Ever since large language models (LLMs) such as ChatGPT have become widely accessible, it has been inevitable that they will impact education. Many commentators have focused on concerns about students using such tools to ‘cheat’ on their homework and unfortunately neither humans nor AI are very good at detecting AI-produced work. But what about the opportunities that AI tools can bring to teachers? LLMs have been shown to improve consultants’ efficiency and effectiveness; what might this look like in the classroom?
Finding out more from Ankita Sengupta
To gain some insights into practical applications of AI in the classroom, we spoke to Teaching and Learning lead Ankita Sengupta. Read on to find out more about how Ankita uses AI:
- To provide feedback on her own teaching
- To improve existing resources
- To support marking and feedback
AI as a coach to provide feedback
Ankita explained that one of the things she most missed from teacher training was the regular lesson observations and feedback provided by her mentor. She now uses AI to fill that gap, but without the associated pressure of being observed. She chooses a specific part of a lesson and a particular focus, records the audio on her phone, uses an AI tool to transcribe the audio and then adds the transcript to AI with a prompt to give feedback. To ensure that the feedback is useful, the AI is prompted to refer to a specific teaching strategy. For example, she recently wanted to focus on her use of Cold-Calling in the classroom, so she provided the AI with the following prompt:
“You are an experienced Chemistry teacher who is knowledgeable on educational research and pedagogy, particularly the research on using cold calling for inclusive classrooms as discussed in this blog post. You will analyse the lesson in the transcript attached and then give constructive feedback to improve cold-calling strategies for student motivation, engagement and progress. You understand the importance of feedback being actionable and easy to implement whilst also being aligned with the existing teaching style and content being delivered in the lesson.“
“ChatGPT did conjure up a phantom pupil, but its coaching still landed as it spotted that my pose → pause → pounce → bounce routine was more pose → pounce. So my focus for next lesson will be to add in more of a pause and bounce to another student to ensure that every voice gets some airtime.”
Ankita was quick to point out that the students were made aware that she would be recording part of the lesson and how the recording would be used. As only student first names were used in the lesson, she wasn’t concerned about giving personal data to an AI but nevertheless, she urges anyone wanting to try this in their own classroom to first check their school’s AI policy.
Using AI to improve lesson resources
ChatGPT has been shown to support teachers with lesson planning but, equally, AI lesson plans can contain factual errors or promote outdated ideas such as lessons centred around learning styles.
As an experienced teacher, Ankita doesn’t feel the need to ask ChatGPT to plan lessons from scratch, and she is also reluctant to abandon the resources she has built up over her years in the job. Instead, she uses AI to adapt her existing resources to tighten alignment with particular pedagogical principles. For example, she recently asked ChatGPT to critique a resource based on Rosenshine’s second principle to present material in small steps. ChatGPT recommended adding in a worked example at the start, a framework for students to refer back to and a list of common mistakes.
“The tweaks were easy to slot in and viewed through a student’s eyes, blindingly obvious. Why didn’t I spot them sooner?“
Using AI to support marking and feedback
Marking and giving feedback on student work takes up a great deal of teacher time and is hence a prime target for the efficiencies that AI can bring. AI tools have been shown to give feedback that is as good as human assessors, but there remain problems with AI hallucinations, meaning that any feedback needs to be carefully checked by an expert teacher.
Ankita uses ChatGPT to speed up her marking and feedback in several ways:
- To transcribe hard-to-read student handwriting
- To provide suggested sentences to help students to improve their work (rather than just provide the correct answer)
- To produce low, middle and top mark answers for students to critique as part of a feedback session.
Once again, Ankita was careful to point out that she let students know she was going to be using AI to support her marking in this way and didn’t feed any student’s personal data into the AI.
Generally, Ankita only uses AI to support her marking where it is going to have the most impact, in her case extended response questions, which are the most time-consuming to assess. She emphasised the importance of training the AI carefully by feeding it a specific marking rubric, testing the AI on a few scripts she has marked herself and re-prompting the AI where it falls out of tolerance. Only then does she allow the AI to mark the remaining scripts.
“I like to think of AI marking as an eager PA who starts with zero Chemistry teaching expertise. I give it a quick quality-control briefing and it clears the bulk of the marking grind, freeing me for the feedback which moves learning forward”
Final thoughts
AI has the potential to support teachers to be both more efficient and more effective at their jobs, however, careful thought is needed to decide which processes AI can help with and which are best left to human experts. As Ankita’s examples show us, one way to do this is to feed the AI evidence-informed pedagogical principles and ask it to provide analysis, critique and feedback. It is then up to the teacher as the experienced practitioner to decide whether or not to act on this feedback.
A huge thank you to Ankita for sharing her expertise with us! If you want to find out more about using AI as a teacher, then do check out our Guide to AI in Education or enquire about our AI workshops.
