Build successful lesson plans using AI

No comments

Ever dreamed of having your own personal teaching assistant? Well, now you can, thanks to AI

An illustration of a science teacher in front of her class with lots of colourful ideas

Source: Adapted from © Shutterstock AI Generator/Shutterstock

From our picture editor: I created this image by entering text and style prompts into an AI image generator. Initially, it produced nightmarish images, but by refining my prompts a few times, I got an image I liked. I had to spend time in Photoshop correcting the very wrong elements and strange inconsistencies in the image – even using some of Photoshop’s new AI tools to help. Although I had a good first impression of this image, every time I subsequently looked at it, I discovered something else to amend.

There’s a new kid in the classroom, but not everyone is convinced they’re playing nicely. Artificial intelligence (AI) is being used more and more in education, with tools like ChatGPT and Bard getting top marks for tasks such as lesson planning, creating quizzes and composing emails or letters.

Outsourcing some of the thinking around lesson planning to AI hugely benefits time-poor teachers, allowing them to use their expert knowledge to fine-tune a plan for the unique class of pupils they’re preparing a lesson for.

I’ve heard it said that lesson planning with AI will de-skill teachers. But any output from an AI tool must be looked at with a critical eye, which requires the expert knowledge of the teacher. Is the lesson appropriate for the pupils in your class? If not, how will you adapt it? Are the teaching strategies appropriate to the aims of your lesson? Does it correctly meet the specifications of the awarding body who will write your learners’ assessments?

Is using AI to support lesson planning going to de-skill teachers? No

As an example, I asked ChatGPT to compose a lesson plan using the following prompt: You are an expert chemistry teacher in Scotland. Write me a lesson plan for teaching the following part of the SQA National 5 Chemistry curriculum: Fuels burn releasing different quantities of energy. The quantity of heat energy released can be determined experimentally and calculated using, E_h = cm∆T. The lesson is 50 minutes.

Five ways with AI lesson plans

Use role play. Get the AI to be who you want it to be: You are a chemistry teacher in Scotland teaching Higher Chemistry …
Be as specific as possible: The lesson should be 50 minutes in length, and please include starter questions to check for the required prior knowledge of …
Don’t accept the first plan. Use prompts to refine it, eg: Can you rewrite but make the practical a teacher-led demo instead of a class practical?
Review and edit what is produced. Is the terminology correct for your context, for example? Is any suggested practical appropriate and risk assessed?
Experiment! Take the time to play about with AI tools and find out what they are capable of.

The plan generated from this prompt was a good starting point for a lesson on this area of the curriculum, and the suggested lesson objectives were appropriate. It began with an explanation of the concept of heat energy as applied to chemical reactions and included a standard practical for measuring heat energy released during combustion. However, there were some context-specific issues. For example, in Scotland the SQA expects pupils to be working in kJ for energy and kg for the mass of the substance being heated, whereas the ChatGPT lesson plan used J and g. In Scotland, the SQA usually uses negative exponents in units (eg kJ kg^-1°C^-1), but the generated lesson plan used solidus notation (eg kJ/kg °C).

ChatGPT produced similar errors when asked to compose a lesson plan to teach metallic bonding for SQA National 5 Chemistry. It suggested teaching pupils about metallic properties of ductility and malleability, which is not required knowledge in National 5 Chemistry.

In conversation with a robot

We can also use AI tools to generate resources to support already-planned lessons. Again, care is needed as AI tools are not immune to pupil-like errors. Below is a real example of a discussion I had with ChatGPT after I asked it to write a question to check for understanding of the use of the relationship E_h = cmΔT.

ChatGPT: If the temperature of a substance increases by 15°C and the mass of the substance is 2 kg, what is the heat energy change (in kJ) using the specific heat capacity of 4 kJ/kg °C?
a) 8 kJ
b) 30 kJ
c) 60 kJ
d) 120 kJ
Answer: c) 60 kJ

Me: Are you sure you have the correct answer?

ChatGPT: I apologise for the mistake in my previous response. The correct answer should be: d) 120 kJ. Thank you for catching that error, and I appreciate your understanding.

I have slowly been introducing my student teachers at Edinburgh Napier University to AI tools and will continue to do so as they progress through their PGDE. ‘Don’t we want our student teachers to develop lesson planning skills?’ I hear you ask. Yes, of course we do.

I work with my student teachers to evolve their lesson planning skills using a (dreaded) lesson plan template. We then use AI to progress these skills further. For example, my students will generate two different lesson plans for a specific lesson, and then compare the plans. Or they will generate a lesson plan and then explain how it can be improved with evidence or research to back up their suggestions for improvement. It would be naive to think that my student teachers will not be using AI tools, so taking the opportunity to illustrate the benefits and pitfalls of using them to generate lesson plans and resources is a no-brainer for me.

So, is using AI to support lesson planning going to de-skill teachers? No. Good teachers will be using AI tools critically and adapting the output for their curriculum and for the pupils in front of them.