By Dr Hamideh Hamdi Khosroshahi
Feedback Without a Staffroom
Reliable feedback has always been essential for any teacher’s professional development.
In the 1980s, Donald Schön introduced the term “reflective practitioner” to highlight the teacher’s desire to see themselves from the outside so they can grow on the inside. Over the years many new methods have been introduced for reflective practice, including peer-observation frameworks, diary studies, post-lesson reviews, and recorded micro-teaching sessions.
However, the traditional teaching model seems to apply to fewer and fewer teachers nowadays, in part thanks to the pandemic and the upsurge in new technology. Those teachers working in micro-schools or private online setups have fallen through the gaps of reflective practice, and have been left isolated.
This could be where technology comes to the rescue. Tools like TeachFX and Project CAFE now listen in (with our permission) and show us when we’re hogging the mic, when students have gone quiet, or when a brilliant question sailed past unused. UNESCO’s 2024 brief is excited about the promise offered by such tools — AI as a pocket mentor — but also waves a red flag about data privacy and hidden bias. The researchers end, as always, with “More classroom trials, please!” We still need to know whether these dashboards actually change teaching, whether busy teachers can or will use them, and what ethical potholes might be lying in wait.
And while the academics debate, a huge crowd of freelancers and private tutors end each lesson, shut the laptop, and stare at a lonely screen — no staff room banter, no colleague to whisper, “Hey, you talked a bit much today.”
Putting AI in the Coach’s Chair
It was late on a Friday in March. I completed my sixth one-to-one of the day, drained from a forty-minute struggle with the present perfect. The student had logged off, saying I have gone. I felt a combination of fatigue and self-doubt. I went to ChatGPT. I remember pasting in my lesson notes and typing: “Acting as my mentor, list what went well and what I could improve.”
I expected bland praise from YES MAN ChatGPT. Instead, I read that I had rephrased the learner’s ideas many times instead of promoting self-correction. It added that I needed to consider pausing longer after an error. This was not a senior teacher, an observing friend; it was a chatbot that made me laugh and then reopen the plan for the next day.
During the same experiment, I did not give ChatGPT a recording of my lesson but only the skeleton of my lesson: the activity sequence, the worksheet, and some of the questions I had planned to ask. I asked ChatGPT to comment on pacing and questioning, and only then did I learn that I had asked twelve closed questions in a row and had to mix more open prompts to push the learner to use the grammar for themselves.
It wasn’t all glowingly useful, though. Some guidance was far from suitable. For example, it suggested using puppets for a lesson with a trained engineer, along with a five-minute dance break.
I have a postgraduate degree and an extensive teaching record, and a research background; however, I realized even the best teachers still need self-reflection. It had been a long and exhausting Friday, and all I wanted was to finish the day, but the AI feedback deflated my arrogance and made me more mindful of my own teaching practices. In the following sessions, I was more aware of what I was doing rather than only focusing on my learner.
The days I feel extreme fatigue, I picture ChatGPT raising its virtual eyebrow and muttering to me to use my puppets.
Crafting a routine with four trusty prompts
Within two weeks, my debriefs settled into a rhythm. First: the lesson debug. I copied the outline, added a few student quotes, and dared the AI to name three strengths and three fixes. The model framed feedback as questions like: “How would you tighten your instruction?” rather than “Yikes, you babble!” I felt provoked rather than scolded, and motivated to make some tweaks. Next: detecting biases. Awkwardly, I later realized that many of my role-play scenarios included male managers. I asked the bot to scan my examples for gender or cultural biases, and it immediately flagged the imbalance and offered practical female-led cases.
Whenever tomorrow’s schedule included mixed-ability mayhem, I employed my scaffold booster prompt. I told my loyal buddy, “I’ve got one B1 who learns best with visuals and one C1 who is obsessed with analyzing everything — same reading text, help me out.” Seconds later, the idea emerged: “Color-code the key points for the B1 — green for ideas, amber for details, red for examples — while C1 detects cohesive links like however and therefore.” It further guided me to ask both learners to link those key phrases with C1’s linking words to build a roadmap of the text.
Of course, I added one last layer: emotional debrief. I prompted the model to ask some questions about my feelings during the lesson. “What did the learners do today that made you smile, and how can you create those moments again?” The question hit home. I remembered the moment my four teens had burst into laughter when I made a role-play twist, so I knew inserting surprise tiny “role-play” surprises would cause a shared joy in lessons.
Why it matters for the isolated teacher — and for everyone else
As a mentor, AI can be beneficial for freelancers who teach their lessons alone, in a kind of vacuum, without peers to come in and observe or simply to bounce ideas off. The scope can extend further and include corporate trainers who can review their work using their phones as their mentors. Novice teachers can prepare for a demo lesson by pasting their outline and receiving an instant second opinion beforehand. It can help those who, for one reason or another, lack confidence in the grammar they are teaching, by providing clarification before the lesson begins. Finally, teachers with intense schedules can use AI for self-reflection without needing to find another human to take part – and, importantly, without having to hope that this other human happens to be free at the same time as the teacher is.
Limitations, blind spots, and comic errors
Just as there is no perfect mirror, my algorithmic confidante can also be hilariously wrong. There are times when my buddy falls into the never-ending generational cycle of providing feedback like: “Excellent lesson!” — until I push: “Be concrete. Quote lines. Time my wait-time.” There can also be cultural fails. Once, it rewrote a local retail case study into a Silicon-Valley fantasy that ignored European labor laws. There is also privacy: I anonymize transcripts and strip company names before passing the information through the platform, which can be time consuming. Despite its capabilities, ChatGPT can never replace the human ear when exhaustion signals burnout rather than poor planning.
A sample AI-guided reflection in action
The other day I taught a forty-minute lesson on modal verbs to a trio of B1 logistics managers. After class I copied my plan and the first ten minutes of the transcript into the model, asking for feedback on clarity, interaction, and next-step tasks. I received applause for my student-generated examples, but then was asked, “At which point did you hand control of modals from yourself to the learners?” Fair point — I noticed I never did and had kept control the whole time.
My model friend proposed an instant fix: invite each student to write a brief note about a real shipping problem. They swap notes and rewrite the duties using a range of modals — must, have to, need to, and don’t have to. Then, in a third round, a “compliance officer” rewrites the duties as a formal email. Same grammar, real content, and more student talk, so I implemented it the next lesson. Participation jumped, the mirror had pushed me to let go earlier, and the class blossomed.
Ethical side notes worth our vigilance
Three concerns surface repeatedly.
The first is data security. Although transcripts contain no bank details, they do hold voices, names, and sometimes sensitive demographics. I now abbreviate names to Student A or B, remove company identifiers, and store original files offline.
The second concern is emotional substitution. AI can ask smart questions, but it cannot notice when I slouch in the chair, nor can it suggest a walk or a therapist. A teacher’s wellbeing still needs human eyes, a colleague’s laugh, or a supervisor’s phone call.
The third is… well. I worry sometimes that all this data that I am feeding the AI is training it to take my place. It learns so quickly, and I give it so much data, that I have to decide if the tradeoff is worth it – on the one hand, my teaching is getting better thanks to this AI-supported reflection; on the other, am I reflecting myself out of a job five years down the line?
Looking ahead: toward a teacher-coach console
In the near future mainstream learning-management systems will include a tab for reflection. Teachers will upload a Zoom recording, click “Analyze,” and receive a color-coded heat map of turn lengths, student-talk time, and wait-time. There will be a prompt bank that can be calibrated to their lessons. Until then, ChatGPT is my mirror – often warped, often flattering, but far more revealing than the black mirror of my empty Zoom meeting when the students have all departed. For the deserted teacher, finishing the day with a dialogue means more than ending it quietly.
References
Petrilli, M.J., (2024). Next-Gen Classroom Observations, Powered by AI: Let’s go to the videotape to improve instruction and classroom practice. Education Next, 24(4).
Schön, D.A. (1983) The Reflective Practitioner: How Professionals Think in Action. New York: Basic Books.
UNESCO. (2021). Artificial Intelligence and Education: Guidance for Policy-Makers.
Author Biography
Hamideh Hamdi Khosroshahi has a Ph.D. in Teaching English as a Foreign Language. With more than ten years of teaching experience in schools, universities, and refugee education programs around the world, her research looks into how language, identity, and inclusive teaching come together to empower both teachers and students. By combining hands-on teaching with academic research, she promotes more inclusive methods of teaching English.
