Roll Up Your Sleeves: The Practical AI Toolkit for Busy Educators

I want to say welcome everyone again to Roll Up Your Sleeves, the practical AI toolkit for busy educators. My name is Sophie. I'm from AIX Education, and I'm here with our amazing speaker, Rebecca Bzma. Before we begin, please note that we'll be recording this session. And if you prefer not to appear in the recording, feel free to DM me, and I'll use the editing magic to protect your privacy.

Now, a little bit about us. AIX education is a student-run initiative dedicated to bringing together students, educators, researchers, and founders to collectively explore the future of AI and education. We host events like this webinar, as well as conferences, and we also publish a bi-weekly newsletter to share the latest resources and research in this very exciting field. Today we're thrilled to have Rebecca with us, a practical AI strategist, seasoned keynote speaker, and chief innovation officer at Amplify and Live Innovation. Rebecca has trained thousands of educators across North America and will be guiding us through practical, ethical ways to integrate AI into your classroom, regardless of your technical background.

And on that note, let's welcome Rebecca.

All right, I'm just going to share my screen, my screen here very quickly and make sure I can get it where I can see it. That will be important, I think. Thank you so much for joining us today. I am thrilled to be here talking to you. And if someone could just give me a verbal, uh, confirmation that you can see my screen, I would appreciate it. Anybody?

The screen share is working perfectly.

Excellent. All right. Well, let's dive right in. A little bit about me. Um, I am—I speak to a lot of people about AI. I research AI ethics at the University of Edinburgh currently. And so a lot of what I'm planning to share with you today will be based in some research. Um, and I'm—have a lot of that linked into this presentation, and I'm happy to share this deck with you at the end of our conversation today because I have linked in a lot of things I think you'll find useful. If you have questions as we go along, uh, please ask them. Sophie will try and interrupt. I do get on a roll sometimes, but I like to address things as they come up if at all possible. Uh, I talk to a lot of, uh, particularly communicators, public relations experts, but mostly educators and educational leaders about ethical AI but practical AI strategies. So the reality is this: AI—what we're going to talk about today, uh, a lot of it is very practical, and I think once you understand how to use it effectively, you can be a lot more effective in everything you're doing moving forward.

So, our road map for today: I'm going to show you how to get some quick wins with, uh, AI, how to use it strategically in your classrooms. I had a lot of questions presession about feedback and assessment. And so, I'm going to talk about that a little bit with some strategies, but also some ethical cautions. And then I'm going to share some of my favorite tools with you. Some of them were in our presession conversation. Uh, some of them you may have heard of, some are brand new, but I've been adding things to this presentation right up until 5 minutes before we started. So, let's dive in. Oh, and a friend of mine based in New York and I, uh, from Amplified and Elevated Education have an education and AI focused podcast you're welcome to listen to. It's called AMP211.

So, I want to start by just giving you a quick reminder, if if you're unfamiliar, of how we kind of ended up here. Uh, it all started, the magic kind of started back in 2022 when, with this one tweet, OpenAI launched Chat GPT, and it hit a million users in 5 days, which—it was unheard of at the time. The next closest or fastest-growing platform was Instagram, which hit a million users in about two and a half months. And generative AI was a brand new type of AI that basically consumes massive amounts of data, finds patterns in what it learns, and then creates brand new content customized to the questions that you ask. That's kind of a simplified version of it. Since that day in November 2022, um, this is probably actually old now. As of February, it was 400 million weekly users for Chat GPT. Uh, Chat GPT released something just yesterday that said they were up to 500 million weekly users, which is half a billion people, and Sam Altman tweeted this, uh, on March 31st that they added 1 million users in one hour. So this is a fast-growing platform, and it's about to grow even faster because as of this morning, uh, Chat GPT is available for free—the pro version—to every college student. So, this is something to put on your radar because this means that if your students were not already using it, they probably will be now, especially this pro version that gives them basically unlimited access.

So, let's talk a little bit about what this might mean for you, how you can use it, and ways that it might be helpful. I want to start by just talking about some quick wins because I come across a lot of different audiences and people and sentiments and feelings about AI in general when I'm speaking, and the main sentiment I get is that people are overwhelmed, and everybody feels like they're behind. And so I'm here to tell you you're not behind. By the end of our session today, you'll know more than most people about AI. So it's okay to start small, and if you're overwhelmed, everybody is. Everyone's figuring it out as we go. Don't worry. This—I liked this quote, and this is about a year old, but Amy Webb was on the Brené Brown podcast, and she said that—um, she said that we're all kind of going through this transition, and we're collectively this Gen T, and we don't know how long this period of transition is going to last, but everybody's going through the same amount of change right now, so to not feel like it's just you.

So I want to start by talking—just the quickest win that you can get with generative AI is using it as an admin assistant. So using it to help you get started on emails, using it to summarize big documents for yourself, using it to create outlines and drafts, all these common, time-consuming tasks and communication drafts. So if you find a way to save yourself 20 minutes a day, let's say, with AI on an admin task, that actually ends up to about two and a half weeks of a working week a year. So look for these small wins. There's a lot of different ways that you can use AI for—a lot of different ways to help you get through these things a little bit faster. Uh, but how exactly do you do that? This is where we kind of dive into it. So, Chat GPT is the most common type of generative AI or large language model for doing things like this. But these are kind of the big three that you should know about. There's Chat GPT, there's one called Claude AI, and Google has one called Gemini. There's several large language models out there. These are the biggest three, more or less. Microsoft Copilot is actually Chat GPT. It's built on top of Chat GPT. So, I don't put them in their own category, but it does have a lot of functionality as well. But these are the three that I think you should know about. And I think that every time you're interacting with AI, open up all three. Ask it the same question, all three of them, and see what happens in response and what you can get—how they differ. They have different personalities in a lot of ways and different strengths.

So, there's a few different ways that you can approach getting AI to do things for you. And we got a lot of questions presession, and I want to show you how to answer your own questions with AI. There's a lot of different ways, and we're going to start at the most basic, and we're going to go to advanced. So really, if you don't know—if you've never really interacted with a large language model like Chat GPT, uh, you might feel intimidated when you go and have to start figuring out how to interact with it. But the easiest way to do—to interact with a large language model is say, "Hey, I have to do this thing. How can you help me?" And have Chat GPT give you 20 ideas of how it can help you do the thing you need to do. "I have to plan my daughter's birthday party. How can you help me?" "I have to plan a lesson. Give me 10 ways you can help me." That's all you have to do. Type that in. "I have to do the whatever it is. How can you help me?" That is the most basic way to approach any large language model like Chat GPT or Claude or Gemini. It's free. Just go type something in and start a conversation.

Now, we're going to start moving into a little more advanced talking to AI. Uh, the most common way that we describe that is through prompting, giving instructions to a large language model to have it do exactly the things that we want it to do. And over the last few years, people have come up with a lot of ways to interact with Chat GPT or any of these generative AI models that make them even more effective at responding to you—much like people. And my background originally is in communications, which is why I'm so fascinated with these large language models because how you communicate with them matters. Just like people, there's certain strategies—if you've ever done crucial conversations or any of these—that help you be more effective communicators, and there's certain strategies that you use with AI to give it the best possible responses. Now, if you're brand new, you might want to find this Gemini prompting guide 101. It's free online. But what everyone has kind of agreed is that this is kind of the secret sauce to getting an AI to do something, uh, as—as well as possible for you. Uh, it's four steps. Number one is you assign a persona. So if I wanted, uh, uh, to develop a lesson plan, I would say something like, "I want you to act like the best teacher or curriculum designer in the world." That would be step one. "Design me a lesson plan about Romeo and Juliet." That would be the task. The context is it's for 10th-grade students, uh, at the—in this state, and then the format. "I'd like this to be in a lesson plan that includes these things or bullet points." This is the secret sauce for kind of an optimized version of a prompt. We call them optimized prompts that are designed to give you the best possible output. And this matters, but not as much as you might think because there are now ways to have AI do all of this for you. But I'd recommend that when you start interacting with AI, you start by experimenting with this format. Go first to Chat GPT and say, "Hey, write me a lesson plan about Romeo and Juliet." And then try it in this optimized format because how you talk to it matters. So that's your second option. So remember option one, you just—Oh, option one, you just go and say, "Sorry, I've just got to represent this. Didn't like that." Okay, we still with me here, guys? Sorry, I'm just going to have to reshare. Apologies. You hit that one button, and it all goes sideways, I suppose.

Oh, good. We can still see the notes on the side. Yeah. Let's see. I think I might just have to reopen it up. Apologies, team. We're just getting to the good stuff. Trust me, this is the basics.

All right. If—if that just got anyone excited and have some questions, plea—please feel free to share in the chat. Okay, I think we're good. I just don't know how to get rid of that side window, but I think if we go in and present, we should be good. H—maybe. Wow. Okay. Well, uh, yeah, if you have any questions, I might just need a minute to figure out what's happening with my Canva. One second here.

We have someone in the chat who asked, "Are the slides available to access to follow along?" Yes. Oh, to—to follow along. I'll send them after, but now that we've got it live here, I'm ready to go. Can you see my screen?

Perfect. Yes.

All right. So, option one, just ask Chat GPT how it can help you. Option two is to use an optimized prompt and a prompting guide. Option three is to use prompts that somebody else has already written. And this is where it gets really fun. So my favorite AI rockstar guy is a guy named Ethan Mollik. He's a professor at Wharton School of Business, and he has designed optimized, amazing prompts specifically for educators, and they're at moreusefulthings.com. But he has them all linked here. I have them linked in the slideshow, which I'll share with you, and he has written several papers about the best possible prompts, and he—you can see in here—the ones that will create an AI that will be a teaching assistant for you—a tutor—blueprint—project ideas for classes—quiz creators—active learning co-creators—and he's also developed some for student exercises, and all you have to do is go to this site and copy and paste them into your Chat GPT. So you'll see here there are negotiation simulators—interrogation agents—teach the AI as a student prompts. So you might not have to write your own prompts. You can use some of these optimized, amazing prompts that other people have invested time in writing. And I highly recommend Ethan Mollik. He's also written a great book, uh, called *Co-Intelligence*, that if you really want to learn more about AI, it's written at a beginner level, and it's very, very interesting and very well written. I highly recommend it. So this is option three is to use prompts that somebody else has already written, like Ethan Mollik. Option four is something that not a lot of people are familiar with, but it's the absolute best-kept secret within Chat GPT. They are called GPTs, and it's almost like a mini, hidden app store within Chat GPT. So, it has apps or special versions of Chat GPT for free that are programmed and personalized to do amazing things. So, when you go in, they have lists—almost like the app store—of things. There's one that's an expert in plants, and you can give it any picture of a plant, and it'll tell you what it is. There's one that is a Gen Z translator where you can put in Gen Z lingo, and it'll explain it to you, uh, in normal language. I have to put memes for my kids in there all the time because I have four kids in college. Uh, but there's ones that will explain the rules of games. There's tons of different ones, but I'm going to show you the one that will change your life right now and why you don't have to worry so much about prompting. So, when you're looking at Chat GPT, there's this little thing right here that says "Explore GPTs." That's where you're going to find it. This one is called Prompt Maker. It's there because I used it before. Now, I want to show you—this is not us saying, "Hey, Chat GPT, write me a lesson plan." This is us saying, "Write me the best possible prompt so that Chat GPT will write me a good lesson plan." So, instead of me just saying, "Hey, make me a lesson plan," I'm having this Prompt Maker GPT write me a prompt that will evaluate my lesson plan and identify ways that I can make it more engaging and offer suggestions for improvement. So, it's going to write me an optimized prompt. Take a look. Remember, we talked about a persona. It's giving it this persona—an experienced instructional designer, an expert curriculum consultant, 20 years experience. "I will provide a lesson plan." And it's writing this as instructions on its own. And it will do this for anything. You can make it an expert in anything. It's giving you all of these steps. And all you have to do now that you have an optimized prompt is to copy and paste it with your lesson plan. And you have optimized and made Chat GPT an absolute expert in whatever you need it to do. So use—go in and explore GPTs. This is the Gen Z for meme one. There's all different ones, but that is going to be your other secret sauce.

So just ask Chat GPT how it can help you. You can experiment with some of this prompting on your own. You can use prompts somebody else has written, or you can use Chat GPT or a prompt maker to write prompts for you. Keeping in mind, no matter how you interact with Chat GPT, one of the most important things is that after you prompt and you get an output from Chat GPT, review it and evaluate it for accuracy and bias. Edit it, have a back-and-forth conversation, but make sure it matches the intended audience and purpose. Follow up, iterate to refine the output, but know that you have to take responsibility for the final work. Don't just blindly use whatever it gives you because there's a lot of risks, and risks are all I research and think about and talk about a lot.

So, I'm just going to give you a high level of a couple things to be aware of. Um, and that is some of the limitations of generative AI, these technical harms. And I read a really great paper about a week ago that just came out. Um, and it's called "Don't Forget the Teachers: Towards an Educator-Centered Understanding of Harms from Large Language Models," and it talks a lot about, uh, all the different types of harms that teachers identify and how they are actively working to overcome them. So, I have it linked in this deck, but some of the limitations you need to be aware of. Um, well, there's—there's a few. First of all, uh, anybody know who this is? I don't know if I can see in the chat, but, uh, spoiler alert, it's Caitlin Clark. Uh, she broke an NCAA record last year, and this was an ad that Nike put out. What makes this ad great is that whether or not you know who Caitlin Clark is, you understand because of your wisdom and your lived experience that when they say "break it" here, we understand—they're probably talking about a record—even though it doesn't say that. We don't think she broke a dinner plate. The idea of owning it—we understand through her body language that she's owning the moment, that she's owning her success. But this type of nuance is very, very human. And people have tried to get Chat GPT to replicate an ad like this, and it struggles because there's several things that are limiting for Chat GPT. For example, uh, it's not great at complex social interactions. It struggles with cultural understanding and emotional intelligence, ethical decision-making. You need to know that AI is never neutral—ever. There's always bias embedded in it. And there's a huge risk of inaccuracies and errors. For example, earlier on, about a year ago, Google decided that instead of you having to go through and click all of the links, um, when you did a Google search, they were just going to give you a summary at the top. And they started doing that, and they realized right away that there was going to be a problem. For example, somebody asked the AI overview feature, uh, what they should do if their cheese is not sticking to their pizza, and it told them to add an eighth of a cup of non-toxic glue to the sauce. Why did it do that? Well, because it had found a Reddit sub-thread where somebody mentioned that, and it was funny. So, a ton of people gave it a thumbs up. And because the Google Gemini didn't understand the context, they just knew that a lot of people liked it. They thought it was the best answer, but in reality, it makes no sense in the context. There were several examples of this. For example, uh, UC Berkeley—"You should eat one small rock a day." Uh, the Apple overviews—if you get them currently—like it will give you a summary of—of news and notifications on your phone, and sometimes they're really, really far off. So this is a limitation of generative AI, and which is why you have to fact-check everything, and you have to tell your students or anyone you work with who's using this that they need to fact-check everything. There's been lawyers who have used fake, uh, sources and cases because they trusted the outputs from Chat GPT and—and all other large language models. But the reality is if the AI doesn't know the answer, it's going to just make something up, which is why you have to check everything. In New York, there's been issues because their AI tool has been giving out illegal advice. In—in Canada, there was a chatbot that was giving customers bad advice that led to a Supreme Court ruling that ultimately said that you are fully responsible for every output, uh, that an AI generates—your company, you—which means you need eyes on everything before it goes anywhere.

So my best strategy that I like to talk about for how you approach using AI is to use something like Chat GPT to help get you started for 20%. This might mean—say, "Hey, I have to do this thing. Can you give me 50 ideas on how I should get started? How do I do this thing?" Then moving out of AI, working with your human expertise and all the wisdom that you've gained somewhere else—Word, Google Docs, Grammarly, wherever you're working—but then at the end using AI to gap-check and polish and refine. And I don't mean just spell-check, because that's been around for a long time. One of my favorite ways to gap-check is to think, okay, uh, "Review this document that I just wrote as if you were my boss, my teacher, my—a parent that I have to talk to, and ask me 10 questions you might have about this or tell me what I missed or tell me things that you like or don't like about it. Where are the gaps? What doesn't make sense?" And giving it those extra set of eyes. I like this strategy. I think it works well.

So to wrap up this section, I just want to talk a little bit about how you can use this in a workflow. Let's say you think of a task that you have to do right now anyway that's on your docket that you probably should be doing right now, but instead you're in this webinar, and it's kind of—you're like, "I've got to get to that." Figure out that task. Go to Chat GPT or Claude or Google Gemini and say, "Hey, I have to do this thing. How can you help me?" and/or use a prompt or go try out a GPT to write your prompt for you, and then use that Oreo strategy. Help it get you started. Go somewhere else, but then use it—complete the task with your human expertise—fact-check and review it with your human eyes, and then gap-check it with AI. This I find is a very, very useful strategy.

So before we move on, do we have any questions before we hit section two?

No, I can't see the chat. So, I'm just gonna try—I can—All I can see is Sophie, so we're just going to trust Sophie here.

All right. Currently, we don't have any questions in the chat, but if anyone wants to raise their hand and speak, please do. We would love to hear your thoughts. And I know I'm moving quickly and covering a lot. So, feel free to reach out to me after with questions. Uh, connect with me on LinkedIn or—um—take a look through the deck.

So, let's move on to some classroom power-ups. So, Sophie did a great job of collecting information about questions that people asked before the session and sending them to me. And I thought it would be great to identify ways to answer those questions, but also show you how you can use AI to answer your own questions. People always say, "Hey, can AI help me do this thing?" And I like to show them how to ask AI those things and have AI help them answer those questions. So, two of the big things that came up though were lesson planning and student engagement. So, we're going to focus on some of those things.

So, let's talk about classroom power-up number one, which is enhanced lesson planning. And I want to go back to the GPT section of Chat GPT, which you can go in, and there is a full section all on education GPTs that are already built and already programmed. Uh, or you can just go in and type in something like "lesson planner" or, um, "schedule maker" or all of the different things that you have to do. There is probably a GPT. Keep in mind these are all built by other people. So if—

You look at this top one on the list, I just started typing in lesson. This top lesson planner on this list here, it looks like it has 25,000 uses. So again, think like the app store. This is kind of like the app store ratings. There’ll be some that kind of suck, and there’ll be some that are really, really great. So I found that prompt maker one that I like. You’ll find some that you like, but it gets even better because these are great, but you can build your own GPT.

And it sounds hard, but it’s actually not. It’s building your own little expert in your class, in your job. And actually, the Harvard Business Publishing says that for higher ed, a custom GPT’s key benefit is that it can be built around specific content and use cases business schools or any school wants to deliver. So, what this might look like is you can build a GPT for your class with all of your teaching materials where people can just go in and ask questions and talk to it. There’s really, really great instructions that I’ll link here on how to make a custom GPT that OpenAI posted.

Basically, you just go to my GPTs. You can make as many as you want for all different tasks. You upload all the documents that you want it to reference, and it’s it’s actually very, very easy. You don’t have to know code. It used to be that to talk to a computer, you had to know code, and now you just need to be able to communicate well what you want to do. So building your own GPT specifically for lesson planning might be a great use case for you.

Somebody in the chat mentioned this, and this has been one of my favorite tools since the day it came out, but one way to enhance lesson planning is to come up with great ways to visualize complex ideas for students, for your decks, whatever that looks like, or if you’re a student studying. There is a great tool called Napkin AI, and you just copy and paste in any text and you imagine like you’re sitting with somebody and you’re explaining something in a restaurant. You’re drawing it on a napkin. All you do is push one button, and it will take your information and turn it into a chart. Any chart. It gives you hundreds of options. This is talking about communication processes. And as you can see, it will visualize and graph anything. And from there, it is a free download. You can change the colors. You can change the fonts and get any visual out of any text. So if you’re trying to explain things for students or if you’re trying to understand a complex idea yourself, to make something like this in Canva would take you forever. Now you push one button, and you have everything illustrated. So this might be something interesting to include in your lesson planning.

Another tool that was mentioned before we got here, a great classroom power-up and my favorite AI tool is Notebook LM, and it’s by Google. And I do want to tell you, I know that one of my top questions I get right now about, oh, should I put my data in here? Is my data even safe? For education, like Google education for sure, um, they do not train on your data for this. Your data is secure. But here’s exactly how Notebook LM works. And this I’m telling you is the best tool I’ve I’ve used to date.

So, Notebook LM gives you all of these different notebooks for different things. I’ve got some research in here. So, we’re going to go into my data and AI ethics as a practice notebook where I have about a hundred different PDF journal articles, resources, tools, and you can go in and ask a question and query your information or say something like, “Can you create a lesson plan for me using the material just in this folder?” and it will come back, give you everything that you need, and you can click anywhere on your output, and it will show you exactly where it got the information. You can see these little numbers. It will show me exactly where it found that information so that I can cross-check it and make sure it’s exactly perfectly cited. I can also go over here and click on something like study guide, and it will take everything in my notebook and instantly turn it into a study guide or instantly turn it into a briefing document, anything I want based on hundreds of articles, or a brand new feature added a week ago is this mindmap feature which is great for helping people understand things. You it will mind map everything that you have for content, and then you can just download it instantly as an image that you can then use as a study tool.

So, this is an amazing tool, but it gets better because you can take all of this information and you can turn it into a conversational audio overview, which means a two-person conversational podcast talking about the information in your notebook. So, when we talk about student engagement, how do we make this engaging for students? I want to make sure I’m like scared to stop sharing now because I want to make sure that I have my audio shared. So, we’re going to try this one more time because this is very cool. Share sound. Perfect. We’re good.

So, to make this engaging for students, I’ll tell you a story about one of my sons. I have a son who plays college basketball, and he sucks at math, and he was trying, trying to figure out this business mathematics course, and he really, really was struggling. So, we gave Notebook LM all of his slide decks from his instructor and asked it to create him an audio overview using all basketball analogies to help him understand the content. And he aced the test because it did that.

So, what I’ve done here to show you how to make this engaging is I went in and I said, uh, remember this is data and AI ethics. It’s very, very heavy, dense readings. And I just said, “Can you explain all of this content to me using Taylor Swift references,” and generated a podcast, an audio podcast. So, we’re just going to I’ll give you a listen to this so you can just see how great it is and how how great it sounds here. Hopefully, you can hear it. So, you can give me a thumbs up to see AI, doesn’t it? That’s a fantastic way to frame it. Think about it. When a new Taylor Swift album drops, there’s often intense interest in who collaborated on it, where it was recorded, what inspired certain lyrics. Data sheets for data sets proposes a similar level of background information for the data that powers AI, right? It’s about moving beyond.

So, you can see how if you are an audio learner, this would be an amazing way to keep students engaged, to meet people where they’re at, or for you to consume information in any way that you like. So, it ends up being a very, very useful tool for sure.

Here’s where we’re going to go a little bit more advanced, but things that I really think that you will appreciate and find useful no matter what you do for your job. So about I want to say about three, maybe about six months ago, we kind of realized that AI models, if you added something to your prompt like take a deep breath and work step by step, kind of like people, if you gave the AI time to think before it answered, it would give you a better response. So, I thought some of your presession questions were were tricky, and I thought we would use advanced thinking models or reasoning models from AI to help us answer some of those.

Now, this whole idea of think before you speak works with people, right? We know that if you get a little more time to think before you give an answer, it does a better job. And what’s interesting about this is regular generative AI can’t do something like a crossword puzzle because it starts and works through a problem start to finish. Whereas this reasoning model can think through every possible option and choose the best one. It just takes a little longer to think through every possible option. And Gemini has a version of this, and chat GPT has a version of this, and Claude has a version of this. But this is very, very good. And when I say reasoning and high level, the best versions of this, they say reason and think at a PhD level in any subject that you want, which is huge. If you could have an assistant who’s an expert at anything at a PhD level, that’s amazing.

So, I thought I’d show you this in action using one of your questions that was asked presession. So, you go up here in your chat GPT, there’s a drop-down of all the models, and there’ll be something like 01. So, I asked it, what are some strategies for accurately measuring and reporting student achievement in the age of AI? That was a question we had. And you can see that it’s taking a little while to think before it answers. And then it will give me a response. And the longer it thinks, the harder your question was. So if you really want to play with it, try and ask really hard questions. This only thought for 21 seconds, so it didn’t think this was that hard. But from here then, you could really dive into some of these. And this is how it answered the question. This is information that you could have you have available to you at any time. But these advanced reasoning models, you can say, think through every possible way that this could go right and wrong and give me 50 options. It will think through every possible scenario for you.

In any case, the most major advancements in the last couple of months has been this idea of deep research. And this used to be something that was only available for the highest paid tiers of chat GPT. And then Deepseek, a a China Chinese-based AI company, came along and made this available to everybody for free. And so now it is available to everyone for free. Google Gemini. Actually, I highly recommend you go check out Google Gemini’s deep research feature because they have made it available for free for everybody right now, which is huge. And I’m going to show you how it works. When you go to Google Gemini, there’s a drop-down, and you’re just going to go down to the deep research feature. This is this is amazing. And working in education, you’ll appreciate this.

So, I used one of your questions. How might teachers leverage AI to improve their teaching efficacy and their student engagement? It goes through and it creates a research plan for what I’ve asked it to do. And then we can see it working and thinking in the background. And it takes about 15 or 20 minutes, but it goes through hundreds and hundreds of sources to write me a detailed master’s level research paper using tons of information. So, you can see here that it’s thinking, I found several AI tools designed for teaching assistants. These tools might give me some ideas for this. You can see its thinking process as it does its research. I’ve discovered that AI can be instrumental. Let’s go back to that and just take a quick look at some of these some of the readings the research that it has here that it’s been doing. Oh, or not. I need like a Canva troubleshooter on this call because I can’t get down in here while I’m presenting. Anyway, it goes through hundreds and hundreds of research sources and goes through every possible scenario as we would as researchers. Go through and and look at tons of different sources. It’s important to note here though that things like Google Gemini don’t necessarily have access to the hardcore academic research that we have behind school paywalls, right? So, they might not have access to peer-reviewed secondary sources. You’ll notice a lot of the sources they’re going through here are websites are edtech companies. Some of them are like school AI, some of them are um, you know, free academy.gov, but they don’t necessarily have access to things that are paywalled right now, but it’s doing the best it can.

And here’s what it ends up with about 10 minutes later. This is the paper it wrote for me. Now, you have a couple options. You can either export it straight to Google Docs, or it will do an audio overview for you instantly of what you need. So you’ll see here I could get that as a podcast instantly, or I just hit a button and it’s in Google Docs. And I just want you to take a look as this scrolls through here at the amount of research and how well it’s done this with tables. It’s done all of my citations. It’s done every section. And then you’ll see as we get to the bottom here, it has 163 references all referenced.

So know that your students know about this, and this is part of the reason we’re seeing everything change, right? We need to make sure a you have to make sure probably go through and make sure that everything is accurate in here and fact-check it because it could be wrong. But to have this level of a research paper written by AI in 15 minutes is going to change a lot about what we do. Obviously, the conversations you need to be having with yourself and your students have to do with evaluating the outputs or assigning students to fact-check or critique AI responses for bias and accuracy, comparing sources and analyzing the AI versus human expert sources on the same topic, and then figuring out how would different prompts get me different research papers or different outputs. So these are things to be thinking about.

Another question that we’re going to kind of talk about briefly is that we got presession was how do we avoid widening the digital divide in AI adoption? And again, this week there has been some advancements in this direction because OpenAI just released a couple of days ago publicly for free for everybody what’s called OpenAI Academy, which is free courses and seminars specifically for every group. There’s some for older adults. There’s some several for older adults, but there’s a ton for education. So introduction to ChatGPT Edu, mastering prompts, AI for academic success. These are all available brand new for you for free in the Open AI academy starting now. And also keeping in mind that now students have free access. These are things that were going to help um not make the digital divide worse in a lot of ways because the major roadblock is people understanding how to use it or having access to the best version. So yes, I can share that I’ll share that output with you too so you can take a look. But definitely try your own on a subject you’re interested in.

Someone asked in the chat if I could share that research output, and I absolutely will. Um, but I recommend the first one you do do on something that you know a lot about because you’ll be able to then scan through and tell where it’s right, maybe where it’s not quite right. But yes, I will absolutely share that out. It’s fascinating to look at, but it’s even more fascinating when you make your own. I was at a presentation in Oregon last week, and we just did it Mad Lib style, and I said, “Somebody give me two random topics,” and they gave me moths and they gave me sushi. And I had the deep research write a report on moths and sushi, and it was amazing. I had no idea. There was so much overlap. So, uh, definitely try it with something you know about and something that interests you and see what happens. And check out these courses from the OpenAI Academy. They’re brand new and they will be very, very helpful in helping you feel like you know what’s happening and give you a place to start.

So I want to spend some time talking about feedback and assessment because that was something that really came up in our pre-session questions, and I feel very strongly about some of the things in this section, and I think that you do too based on the questions that were asked. So just to give a quick recap or a scan based on all the speaking that I do and the groups that I train and I work with, this is kind of a quick landscape scan of where we’re at with AI in education right now. AI has destabilized trust in written work, right? Like we’re like, “Oh, did a kid write this? Did ChatGPT write this?” AI detection tools are flawed and very easily circumvented. I could show you in less than 30 seconds how I could get around an AI detector with a fully AI-generated content. Um, so if that’s something you’re relying on as AI detectors, send me a private message. We can talk about it. But there are a lot of issues.

The AI arms race is reshaping how we’re thinking about assessment. Teachers are having to double as system designers right now to try and figure out how to come up with these new ways to have assignments. We’re worried traditional essays are well, we’re not worried, but traditional essays are going to have to give way to a broader range of assessment formats that prioritize things like conversation, reflection, and reasoning over just pure regurgitation. And we’re shifting from this static evaluation to more dynamic personalized relational approaches to learning, and how we grade is going to have to evolve as well.

So one of the questions that came up pre-session is is AI helpful or harmful for students who are already falling behind, and the answer is both. It can be helpful in many ways. Again, I’m doing research. I’m I’m a student right now in half of my life, and I use AI for a lot of things similar to this. For example, you can give a a prompt to students where it can say something like, “Here’s my rubric. Here’s my draft. Evaluate this as if you were my professor and give me some areas for improvement.” And it will review it and evaluate the draft. Students could go in again to that GPT section and find all of those different GPTs that can help them with their learning if they’re so motivated to do so.

But with that said, there is a lot of research coming out right now about what’s called cognitive offloading, which is these risks that because friction is what gives us learning. When we struggle through things, that’s how we learn through friction. Now that we’re moving into these frictionless solutions where something can just give us an output and we don’t have to think about it, there’s a lot of problems when we’re getting these instant solutions without effort. There’s these learning gaps, loss of critical thinking skills. I have a QR code to the research here. And the what the research tells us is that high AI use can negatively impact students’ critical thinking skills. Something I’m sure we all suspected. Younger students are particularly susceptible. So, high school students. We need to be aiming for balanced AI integration, prioritizing active learning and critical engagement, developing metacognition skills related to AI use, and promoting educational strategies that promote critical engagement with AI.

So, how does AI work with assessments then? Well, using AI to grade assessments, I when I talk to a lot of educators, this is their number one biggest time-consuming thing is this idea of assessments. And so there’s a great opportunities, right? Like you can save a lot of time and brain power. There’s some some suggestion that it can protect students from potential bias. There’s these pros, but there’s also a lot of cons using AI. And I mentioned this earlier on, AI is never neutral. There is some sort of bias somewhere in it. And we saw this in I think this was 2021, maybe 2022, during the pandemic in Great Britain, they used an AI tool to grade their A-level exams, and it came out later that there was major flaws in the system, students from public schools were graded significantly lower than students in private schools just because of bias embedded in the algorithms and that they didn’t know about and they didn’t know was there, and it resulted in tons of kids not getting into the colleges they wanted, and then when they went back and corrected it, suddenly tons of kids got into college at the same time, and the colleges were overrun, it caused a lot of problems.

So knowing that an AI tool is never going to be completely unbiased ever should give you pause when you’re thinking about using AI for assessments because there’s bias and accuracy, and you can’t just say well why did you give this kid a 71 versus a 74 and this other kid an 86. You can’t ask the AI that. You can’t look under the hood and know or challenge it. And kids can’t challenge it, the outcome which is a problem. It can undermine a pupil-teacher relationship. There’s some research about that. And we risk then creating assessments just that will fit an AI machine instead of ones that will accurately assess students. So you can use things like this in partnership but not as a sole determinant. There are there are ways to do this ethically, but these are some of the issues that you should know about.

There is kind of a movement right now towards this idea of oh well let’s just grade process versus product. So for example, oh, Grammarly and Turnitin both have this program now where a student has to write within their program, and it records the entire process of the student writing and then plays it. You can see if a student typed fast and then slow and deleted things or copied and pasted in large amounts of data, and you evaluate the process that they’re writing. There are major ethical issues with that as well in terms of then are we forcing and combining kids to operate within a box? Are we putting them in a glass jar and just saying does that stifle creativity? Is it a violation of their personal processes? Do they operate differently within that space than they would in their normal way? Are we just teaching them to be more and more and more confined again so that it fits within an algorithm? So, some things to think about when it comes to that process versus product.

Something else to think about, especially when it comes to AI detectors, is these are some of the main concerns around AI and education that many different groups have brought up. And this is from a Carnegie learning report that came out last year, but students’ number one concern is being accused of plagiarism or abusing AI when they haven’t. And I personally have a son who loses sleep over this because he does not use AI to write things, but he is so panicked that he will be accused and will not be able to defend himself. And when I write something, sometimes I just go put it in an AI detector just to see what happens. And I am always flagged as AI generated because it’s trained on patterns, and my patterns are often very, very standard academic patterns of writing. And the last time before I handed in a major paper, I had this happen, and I was deeply conflicted. I and I recognized why students feel this way because I wasn’t sure if I needed to change the way I wrote and my voice and my style and my arguments to fit an AI detector so I wouldn’t get flagged or risk being accused, and it puts kids in a very, very difficult situation. So for those reasons, I I struggle with the idea of AI detectors, and I it’s important to think about the impact it’s having on students.

This is a great blog post that just came out. I shared it on my LinkedIn, but it’s highly worth reading, and it’s called How Teachers Are Trying to Outsmart AI, and it’s from a substack called How to Survive the Internet, and these are some strategies that teachers are using. So for for assessment and to prevent some of the cheating that we’re worried about, handwritten assignments, live writing in Google Docs where teachers can observe the typing behavior. You should know though that there is AI tools right now that students are using that will intentionally generate slow and delete and go fast and rewrite and are trained to fully mimic student writing. Students find ways around this. If you don’t believe me, go see the research that Dr. Sarah Eaton is doing at the University of Calgary specifically on these things. A lot of teachers are moving more to oral defenses or interviews to verify that students understand. They’re reframing essay prompts to focus on niche or reflective content. They’re rewriting their rubrics. They’re assigning students to critique AI-generated essays than just write them. Or sometimes they’ll use decoy essays. If a kid just ran something off of ChatGPT and turned it in, the teacher’s generating five similar ones and asking the student to identify which one was theirs. These are some ideas that are brought up on the blog. But some things that that people are doing. Again

This paper is about an educator-centered understanding of harms from large language models in education. I'll send this to you because I do want to have some time for questions. But these are some of the strategies, mitigation strategies, that educators are practicing to help avoid some of that harm, um, of LLMs in education and uh how they are adapting to this new landscape with students.

So, I want to show you just real quick. You can link to these once I show you this deck. These are some of my favorite AI tools right now: Chat GPT, obviously, it's the GOAT. It's the best. Especially those GPTs make it really, really great. There is something similar, uh, in Gemini called Gems, but Chat GPTs are the easiest to use right now.

Notebook LM, I cannot recommend this enough. Uh, this is a great tool for yourself. You could get yourself organized. You can do full lesson plans or syllabi from a notebook full of all of your content. If you've written a lot of stuff, you can source your own stuff and have it answer questions. I've done things like put in a huge piece of legislation that I needed to understand and just ask questions, and it would show me right where it answered my questions and cite that right in there. Sorry about the QR code. As we ask questions, I'll go through and grab these and drop them in the chat for you.

Microsoft Copilot is something—again, it's built on top of Chat GPT—but it's a great tool, especially because, uh, it has high levels of data protection, uh, within organizations, so I find a lot of educational institutions gravitate towards this. Claude—I can't say enough good things about Claude—and you should know this is an article from the New York Times, um, that the tech insiders in Silicon Valley—and Claude is what they use—it has like, the vibes are immaculate, if to speak Gen Z. Like it is, it has a great personality. It is viby and it is chill and it's like a good conversational partner. And so it's definitely worth seeing if it aligns with what you think.

Uh, if you're into Burt Reynolds like I am, there is actually a tool called 11 Labs that I use all the time, especially when I'm listening. I want to listen to uh complicated academic journals or go for a walk or have my emails read to me, and it will read it to me in Burt Reynolds' voice. The Medford School District, also known as Medford School District 549, and it has tons of voices. So you can have him read you or any of the voices read you anything. Uh, side note, I'm still waiting for Keith Morrison to get in here. If you know who that is, he is my favorite voice. Definitely, uh, can't wait for that one. But this one is called 11 Labs, and it's an app on your phone, 11 Labs Reader. It's amazing. Uh, also, it does amazing voice-to-text, and it is almost perfectly accurate. And I know this because I took one of the fastest rap songs by Eminem and played it for 11 Labs, and it got every single word perfect. So I tested it very scientifically. So now you know.

Uh, also, if you want for fun, and if you take nothing else away from this conversation today and you're still skeptical about AI, go to sunno.com and have it write you a song. So I did a webinar yesterday for a a PR firm, and their company was called Kirk and Co., and I had it write a '90s hip-hop song about their company, and it took 10 seconds. [Music] So again, this is a free tool. You can write a bedtime song for your kid or a birthday song using all their favorite things in whatever style you want with Sunno.

Grok is one by Elon Musk, and uh, it's pretty great in a lot of ways. The image generator is great. What makes this unique is that it can answer questions in real time. You can say, "What are people saying right now about this bill or this issue?" and it will pull from Twitter data in real time and tell you exactly what's happening in the moment. Uh, Perplexity—someone brought up—definitely a tool worth trying. It's like if Google and Chat GPT had a baby; it will find you news sources everywhere and write you a summary. Midjourney for image generation is amazing. This is one of my sons. This is actually Chat GPT 40. Just got a brand new image generator for free to generate images right in Chat GPT. So, you should definitely experiment with some images. Ideogram is another image generator.

But your top takeaways for today: We're living in this transition era. AI can really be a pedagogical amplifier and not a replacement. Try one tool or use case that aligns with your goals. Make sure that your pedagogy leads AI and not the other way around. You have to be thinking with that ethical foresight, but remember that great education is the result of great teachers leveraging technology, um, no matter how advanced the technology, it's not a substitute. We lived for years in this goods economy where we were working in factories and we were farming, and technology, uh, took some of that labor on for us. Right now we're in this knowledge economy where now technology is going to take some of that knowledge labor on for us, which means we're going to be moving into this relationship economy. And in the past jobs were about muscles, and now they're about brains, but in the future they're going to be about the heart. So, make that the focus of absolutely everything that you do. And find me on LinkedIn. Shoot me an email. This is my favorite thing to talk about. Happy to chat with you with any questions. And I'll make sure that you have all those resources. I'll send them to Sophie, and she'll send the deck and the links to the papers and everything. And I know I talk fast, and I know that's a lot. So, thanks for sticking with me.

Thank you so much, Rebecca. This is amazing. I always use this, um, insight per second as like a metric for measuring a good webinar, and I absolutely loved your presentation and all the tools you recommended as well as the immaculate testing you did with each tool. Thank you so much, and I do want to open the floor to some Q&A. Um, if you have some questions for Rebecca, please feel free to raise your hand and unmute yourself, or you can also type your question in the chat and I can bring that to Rebecca. Oh, I see these questions, too. Amazing. All right.

Oh, resistance when working with faculty. I see this a lot. Like, this is so hard. And honestly, with educators, it's because they have so much to do. I find that sometimes the solution is just showing them like one small thing that's kind of fun or one way that it can make their job easier or something that's already a pain point and just kind of getting them to dip their toes. And I find once people find something like that, it gets a lot easier. But I think making it engaging and finding one way that is directly applicable to them or their role is usually what I find gets people in the door. Um, figure out what their biggest pain point is; show them how to use AI to maybe help with that one thing, and then you can kind of start. And whoever said find your early adopters—100%—figure out because, like I said, I worked in communications in in education for years, and this—as soon as I saw this—I was immediately hooked, and I wasn't the teacher; I was like a director; I was, you know, with all the superintendents and everything, but I wasn't the typical person to be teaching people about this. Figure out who's already into it in your office, in your district, find them and leverage them because the people who are into it are really into it, and you can't teach that kind of passion. So if you have early adopters, nurture those people.

Yeah. And and thank you for this. This is really good. I'm looking forward to reading this article. Providing content in multiple modalities is helpful for all learners. Absolutely. There's definitely exceptions to all of this, and I definitely plan on reading that article. I'm always looking to learn more about things like this. There's definitely, uh, different ways and different learning styles that work best for different things, different modalities. Thank you for sharing that. Improving rather than fixing is definitely helpful. Language—again, that's my that's my thing—words and language. Another strategy for you is to take your exact question and put it into AI and ask it for suggestions and feedback.

Um, yeah, a new generation of assessments. Uh, yes, actually I have. Um, Billy, will you shoot me a message on LinkedIn? I will share some resources with you that I just came across yesterday for assessment stuff that, uh, I've been taking a really close look at. Uh, yeah, because it's hard to teach kids about AI until we fully understand AI and how to use it. Uh, online asynchronous content definitely more challenging. Uh, Shelley, reach out to me. I'll I'll share with you some stuff that I've seen on LinkedIn. I'm not necessarily an expert in this, but I can share with you what I've seen other people doing on this because yes, asynchronous courses is definitely more challenging, especially—yeah—online learning and adult education courses—very, very common.

Let's see. Did we miss anything else? Any other questions? Thank you very much for all the questions and comments in the chat, and thank you for all the great insights and valuable lessons from Rebecca. Um, while some people are thinking about what question to ask, I do want to share the QR code of our feedback form. And if you have anything you want to make sure that we will address next time or if you have more questions for Rebecca, please fill out the form so that we'll we'll make sure to address them. Oh, I lost my chat. We have a question. Let's see, from Shelley: How do we help students develop the middle oral part that uses their human expertise? They're still learning and thus don't often feel confident about their expertise. And what strategies do you recommend? What evidence would you ask students to submit to prove they fact-check the AI output as they use it?

Great question. I've seen this done in a lot of different ways. You're right, though. They don't have the expertise, but still we encourage them to move completely away from AI and do that 60% in the middle on their own. I've seen teachers who have, uh, students hand in—maybe there's a way that you can link to your prompt thread that you used in AI that shows the conversations that you can kind of evaluate the questions they were asking and how they were, um, arriving at proper conclusions or asking the right questions and just kind of evaluating again some of that process—but I would do it in a way that maybe they can use expertise in areas that they do have, if it's minimal, or an interest that they have, so they can have practice doing it, but at least writing or doing things completely away from the AI in that middle part. Um, that's kind of the most important part is not relying on it for every single part of every single thing that you do. So, I wouldn't—it's tricky. I've seen people have them submit prompt threads mostly, but also actually just write, uh, write about their process, what they learned, and grading that or having kids give verbal presentations about the process and how they arrived at certain conclusions, uh, for that middle part. So hopefully that's that's helpful. But I again I have seen some people doing some really interesting things in this space online. So, if you want to shoot me a message, I can point you to a couple of people on LinkedIn that I've seen doing some interesting things here.

Thank you for this continuum, Christina. This is actually great. So, if you guys—Sophie, if you want to grab that link, I would share that out after this session as well with the resources. Christina, that's fantastic. It's a great continuum. Absolutely. We have a lot of great links shared in the chat, and we'll make sure to compile it and share that with um all the attendees in our post-webinar email. And thank you so much, Rebecca, for being so generous with your resources and knowledge. And I hope everyone really enjoyed the webinar, really enjoyed the presentation, and I hope there's one tool that you want to try out after this webinar. And on that note, I want to thank you everyone, particularly our speaker Rebecca for this amazing experience. Thank you. Thanks everyone. Thanks for being here. Thank you. Take care. Take care.