It’s here right now and there’s no escaping it. Like computers, like the internet, artificial intelligence is here to stay. It will come with amazing benefits and significant challenges for teachers. Not utilising it is equal to being left behind. We must prepare our students to use artificial intelligence as an awesome tool for learning, ideating, editing and creating. We would also benefit greatly from implementing it in the same manner when planning lessons, making rubrics and more. In this article, I will be discussing the myriad ways we might go about doing this and providing general background information about AI for teachers.
First, I’ll discuss my background on this subject. Cards on the table, I’m that tech guy, the one you call to fix a printer, projector or similar. I’ve been playing around with AI for many years and was interested in integrating it into my teaching somehow as soon as I could. I have made two apps so far on this website that integrate AI to support student learning. I also have a love of sharing my knowledge with others and have been doing so on TikTok for some time. So here we are. This is what I have so far for you on AI. Fair warning, the facts will change fast, but I will try and make this article as timeless as I can.
Should we be using AI in the classroom at all?
It is a fair question. It is likely your school will develop its own stance on this. You may not have a choice, one way or the other. If it is up to you, or you are in charge of this and are trying to decide, here are some matters to consider…
1) The dangers of AI use in the classroom
Chatbots like ChatGPT are designed to generate generally appropriate responses. Still, they are not made for children and cannot be said to have the appropriate level of safety that would be desirable. Snapchat infamously integrated an AI chatbot into their system that kids would be able to chat with. Kids asked it all kinds of concerning questions which it did not answer in the manner that we might hope. Reporters pretending to be children with the AI, described to the bot a clear situation where a child was being groomed and the bot was only encouraging towards this, for example.
Additionally, these systems can ‘hallucinate’ and do some truly bizarre things that we don’t expect. We don’t understand why they give the answers they do, it is beyond us to do so. This obviously has risks associated with it if kids are going to be using these tools since we give up control over what students will see exactly.
2) AI will be in everything you already use
On the other hand, many (if not all) educational apps are trying to embed AI into their systems to ensure they aren’t left behind. These systems can be adaptive in ways we have never seen before. Students can chat with a bot that helps them like a teacher would, providing hints but never giving the answer. This will free up teachers’ time to work with students who need more support and is a whole different kind of flipped classroom than we have seen in the past. If you want to use many of these tools, you won’t be able to avoid AI. In truth, many services that you have been using for years make use of AI, like algorithms in social media. This fact has only become more opaque in recent times. Services created for children will have better safety features to protect students, and this may make these decisions less challenging.
3) Cheating with AI
The biggest obvious change with AI recently has been in the ‘Generative AI’ space. AI can create texts of its own like poems, narrative texts, persuasive arguments, works of art, videos and more. Previously, we have asked our students to create things like this as evidence of their understanding, but this has now become more complicated. Cheating like this has been possible in the past. You could borrow someone’s essay from a past year or online services connect you with people who would do it for you, for a price. But ChatGPT is so quick, cheap and can be used in class. Some students have taken to memorising essays written by ChatGPT. It’s often possible to tell if something is written by ChatGPT, if you have seen a lot of generated text (and text written by that student) but this will almost certainly change as the AI improves. There are also AI detectors that claim to detect the use of AI, but these are only somewhat accurate and can be difficult to trust.
Final thoughts on whether it should be used
I think some analogies will be helpful for this. We use the internet at schools, but generally with a filter that prevents students from accessing websites that may be inappropriate for them. Similarly, students won’t have access to all AI tools, but they should have access to some. We need to prepare students for an AI future, and there is some risk, just as there has always been with the internet, even if the filters are excellent, they can’t be perfect.
We might also consider the calculator as an analogy. Sometimes students are not allowed to use them and sometimes they are, depending on the task. But our expectations will be significantly higher for what they can do if they use a calculator. If students are going to write a narrative with the help of ChatGPT, the expectation of what they produce will have to be raised significantly, but it would be really interesting to see what they come up with. Students are going to use these tools at home regardless of what we do, so we should teach them to use them to be productive and teach them to use them safely. They are going to use them at work, in the future, so we should prepare them for this future.
How teachers can use AI to save time and to plan
Planning lessons, activities and so on is an involved and creative task. Imagine if you had a really smart person with you who you could bounce all of your ideas off of. Also, imagine if they would type up everything you make together nicely. We are already there, and it is really great.
I asked ChatGPT to create a lesson on gravity for Year 3 students (see above). It isn’t amazing, but what is amazing is you can make suggestions and adjustments. It will update based on what you tell it.
I had it create a task for a unit on gravity (see below)
Followed by a rubric (see below)
All of this took about 60 seconds.
What are the benefits?
Your productivity can be amplified by this technology. You’ll save time and can get more done or put more detail into everything you do. It also can prevent you from getting stuck and waiting around being unsure what to do. You should try it out at least and see for yourself!
Making quizzes and worksheets
You can use the same process for making quizzes, for example, with multichoice answers.
Or a CLOZE passage (see below)
Or spelling lists for a particular sound, such as this one where I asked for words that have the same vowel sound as ‘boat’.
An activity with narrative writing and editing
Students find it difficult to edit their own or their peers’ work because of the emotions surrounding this. AI can write texts with clear mistakes or that are lacking in any way that you choose. Having students fix these stories and make them better is a great way for students to approach writing and editing stories.
I once co-constructed a story with my students and then had ChatGPT create the full story for us. We read the story together and it was great for students to see the way that a plan is supposed to work for creating a story.
For these tasks, I have preferred ChatGPT because it is the most popular and best at the time of writing. That may not always be the case though and Google and others have comparable tools that are worth considering.
Generating images using AI
There are lots of fun opportunities for creating images using AI. Dalle can create amazing images from text.
Bringing students’ stories to life
You could create images of the characters that your students create in class. It can accurately represent their descriptive language so this will hopefully encourage them to be more descriptive (see below).
Creating amazing story prompts
There are endless prompts online for stories, but now you can create anything you can think of (see below).
The ways that you could use this are endless. It will speed up your process, improve your work quality and generally help you to get things done and to do things you never could have done without it.
Specific apps
There are many apps that aim to support students with planning that are more directed and offer more features than a chatbot. See some examples below.
Students interacting with AI directly
There are many tools out there to have students interact with AI in a direct way. We might have them use ChatGPT for example, but there are more directed, scaffolded apps, designed for students to use and this might be a more appropriate place to start.
Getting quick feedback
For example, I made this app for students to get feedback on their narratives quickly without needing to go to the teacher. They select whatever criteria they want and then get feedback on that.
Khan Academy
They have implemented AI into their app. Students can now get feedback quickly and easily that will help them to work through the questions it gives them, along with the video they would normally watch.
Talking with historical figures
Many companies are offering the ability for students to have a conversation with a historical figure. This is a fun approach to learning about history, for example.
https://www.hellohistory.ai/for-education
Bringing drawings to life
https://sketch.metademolab.com/
We’re just at the beginning of this. A lot of companies are trying to figure out how to make great products that add value in this space. Many of the big companies in educational apps are adding AI to their apps.
A word on AI safety
Many people are concerned about what AI will mean for our society. I’m just going to outline some of these ideas in short. You may wish to discuss them with your class and have them consider them although there are no easy answers.
Firstly, some are worried that bad actors will use AI against humanity. Others worry that no bad actors will be necessary and that the road to an AI apocalypse will be paved by technologists who relentlessly seek progress and/or profits. There are also the same concerns there have always been with technologies like this, that jobs will be taken away. Finally, some believe that AIs will have a large impact on matters like financial markets since they can see the whole picture in a way that humans cannot. At the end of the day, we don’t know what will happen. AI could be considered to be a bit like an invasive species. We don’t know the effect that it will have on our ecosystem, to have a human-like and yet non-human intelligence among us.
On the positive side, AI may be used and is being used to cure diseases. Many futurists predict a future in which humans will not need to work at all and could live on a universal basic income since all of the work could be done by AI and perhaps physical robots whose ‘brains’ are of AI. Personally, I have seen the benefits of AI, I have created things I could not have created before and saved a lot of time. I feel excited about the possibilities while recognising that there are dangers that need to be managed.
Finally, it’s interesting to discuss with students how we think about AI. I read the book called ‘The Wild Robot‘ with my class. In the book, the robot is kind of like a human and kind of not. The book teaches you to be sympathetic to the robot through its story. The conclusion I have come to regarding how we treat robots is that the way that we treat robots is a reflection of us. These are our creations. I don’t believe they are alive, like a human. I don’t believe that they ever can be like a human in this way. But it does not mean that we should treat them poorly. This kind of moral and philosophical question is an interesting one to have with your class and I can recommend it from experience.
Final thoughts
I hope you’ll see this as your time to start trying out the technology and seeing what it can do. If you haven’t started playing around with it yet, you should give it a go. The more familiar you are with the technology, the easier it will be for you to teach (as is the case with anything you want to teach). Start considering how you could use it to save time and also how you could make your class a more exciting place that prepares students for an AI future.
Glossary
Here are some terms that would be helpful for you to know as you are considering AI and its use in the classroom. These aren’t the official definitions, just my own.
Artificial Intelligence: When computers can do what we used to think that only humans could do. For example, chess used to belong in the domain of humans alone but AI has been created to play chess with humans (and far exceeds the abilities of humans although I wouldn’t consider that to be necessary to meet the definition). The AI available today can write poems, create works of art, PowerPoint presentations, realistic videos, write computer code and more. We could compare this to Microsoft Word which allows humans to create a document and is a tool, like a hammer, but isn’t intelligent, like a hammer doesn’t hit the nail itself or build a building by itself.
Chatbot: AI you can talk with. This is a way to interact with AI. You can chat with it like a person using text and ask it to do things for you. By contrast, you can’t talk to AI designed to play chess because it is designed to play chess and not to chat.
Generative AI: AI that creates (or generates) something. This has been the primary innovation leading to the current focus on AI. In the past, we have had AI that would make suggestions on what you should buy online, and what posts you should see on social media, but now we have AI that can generate a narrative or generate an image. This generation is what makes modern AI unique.
Large Language Model (LLM): Large language models are ‘trained’ on large amounts of text in ‘data sets’. They provide statistical responses to whatever you ask. This technology is used in autocorrect to finish your sentence. For example, given the stem, ‘In my garden, I planted a’ it will probably guess the next word is ‘tree’. When you ask a question of an LLM, it will develop a probable answer to your question that is statistically likely (although there is far more going on than this of course). Truth is less important than looking true, a lot of the time it will give a truthful answer regardless.
Algorithm: Rules that a computer follows to make decisions. Algorithms decide what to show you online, for example, and they often change and grow over time based on what users do through a process called machine learning.
Machine Learning: Whenever software changes what it will do based on things that have happened in the past. The machine is learning and improving through experience.
AI safety: There are dangers associated with AI that many are concerned about. For example, AI might choose a way to complete our goals that we don’t expect and can’t detect. It might do bad things to obtain good outcomes. AI safety involves the consideration of whether AI is safe for humans or and considers the potential dangers of AI.
Alignment: considering whether AI is aligned with those who are using it (and the good of humanity generally).
Fine-tuning: The process of making AI better (either more accurate or aligned) by providing it with feedback about the quality of its outputs.
Neural Network: A way of creating AI that mimics the workings of neurons in the human brain. It is primarily about creating stronger or weaker connections between variables. This is different to a rule-based system that follows a set of rules created and managed by humans. It is harder to understand why a neural network does what it does since they are generally trained on a large amount of data.
Natural Language Processing: AI that is trained to analyse language by considering aspects beyond word meanings and grammatical structure. For example, it would know that in the sentence ‘I live for live music’, the first ‘live’ is a verb and the second one is an adjective since it can understand the context.
Artificial General Intelligence: AI that has a general ability to work at or above human-level intelligence. This is non-specific, it will create works of art, book flights, file your tax return, make a presentation, play chess, talk like a friend and so on.
Hallucination: When AI makes up a fact and presents it as true. Especially with LLMs, they are using statistical probability to make something that is convincingly true. Fine-tuning is being used to align AI models with our requirements of truth when we are looking for facts. This can be complicated, since we often require AI to make up facts, for example, in a fictional story. Newer models provide sources for their information and fact-checking is built in as a solution for this.
Prompt engineering: The particular ways that you choose to phrase your questions to chat bots can affect the quality of the outputs you receive. Prompt engineering means working on the questions you will ask to get the best answers possible.
AI model: Many companies are working to create AI models. These are made with different data sets varying in content and quantity. Generally, the more data, the better the model. GPT-3.5, GPT-4 (OpenAI), Gemini (Google), LLaMA (Facebook) are some examples of different models.
Image Generator: Generates an image based on a text prompt. Dalle 3 by OpenAI is one example.
ChatGPT: LLMs can exist and be used without a chat interface, however, being able to chat with LLMs is what made them become popular so quickly. This method of communication with the AI makes the most sense for use currently since we are used to chatting with text.
Conversational AI: This is AI that you can chat back and forth with like ChatGPT. You can’t chat with a chess-playing AI, or one that generates video.
Generative Pre-training Transformer (GPT): This is one of the key technologies that led to generative AI models like GPT-4. It was created by Google and allowed AI to train on large amounts of information with little supervision. OpenAI used this technology to create a generative model that would produce text of its own. This was integrated into a chat system, ChatGPT.
Dataset: A large amount of information that is used to train an AI Model.
