Since I can’t really enter into meaningful dialogue with ChatGPT. I need to draw upon its analytical ability and then use that for self-reflection. So my plan is to draw upon frameworks of unseen observation and narrative journaling I want to create a GPT that can:
• Analyse a transcript of my lesson in relation to stated lesson and self- reflective aims and produces a reflective diary entry.
• Compares my own (hot) evaluation of a lesson to the journal and highlights similarities and differences
• Provides me with reflective and mentoring questions
At this stage the analysis is the main focus, once I am happy with that I will work on the other two. I’m of course making the assumption that ChatGPT has been fed enough data about education and ELT for it to be able to analyse things with some knowledge. If not then the experiment dies as I can’t imagine the many methodology book writers being thrilled at my asking them if I can put their work into the AI!
Other assumptions I am making, I know that Gen AI won’t actually understand the transcript I give it. Many people are fooled into thinking Gen AI models understand the content that they use and generate, but they don’t. Equally, I need to be wary of hallucinations, those times when the GPT simply makes stuff up. One consideration this also raises is the fact that it is very likely that if I gave the GPT the same transcript and prompt a number of times, the output is likely to be different. From the mentoring aspect, I made rule that I had to use the first one I produced for the mentoring though as part of the overall experiment I might produce more outputs.
This an important consideration as Gen AI can remember and even build on previous interactions and thus many outputs on the same task could skew potential patterns in my teaching.
To collect the data, I use an audio record the lesson on my iPhone, as a podcaster I have some mics that can be used with the phone which should ensure clarity. I have the students’ permission to record and I explained to them what I’ll be doing with the recordings. From an ethical point of view, I also want to ensure that no student data goes into ChatGPT so I will remove names from the transcript.
To get the transcript, I use MacWhisper as this app means any data is simply on my computer. The only edits I make to the script are those that remove student identifiers. I then put the script in a word.doc. If you’re interested a lesson transcript is around 11 to 13 pages long ( it’s a 90-minute lesson)
Now I just need to get ChatGPT to analyse it.
In the couple of week’s leading up term, I started playing with prompts. There is plenty written about good prompting. I prefer the simple context, task, output approach . On their academic skills page the University of Newcastle refer to prompting an iterative process which I very much agree with, a prompt needs honing. So for the initial prompt I did some trial and error using lesson found on YouTube.
In terms of context, the prompt would need to know a bit about me, and the class. In their framework, Anderson and Kamili suggest setting the backstory, which in this would be the context of the prompt, with questions such as:
What are some of the major event so of your teaching career?
How would you characterise yourself as a teacher
What is your personal history with the class?
Using those as a basis gives this context:
“You are an English for academic purposes teaching as university level pre-sessional class. Though you have thirty years of experience in teaching ,this particular class type is new to you. Your experience as a teacher and teacher trainer means that while you plan lesson thoroughly, you don’t mind going with the flow if a teaching opportunity arises. This is the first time you are teaching the class and you want help reflecting on your lessons. As you have 30 years of experience the reflection should be focused on more advanced teaching matters. The students in the class are multinational adults and likely post-graduates from universities in their home countries. They come from a number of different educational backgrounds but all have a first degree and should be around IELTS 6.5 level. They are taking the class to prepare themselves for likely master, and doctoral studies either at oxford or another UK university.”
Now for the task – the reflective journal. Generalising from self-observation cycle, there is a stage prior to observation in which the observer and the collaborator meet to discuss the lesson so to replicate this I used this in the first prompt for this in the first prompt.
1. Ask what the focus of the lesson was.
2. There should also be a self-observation goal. Ask what that goal was.
Next we need the main task itself and the output.
“Once you have all these pieces of information ask for the transcript to be uploaded. A transcript of the lesson will be uploaded. Once the transcript has been uploaded,
Based on all the information the transcript create an authentic looking diary entry telling the story of the lesson.”
When you put all that in chatGPT, this is the response:
ChatGPT: “That sounds like a really rich context for reflection! Let’s start gathering the essential information. To begin, please tell me:
1. What was the focus of the lesson? (For example, were you focusing on academic writing skills, seminar discussion strategies, critical reading, note-taking, vocabulary development, or something else?)
2. What was your self-observation goal? (This could be a specific aspect of your teaching that you wanted to observe and reflect on, such as your use of questioning, how you manage transitions between activities, your responsiveness to students’ needs, your time management, how you scaffold tasks, or how you deal with cultural differences in communication styles.)
Once you’ve shared these details, please upload the transcript of the lesson. Once I have that, I’ll craft an authentic, reflective diary entry narrating the experience of the lesson.
Looking forward to hearing more about it! “
And in the next part I’ll show you what happened.