In this series of blog posts, we will explore five ways in which you can use Chat GPT to revolutionise your classroom. We’ve already looked at how to create more personalised learning experiences in part 1, so this time we’ll be testing Chat GPT as a virtual teaching assistant.

In these blog posts, I will include screenshots of Chat GPT responses as well as usable prompts for teachers. Keep in mind, as Chat GPT is continuously updated, responses to these prompts may vary in content or format. Also, the version of Chat GPT used in these blog posts will be the free version, not GPT-4, in order to make these methods more accessible to all teachers.

Chat GPT serves as a virtual teaching assistant, available to students at any time. Teachers can integrate Chat GPT into digital platforms or learning management systems, allowing students to ask questions and receive instant responses. Students can seek clarification on concepts, request additional explanations, or ask for guidance on assignments. Teachers can also use Chat GPT to create automated quizzes or assessments, enabling students to practice and receive immediate feedback. This virtual teaching assistant feature empowers students to take control of their learning while freeing up teachers’ time for more individualized instruction.

Above is Chat GPT’s description of its own capabilities, so I decided to put them to the test by asking it to demonstrate its capabilities as a virtual teaching assistant. Here was the result:

I started off by asking it to “Demonstrate explaining concepts for students who may need some extra help.” It chose to clarify the concept of photosynthesis. Below is a screenshot of the majority of the answer.

This seemed like a pretty thorough response, but it also requires a substantial amount of background knowledge in biology to understand. If I wanted Chat GPT to retain this level of discourse (which is quite advanced), I could ask it to clarify certain vocabulary terms further. Instead, I prompted it to “make the explanation simpler”.

The output generated was much more child-friendly, and of course you can also ask Chat GPT to tailor its response to a specific year-level. If each student has access to Chat GPT or an AI-powered tutor, they can do this themselves, and further specify what they need. If not, a teacher could potentially open the class up for questions to input together, but this may not be the most expedient use of class time.

As more AI technology becomes available, honed, and approved for classroom use, which is already starting to happen in some schools, there will be more tools available with which pupils can engage. Ideally, this will democratise AI usage in classrooms, but obviously there is a long way to go before all pupils will have access to these kinds of tools. Access is something to keep in mind when using Chat GPT as a virtual teaching assistant, as the digital divide will prevent some students from getting this kind of help at home.

Speaking of access, another of Chat GPT’s demonstrations of its virtual teaching assistantship capabilities came in the form of inclusivity, meaning it can be a potential tool for SENDCos as well, with some caveats.

In the example above, Chat GPT focused on learning assistance for students with visual impairments. Unfortunately, in its current form, Chat GPT would need additional human or software support to help visually impaired students, because it cannot read aloud the text that it generated. Teachers can feed the audio descriptions into other AI text-to-speech software, such as Speechify, NaturalReader, or voice dream.

For audio-impaired pupils, Chat-GPT may be even more useful. However, in its current form, GPT 3.5 is limited to text-only generation, so even though it says it is able to “provide a visual diagram that illustrates the different stages of the water cycle”, for example, when I asked it to display the diagram, it was only able to provide the following response:

There is still a lot of work to be done to make Chat GPT more accessible and helpful for inclusive education, but AI-powered tools for pupils with SEND do exist, and they are rapidly improving. To give two non-specialised, mainstream examples, Bing can now generate images for free, and if you have purchased GPT-4 from OpenAI, you now have access to Code Interpreter, which is an incredibly powerful tool that can be used for a range of activities, from data analysis to the personalised generation of visual aids from documents or datasets that you can upload to it.

In part 1 of this blog series, I demonstrated what kind of feedback Chat GPT could give for teachers or students in the context of personalising their learning experiences. As a virtual teaching assistant, it can expand upon those functions by giving specific feedback to student work. For this trial, I gave Chat GPT an example response about virtual reality (VR) from one of my own students. First, I asked it to give some general feedback based on the passage provided below.

Here is some example feedback:

This kind of personalised feedback on a written passage is invaluable, especially considering it took a fraction of the time in which a human teacher could have done the same. The points are all tailored to the specific piece of writing, and in my opinion, are good suggestions. Even with the addition of scanning feedback like this and adding a few points of their own (which would be the responsible thing to do, as opposed to fully relying on Chat GPT), teacher workload can be lowered immensely.

In addition to the feedback, Chat GPT also provided a revised version of the passage, which you can read below. This has been a concern with some teachers trialling AI tutors in class, because they understandably don’t want the AI tool to do the work for their students. Teachers will have to be careful in navigating this territory, as pupils will no doubt be tempted to simply pass over reading the suggestions for improvement and copy-paste the revised version for submission.

Even when told to play the role of a teaching assistant, Chat GPT still provided a fully revised version of part of the text, which means that unless specifically prompted to NOT give a revised text, it may be a little bit too helpful (i.e. providing the answers).

Remember, you can very much customise your Chat GPT responses by giving it incredibly specific prompts, but this will be difficult to manage if students are allowed free rein with the tool. One way around this may be to provide a ‘learning pathway’ for your pupils by giving them pre-determined prompts to which they can add their own work, but this will require some trial-and-error testing.

Overall, the foundations of being a useful teaching assistant are built into Chat-GPT, but the execution is slightly problematic. It works best when tasked to explain concepts for students, as having a conversation with a chatbot is infinitely more entertaining and memorable than combing the internet for information. But for other areas of teaching assistantship, such as learning assistance for pupils with SEND and generating feedback, Chat GPT is lacking. This may be addressed in the future by new plugins for the generative AI tool.

For teachers, when using Chat GPT, always be specific with your prompts so you can keep Chat GPT and its outputs focussed and on topic. Also, remember that you can tell Chat GPT to act in a certain way by explicitly telling it to “play the role of…” – this can be a really powerful tool whether you want to use Chat GPT as a teaching assistant, or as an interactive roleplay character (more on this in later posts). Once again, don’t input any personal information, and don’t forget to double check Chat GPT’s output, as it is able to generate misleading information.

Chat GPTs strengths as a teaching assistant will shine brightest when its functions are incorporated into other applications and products, or when new plugins are developed to increase its capabilities. At this point, if allowed access to the tool, students could use Chat GPT to clarify concepts for them, but in reality, its capacity for distraction might overcome some pupils’ desire for learning or clarification. Other, more targeted education AI tools such as TeachMateAI and LessonLab might be more helpful for teachers for now.

Any questions? Comments? Suggestions? Feel free to send me an email me at Luke.Kemper@haringeyeducationpartnership.co.uk

Luke Kemper is Insight and Intelligence Lead at HEP. He recently graduated from the University of Cambridge with an MPhil in Education, Globalisation and International Development. Before that, he worked for seven years as a university lecturer and high school teacher in China and Poland.