HANNAH OWEN, CLASS TEACHER, RIVERSIDE PRIMARY SCHOOL; PHD RESEARCHER, UNIVERSITY OF CAMBRIDGE, UK
My first experience of using artificial intelligence (AI) almost crept up on me. I was aware of some of the AI tools available, particularly ChatGPT, but had yet to try them out. Yet one morning, as I was preparing a spelling activity for a group of children, I realised that instead of spending the next 10 minutes writing a short paragraph that contained words from the spelling pattern that we were learning, I could generate the text using AI.
It appears that I am not alone. A recent study published by the Education Endowment Foundation (Roy et al., 2024) found that in November 2023, 42 per cent of teachers reported using AI tools to help with schoolwork. In fact, the number of teachers using AI increased from 35 per cent of teachers in August 2023 (Fletcher-Wood, 2023) to 57 per cent of teachers in August 2024 (Hallahan, 2024). Given the success of my own early attempts at using generative AI (GenAI), I am interested in investigating how AI tools can support the work of teachers in planning and preparing lessons.
A Department for EducationThe ministerial department responsible for children’s services and education in England policy paper on GenAI in education (DfEDepartment for Education - a ministerial department responsible for children’s services and education in England, 2023) suggests that AI tools have the potential to reduce workload across the sector, thus freeing up teachers’ time and allowing them to focus on teaching. Indeed, the trial commissioned by the EFF (Roy et al., 2024) sought to examine the effect on teacher workload of using ChatGPT for lesson and resource preparation (LRP), in comparison to planning lessons without using any form of GenAI. They concluded that the teachers who used ChatGPT for lesson and resource preparation saved 25.3 minutes per week on average (Roy et al., 2024). However, the proportion of lessons planned using ChatGPT was significantly lower in weeks six to 10 than in weeks one to five, which suggests that the use of ChatGPT declined over the trial period. According to the data, the proportion of lessons planned using ChatGPT decreased from 39 per cent to 29 per cent across the two blocks. So, I’m curious: if using GenAI saved teachers time, as is claimed, then why did its use decline over time (Roy et al., 2024)?
This has prompted me to think about the challenges that might face teachers when using ChatGPT in LRP. As a researcher in teacher learning, I am keen to understand how teachers engage with AI tools and embed them into their practice, but what does that mean beyond knowing about GenAI and the tools themselves? What do I need to know as a teacher in order to engage productively and effectively with ChatGPT in my lesson and resource preparation?
My observations have emerged from interrogating my own teaching practice, as I engaged with the use of ChatGPT in lesson and resources planning for my own class. Studying how teachers work has prompted a widespread interest in teacher identity (Beijaard et al., 2004) and I have reflected on my own experiences by considering three different aspects of a teacher’s identity. It is acknowledged that differences in stance can impact on teachers’ pedagogical reasoning (Horn and Kane, 2015) and so may, in turn, shape how teachers implement GenAI tools in their classroom practice.
Drawing on research (Beijaard et al., 2000; Akkerman and Meijer, 2011; Horn and Kane, 2015; Vermunt et al., 2023), these three representations are: (a) the subject matter perspective, which is where teachers focus on the propositional knowledge of the subject that they are teaching; (b) the student development perspective, where teachers prioritise understanding and supporting their students learning; and (c) the teaching perspective, which centres on the understanding and application of different teaching methods. Akkerman and Meijer (2011) describe how the multiplicity of perspectives is important to how teachers view their professional work, and it seems an appropriate place at which to start exploring the knowledge and understanding employed by teachers when using GenAI tools.
A subject matter perspective
As I adopted a planning-as-pacing approach (Horn and Kane, 2015), knowledge of the subject – or, more specifically, knowledge of the topic that I was teaching – was the starting point. Drawing on curriculum documentation, the learning was framed by expectations of what the children needed to know by the end of the unit of work, and organised into a meaningful sequence. It was this framework that informed my prompts in ChatGPT.
Sifting and discarding
The amount of information generated from my requests for explanations of scientific concepts and misconceptions was extensive and required refinement. For example, the prompt ‘I am teaching my Year 5 class about friction. What are the common misconceptions about friction?’ returned eight separate misconceptions regarding friction – far too many to cover in a session that was focused on revisiting learning from previous years. I had sought to expand my subject knowledge, but ended up needing to use my existing subject knowledge to sift through the points provided and discard those that were not relevant to the session that I was teaching.
Fact-checking
While all the information provided was related to the topic, some suggestions referenced scientific concepts that had yet to be introduced to children in Year 5. For example, the prompt ‘Give me five quiz questions for my Year 5 class on friction and gravity’ generated questions in a range of different formats; however, some referenced concepts to which the children had yet to be introduced. It was clear that I needed to have a good understanding of what a child in Year 5 needed to know in order to check the appropriateness of the questions being presented.
A teaching perspective
Once I had selected the content, I turned my focus to the conception of learning activities that would support my students to build their knowledge and understanding of the topic. I perceived this as an opportunity to use ChatGPT to provide activity ideas for my lesson on gravity. Using the prompt ‘I am teaching my Year 5 students about gravity. What are some activity ideas related to gravity?’, I was given a list of 10 different ideas that would demonstrate the concept of gravity to my students. Again, this required sifting and discarding, but now thinking about the optimal organisation of teaching actions into a lesson (Krepf and König, 2022).
Structuring
Research suggests that when planning a lesson, teachers must anticipate the organisational process of teaching, which includes the organisation of the content into phases and sections so that the pace of learning meets the needs of students (Muijs and Reynolds, 2011). These phases or sections must be connected in turn to maintain the flow of the lesson (Krepf and König, 2022). Consequently, in my own practice, I needed to review the activity idea that I had selected and identify how it fitted within the structure of the lesson that I conceived, organising it into actions that could be enacted either by myself or by the students.
Visualising
Anticipating the organisational process of teaching includes a process of mental simulation, where an idea or activity is ‘rehearsed’ in preparation for teaching (Dudley, 2013). In this case, I took the idea generated by ChatGPT and imagined it taking place in my classroom. Through using visualisation (Mutton et al., 2011), I was able to see these actions taking place in my mind’s eye, envisaging questions that I might need to ask, predicting children’s responses and establishing expected behaviours. This ensured that the lesson was not only coherent but also appropriately pitched for the students in my class.
A student development perspective
This perspective conceives of planning as building from students’ current understanding (Horn and Kane, 2015). While this was not an approach that I adopted at the beginning, during the course of my lesson and resources preparation, I needed to start matching the ideas and activities that I had developed through ChatGPT to the needs of the children in the class. For this, I needed to draw on the detailed knowledge of my students that I had gathered over time from classroom practice. Without this knowledge, it would have been difficult to anticipate the ways in which what was planned would unfold in the classroom (Mutton et al., 2011).
Hypothesising
In one of my lessons, I was proposing to investigate the concept of water resistance, and prompted ChatGPT to provide me with an explanation of this scientific concept. When presented with this explanation, my first response was to revisit the prompt, as I hypothesised that many of the class would require additional scaffoldingProgressively introducing students to new concepts to support their learning to understand the key points. An edited prompt was used but, again, the explanation made some basic assumptions about the children’s understanding, and I predicted that some individuals would need additional information to make sense of this concept.
Scaffolding
This led to my next prompt, as I was keen to include some images that would scaffold the children’s understanding of water resistance – in this instance, focusing on the difference between objects that were streamlined and not streamlined. Although the images were relevant, they were quite complex representations and might have led to further misconceptions. As a result, I used other sources to supplement and support the children’s learning in the lesson. For a primary practitioner, this is perhaps not surprising, as often we are seeking to demonstrate abstract concepts in concrete and pictorial ways, which at present appears to be beyond the capability of ChatGPT (Roy et al., 2024).
Conclusion
At the beginning, I started by asking the question: ‘What do I need to know as a teacher to engage productively and effectively with ChatGPT in my lesson and resource preparation?’ I can see that all aspects of professional identity have been employed when engaging with GenAI tools for LRP. Indeed, the planning process has brought these threads together, revealing their interactions in the deliberation that followed from the ChatGPT prompts. The decisions that I made were framed by judgements about what needed to be done for the good of the students in my class – the domain of practical wisdom or ‘phronesis’ (Biesta, 2015). So, when considering teacher engagement with GenAI, the focus should not just be on developing tools and techniques, but also on investigating the meta-cognitive processes that facilitate their use in the intense, complex reality of classroom life.
While I am still at the start of my journey of exploration, I have identified practices to which I will be returning. For example, the vocabulary list that ChatGPT provided to share with my students at the beginning of the unit of work was a ‘game-changer’. It really did save me time, but there are also times when it is simpler and quicker to use pre-prepared or published resources to support LRP. I am thus reminded of a quotation from McIntyre (2000, p. 10): ‘good teaching is not simply a matter of learning the appropriate ways to act in order to get things done. It is at least equally a matter of learning how to decide what to do.’ And as I reflect on the successes and failures of the last few months, I am reminded that a tool is only ever as good as the craftsperson using it.
The examples of AI use and specific tools in this article are for context only. They do not imply endorsement or recommendation of any particular tool or approach by the Department for Education or the Chartered College of Teaching and any views stated are those of the individual. Any use of AI also needs to be carefully planned, and what is appropriate in one setting may not be elsewhere. You should always follow the DfE’s Generative AI In Education policy position and product safety expectations in addition to aligning any AI use with the DfE’s latest Keeping Children Safe in Education guidance. You can also find teacher and leader toolkits on gov.uk .