

(b) a description of precisely which AI tools were used (e.g. ChatGPT private subscription version or DALL-E free version),

(c) an explanation of how the AI tools were used (e.g. to generate ideas, turns of phrase, elements of text, long stretches of text, lines of argument, pieces of evidence, maps of conceptual territory, illustrations of key concepts, etc.);

(d) an account of why AI tools were used (e.g. to save time, to surmount writer’s block, to stimulate thinking, to handle mounting stress, to clarify prose, to translate text, to experiment for fun, etc.).”

This almost feels as if there are too many barriers to use for a subject (or level) where you may be writing assignments of 6000+ words. Especially if you have to reference every section of text AI helped with. The follow-up question is then do you have to highlight every time spellcheck made a correction? Do you have to disclose that you used a service like Grammarly to make minor amendments? That is of course hyperbole. I think that if you get AI to generate starting paragraphs or change substantial amounts of your work, yes, this should be explained in an appendix. For re-writing a sentence and making grammatical choices, then no. That raises another question, how many sentences constitute a substantive change? 

Are AI tools always correct?

The final point relates to whether they are always correct. Now I know these tools are better than me at the construction of language. I have always been better at face-to-face conversations as a form of communication. During a conversation I can see when people don’t understand me, so adapt how and what I say accordingly. Writing does not afford the writer or reader the same luxury. So why do I ask if they are correct? As I have already mentioned, ChatGPT sterilised my voice. When asked to reduce the word count, it did what it was supposed to do but it strips out the humanity of the author (unsurprising being that it is a computer).

Another example (inside a paragraph that starts “The final point”, I do get the irony) is that Grammarly made corrections to the generated text from ChatGPT. 

This is a screenshot of a reply from ChatGPT placed within a box saying 'Both "that allowed" and "allowing" are grammatically correct in the given sentence. This box appears on the lefthand side of the screen on the righthand side is a box created by Grammarly. Grammarly has written Formality, then says to engage with the material creatively. The to-infinitive to engage has been split by the modifier creatively. Avoiding split infinitives can help your writing sounds more formal.
Another “another example” of a final point is: I put the above paragraph into ChatGPT to ask if I had used irony correctly. This is what it said:

Yes, the use of the word “irony” in this context is correct. Irony is a literary device that involves a contrast between expectations and reality. In your statement, you are highlighting the irony in two situations:

The first example of irony is that despite recognizing that language generation tools like ChatGPT are better than you at constructing language, you also acknowledge that you excel at face-to-face conversations where you can adapt and clarify your communication based on people’s reactions. The irony lies in the contrast between the effectiveness of these tools in writing and your personal strength in face-to-face communication.

The second example of irony is that Grammarly, a tool designed to enhance writing and make corrections, ends up making corrections to the text generated by ChatGPT. Here, the irony arises from the fact that a tool meant to improve the quality of the text is itself making changes to the output of another tool that is supposed to assist in generating text.

Overall, both examples demonstrate a contrast between expectations and reality, highlighting the ironic situations in your statement.

The system had not understood how I had used irony in the sentence but said that the overarching point I was making was ironic.

This image is of a person standing, with their eyes focusing as if they're looking at someone else who might be sitting down, because their chin is tilted down a little. The wording across the top of the image reads: I am still none the wiser if I used irony correctly in the parenthesis when I keep adding examples to a final point. The wording along the bottom of the image reads: And at this point I'm too afraid to ask.

Conclusion

In conclusion, ChatGPT ‘saved my life’ by allowing me to interrogate my writing in a way I have never been able to before. We should promote ChatGPT to students as a resource that can help tighten up their drafts before needing academic support. It should be utilised to alleviate the burden on academics, who are often asked to critique early drafts of work. Academics should hopefully then only see much more polished versions of submissions that require less input from them. 

As a final example, I didn’t like my last sentence. ChatGPT gave me another version:

It should be used to relieve the burden on academics who are frequently asked to critique early drafts. Ideally, academics would primarily review more polished submissions that require minimal input.

I didn’t like that version either. Maybe then I should just delete it? That is where the human side wins out over AI, your sense of self and how you want to sound when writing.

AI is here to stay (and take over the world if the Terminator documentaries are anything to go by), but actually, that is alright (AI being here as a tool, not taking over the world). At levels 4 and 5, we need to find creative ways to incorporate AI tools into the assessment process (which may mean fewer essays). Levels 6, 7 and 8, it’s about using AI to help improve our own voice but not lose it. 

The ChatGPT re-write: How ChatGPT Revolutionized My Writing Process – The AI version.

Credit Image: Photo by rupixen.com on Unsplash

Image in the text created by T. Langston using imgflip.com

© 2024 Tel Tales

Theme by Anders NorénUp ↑