Guide to Integrating GenAI for Deeper Literacy Learning

An Expanded Set of Pillars for High-Quality Literacy Content blog post

SAP’s e² Instructional Practice Suite™

Robert Bjork research: Memory and Metamemory Considerations in the Training of Human Beings (1994).

Prompt Framework for Educators: The Five "S" Model

Amanda Bickerstaff

Amanda is the Founder and CEO of AI for Education. A former high school science teacher and EdTech executive with over 20 years of experience in the education sector, she has a deep understanding of the challenges and opportunities that AI can offer. She is a frequent consultant, speaker, and writer on the topic of AI in education, leading workshops and professional learning across both K12 and Higher Ed. Amanda is committed to helping schools and teachers maximize their potential through the ethical and equitable adoption of AI.

Carey Swanson

Carey is a Director on the literacy team at Student Achievement Partners. Prior to joining SAP she was a school leader and classroom teacher for many years. Carey has also had consulting roles supporting leadership development, instructional coaching, and professional learning. Carey has a BS in film from Northwestern University, an MS in Teaching from Pace University and an MEd in School Leadership from Teachers College, Columbia University. Carey is passionate about fulfilling the promise of learning that public education offers, and aware of the ways in which current conditions fall short of this promise for many students. She believes collaborative work, centered on students, focused on reading and research can advance equitable instruction and move us closer to fulfilling this promise.

Jasmine Costello

Jasmine is a Product Manager at Student Achievement Partners. Prior to this role, Jasmine was a project manager, education researcher, and classroom teacher in Philadelphia. She has an MS in Effective Teaching and Learning from the University College of London where she attended as a Fulbright scholar. Jasmine is committed to finding creative solutions that serve students who have been historically and systemically underserved by our country’s education system. She is passionate about working towards greater educational equity and creating tools that support all students to thrive as their authentic selves.

Mandy DePriest

Mandy is a Curriculum & Content Developer at AI for Education. She has over 15 years of experience in public education, having served as a classroom teacher, library media specialist, and instructional coach. She has also taught education technology courses in higher education settings as well as professional development workshops for teachers on the transformative power of technology. She is committed to ensuring that students are prepared for the dynamic demands of the future by leveraging the power of technology-driven instruction.

Amanda Bickerstaff
Hi, everyone. Hello. It's going to take a moment for everyone to get in, but we're really excited to have you here today with us. I'm Amanda. I'm the CEO, co founder of AI for Education. And we have a very exciting webinar today, all focused on AI and literacy. So we'll give everyone just a moment to get in. It's going to be a big group. We had almost 2,400 people sign up for the webinar, so. So it should be a really nice time together. But I just want to say hello again. I'm Amanda. I am the CEO and co founder of AI for Education. Today is really exciting.

Amanda Bickerstaff
We're going to be diving into an enormous topic that is coming up all the time, especially in our ELA classrooms, about how can we start to think about using generative AI in ways that could potentially deepen literacy learning itself. A lot of times we hear that ELA teachers are. Our English teachers are a little less comfortable with this technology, but we know that is actually not the case all the time. We've seen really great early adopters and also just so many cool applications that we're really interested in digging into today. I'm going to be joined by three amazing ladies. We got Mandy DePriese from my team at AI for Education, and then we've got Carrie and Jasmine from Student Achievement Partners. So what we're going to do is we're going to move to the next slide. We're almost 300 people, everyone.

Amanda Bickerstaff
And you already know if you've been here before, like Kat. Nice to see you, Kat. And Laura and Eric. But what you do is say, hello, there's a lot of us. This is your community of practice. Our favorite thing about these webinars is coming together, of course, sharing our knowledge, but also having you all share it with each other. But because we are already over 300 people in the chat, what we want you to do is make sure the chat is about the conversations you have with each other. And we will be putting in resources and the guide itself and all these different pieces to support you. But also that's where you should be sharing your resources and ideas as well. But I will say that we're going to do a little bit of prompting today.

Amanda Bickerstaff
Mandy's going to do an example from our new guide, but also, if you have a question for one of us. So for Mandy or Carrie or Jasmine or myself, please use that Q and A function, because we won't be able to follow the chat all that closely. Because it's moving very quickly. So that Q and A function is going to be how you are sure that I see it. So what I'm going to do is we're going to talk a little bit about why we're here today and we can move to the next slide. Is that this is pretty great. We at AI for Education are an organization that is focused fully on the responsible adoption of AI, generative AI specifically in our schools, whether that's K12 or higher ed. And we have learned so much.

Amanda Bickerstaff
We have seen so many people kind of grappling with these big questions about how generative AI is going to potentially transform learning. And we often think about how can we support that. But we are not. We are all former educators, but we're not literacy experts. But what I'm really excited about is through our partnership with student achievement partners who are going to be those experts in ELA literacy and also math and instruction, high quality instruction. We just wanted to start navigating, like, what can we do together? And I'm going to call Carrie and Jasmine and Mandy up on the screen with me so we can start to like contextualize this with the people doing this work.

Amanda Bickerstaff
But what we wanted to do is like, could we combine our two companies, our two sets of expertise to create something that we hope is going to be really supportive? We have created for the first time ever, a pedagogical guide to generative AI in literacy classrooms. And what's really exciting about this is that I will say the caveat here is this is a first version. It is going to be our best shot at this to start taking the use cases that we see that are already happening and some opportunities in RVI and then layering them on these high quality instructional practices that lead to students deepening their cognitive load in that productive struggle that we know is so important while not offloading and. Or just having some counterproductive struggle that's not healthy.

Amanda Bickerstaff
So we want to make sure that we're looking at with this guide, what are students learning? What is that productive struggle? But then where are those places that generative AI specifically could potentially help? The guide itself, which you'll see in a moment, is going to have. It's really going to be focused at the level of once you've built AI literacy and then starting to do this either as a modeling activity or with your students and in your instruction. But I want to say we're going to introduce everybody right now, so I'd love to have we're actually going to start with Carrie. Carrie, can you introduce yourself? And then we'll go down the line from Jasmine and Mandy.

Speaker 2
Yes, absolutely. Hi, everyone. So nice to have so many people here talking about this topic. My name is Carrie Swanson. I am based in Ann Arbor, Michigan. I was a former teacher and school leader in New York City for a lot of years, and I work on lots of really exciting literacy projects at Student Achievement Partners.

Amanda Bickerstaff
Hi, everybody.

Speaker 3
Yeah, thanks so much for joining us today. Super excited to dig into this with you all. My name is Jasmine Costello. I'm a product manager at Student Achievement Partners. I'm based in Philadelphia and was a previous teacher here in Philly.

Amanda Bickerstaff
Right. And then Mandy.

Mandy DePriest
Hi, everyone. Great to be here with you today. I'm Mandy DePriest, content and curriculum developer at AI for Education. But prior to that I spent 15 years in public schools in Arkansas where I spent six years as a fifth grade literacy teacher.

Amanda Bickerstaff
So.

Mandy DePriest
And I'm a former librarian as well. Literacy has always been my heart. I'm excited to see how our two platforms can merge and produce something really exciting.

Amanda Bickerstaff
And I think that this is really the goal here is like, how can we take some collective knowledge about these two areas and start creating a set of, like, thoughtful intentions around this work? And actually what you'll see is if we can pull up the guide. I think it might be worthwhile pulling up the guide just really quickly. So here's a resource. We're going to drop this in the channel, but Mandy, if you can click on it, I just want to walk us through just really quickly of how the guide was developed. And what you're going to notice is that this is meant to be a collective process. So what we have here, and thank you, first, I want to shout out some people. We had Dan who put this together from our team.

Amanda Bickerstaff
We had a lovely copy editor that helped out to make sure that you as ELA teachers will not get mad at us because we misspelled anything. I'm sure you'll find it if we did. But what we'll see is, if you scroll down, Mandy, is what we have created is this guide to iterating Generative AI for deepening Literacy Learning. And so we actually set out the goals of the resource. This is designed to start having you having literacy teachers actually be able to start implementing this and giving us feedback about what works or not. And in fact, at the end of the document, we do have a form where if you decide to use this or you read it and have feedback for us, we want to make sure that we're actually incorporating it.

Amanda Bickerstaff
And then the next version of this guide will incorporate those best practices or those evidence bases. And if you roll down, one of the reasons why this exists is that chatterty is happening and there are very few evidence based approaches yet. So what we want to do is go back to the literacy evidence, right, of what is high quality literacy instruction, and then layer that on this so we have some key concepts also, again, I just want to point out, if you're going to be doing this with students, this comes after basic foundational AI literacy has been done at the teacher level as well as the student level. And then this reinforces it, but you can't skip a step that's going to be credibly important.

Amanda Bickerstaff
And then if we keep going, what we'll see is that we have, and this is what's really going to be taken through, is we're talking about the difference between productive and counterproductive struggle. And as we have Chris in the chat saying that no pain, no gains, ideas absolutely on the right track. And then what we have is some high quality instructional, like literacy instructional content and approaches, which is what we're going to first focus on with Carrie and Jasmine. And then there is a section on what does it look like when you start thinking about that from the perspective of integrating generative AI. And the first thing is of course, making sure should you even.

Amanda Bickerstaff
We're not suggesting that you do it all the time, but if you do, here are some examples of a best practice approach which is going to enhance the student's cognitive load, their cognitive work, and then a version where it's very easy to offload that because we now have that magic essay button, everybody. And so what we want to do is see that together. So what I'm going to do is we're going to actually shift it over and the first section, we're going to talk about those high quality content strategies. So Jasmine and Carrie, take it away.

Speaker 2
Thanks, Amanda. So our starting point for this conversation, that was the starting point for the resource and honestly is the starting point for all the work we do at Student Achievement Partners, really offering a sort of definition for what high quality literacy consists of. And I think that might sound a little bit basic if you're not kind of in the nuance of teaching and learning. But we actually know that whether you're an ELA teacher, a reading teacher, writing teacher, an elementary teacher, kind of understanding what that means in terms of the content you teach really varies from school to school, classroom to Classroom. And to combat that, Student Achievement Partners has really kind of done some work around what does it mean to teach literacy content in ways that are equitable and rigorous. And we have outlined these five buckets.

Speaker 2
So we're looking at foundational skills and what allows you to become a fluent decoder and be able to have fluency with grade level text over time. We're looking at resources where the content itself is in service of building knowledge about the self, about others, about the world. We're talking about content where we're making use of instruction diverse and complex rigorous texts as the source of what we're studying and talking about in the classroom. We're talking about developing oral language skills and listening skills for all students. In particular, really thinking about that oracy development for multilingual learners within this bucket. And then we're talking about writing, both volume of writing, quality of writing and range of writing. So we have a blog post if you want to learn a little bit more about these five areas.

Speaker 2
But as we're thinking through all of these kind of optimal uses of genai and how it can support instruction, we'll be referencing these five areas as our focal points.

Speaker 3
Yeah, thanks Carrie. And also if you want to dig in more into our definition, our expanded definition of high quality instruction, we have one for literacy and math. You can visit our E Squared page that I think Tori will put in the chat. Yeah, perfect. So that has the, our frameworks which are the foundational research that defines this high quality instruction that all these pillars are grounded in. So we are going to now talk a little bit about how this shows up in. Yeah, in like these AI tools that we're seeing really come onto the scene and the relationship with cognitive lift, as Amanda mentioned at the start. So we know there's so many tools and resources that are evolving in the gen AI space.

Speaker 3
And while there's a range of quality of those tools, many of them are really exciting and have great potential to support high quality teaching and learning and they can really increase efficiency and effectiveness, which we know is priceless for teachers and educators that have little time and need to focus on what's most important. So as we touched upon that, it's essential these AI products that we're seeing pop up in the education space enhance our work by supporting students cognitive lift rather than replacing it. So basically we have to ensure that these AI tools are not positioned to do the thinking and the work for the students. So if we dig into this a little bit more, we can start by defining what we mean by productive struggle and counterproductive struggle.

Speaker 3
I know many of you have thoughts of this and your own kind of definitions of this as well. But just to make sure we have shared understanding, we know that productive struggle is really meaningful interaction with challenging grade level content that supports student learning, pushes students to expand their thinking and grow in their skills and their knowledge. Right. Well, counterproductive struggle is when students are challenged by the work in front of them, but they don't have adequate support to be able to access that content. Content and make progress. And so this can, as we know and have seen, can lead to frustration for students and actually be really damaging to their self esteem and identity as learners.

Speaker 3
So when we think about positioning these AI tools in teaching and learning, we can use this as a helpful frame of like, is the tool helping to alleviate a student's counterproductive struggle while not replacing a student's productive struggle?

Speaker 2
Yeah, absolutely. Yeah, there's a lot of examples in the guide, I think, just kind offering an illustration of this, of what Jasmine was talking about. We know there's a ton of tools right now that will kind of level text at the touch of a button. And we also know that when those tools are used to constantly limit the text that kids are exposed to, we actually create some real inequities in the classroom. If kids are only exposed to low level text, they're not actually experiencing the kind of complex, rigorous syntax and ideas, especially when some kids are always limited and other kids aren't. So that's kind of an example of that leveling actually being counterproductive and kind of reducing productive struggle in the classroom in a way that creates inequities.

Speaker 2
What we're excited about in terms of this resource and getting to learn collectively from lots of educators about how it kind of lands in real classrooms is we've really tried to throw out some common use cases that we know exist already, show or demonstrate some of those counterproductive examples, and also offer some guidance around more productive examples. So Mandy's actually going to get to walk us through some of those in real time today.

Amanda Bickerstaff
I would love to. I just want to dig into a couple of things before we go over and just from the chat as well. So we talk. Let's go back, Mandy, if you don't mind, let's go back and let's talk about this idea of counterproductive and productive struggle. So we talked about the skill sets. You can think of this in other ways as the zone of proximal development that amount of challenge that is just enough, but not so much that it's going to discourage students or be unmanageable. But then the other side, there's also. Mandy, do you want to talk a little bit about Robert Bjork's idea of desirable and undesirable difficulties?

Mandy DePriest
Yeah. We came across Dr. Bjork's research when were researching how to redesign assessment in the age of AI. And it's this really helpful framing that helps you identify those points where it's described often as friction that is positive for students. It's good for them to grapple at the edge of their ZPD because that's how they're gradually going to build up more complex skills. But there are also obstacles like things like a lack of background knowledge or translations or things like that might just be getting in the way. And those are not productive struggles. And like Jasmine was talking about earlier, can lead to feelings of insecurity and poor sense of themselves as a learner. And AI has real potential to alleviate those while the teacher is able to make sure that it is zoomed in on those aspects of productive struggle, that positive friction.

Amanda Bickerstaff
And I'll say that like I will, like right now I do want to kind of identify. Someone said in the chat this idea that the technology itself not being responsive, like gen AI is really new and kind of inconsistent and unreliable and it can be quite frustrating. But it also can be on the other side. Even learning how to use generative AI correctly is productive struggle because it requires us to have resilience. It's learning how to approach things, how to reframe, how to give feedback. How long have we been trying to create kids that are great kids, support kids to become better questioners, to ask better questions and know how to do that. But I also want to take one step back as we talk about skills, but a lot and we talk about content, but also we really want to talk about mindsets.

Amanda Bickerstaff
And the idea of mindsets here is that if I give it all away, like if I am a student or even a teacher and I'm giving it all away, I'm kind of pushing this into offload. Like, is that what I need to be doing? Like, will I be able to do the next step? Will I be able to meet my own goals? Is it going to be meaningful? Because that decision making process of what we need to keep versus what we can start to give away when we build fluency and capacity is something that's going to be a bigger and bigger question as we go forward. So we'll now kind of lift over to Mandy's piece where we're going to really dive into these worked examples.

Amanda Bickerstaff
And can I just say, like, one of the things we love about this is that we are learning with you all. And so Mandy actually. So the way that we designed the guide is that Jasmine and Carrie are going to be the SAP experts on literacy. All of those high quality content approaches which you see here around foundational skills and fluency. And then Mandy and the team actually went and, like, tried it out. So Mandy actually went and created worked examples. So in the document, you'll actually find chats that Mandy did in real time that have been slightly edited for, like, ease of reading. So you can actually see how to do it with a chatbot. So we just want to say thank you to the team for really coming together in this unique moment to kind of create this resource itself.

Speaker 2
Yeah, I was just going to say it's like the guide. It's very meta. The guide included using the guide. You can see how we've used the guide and then you can give us feedback about how you're using the guide and then we can refine based on what we've learned.

Mandy DePriest
Yeah. The thinking behind these worked examples, Amanda, if I could just elaborate on that a little bit, is that it's one thing to just say to teachers, oh, you use AI to create leveled text or use it to chat with a fictional character or whatever, but without like a deep, you know, experience of using the tools, especially if you're new to the tools, that can be a little vague and hard to visualize what that would look like. So with these worked examples, what we wanted to do was give you just a concrete, like, it's just a chat, literally. And you can see what I put in to the chatbot, its response. And then we also provide a little explanation as to like, what my thinking was. You know, how I'm going to formulate my next response to the chatbot.

Mandy DePriest
And some of them are quite long. If you stick with them, you'll really get a sense of how to meaningfully interact with these bots. Because the magic happens when the teacher applies their professional judgment in coordination with the chatbot. These tools do not, like Amanda said, we don't give it to the bot, like by themselves. They're not going to be able to deliver any kind of meaningful instructional impact. It has to be the teacher's expertise and knowledge of the student coming in and working with the bot in relationship to create content that is going to be directly Relevant and meaningful for students to the individual level.

Mandy DePriest
I remember when I was teaching fifth grade literacy, I had, you know, 75 students and they were at 75 different levels and they had 75 different interests and it was so hard to find materials that were going to be engaging for them. And now I'm able to like generate passages about robots and slime and unicorns and whatever else we're all interested in, at whatever level we're interested that we need to access that text. So should we look at one of the worked examples, Amanda? Should we dive in?

Amanda Bickerstaff
Actually, can we just go back to the resource again? I just want to situate this with that. Also. I want to be a slime unicorn. And I've never heard those two words together in this way, but I'm sure.

Mandy DePriest
I had to make a passage about that.

Amanda Bickerstaff
I do use unicorns quite a lot, everybody. But I like. Do you want to just show, like for the worked example, let's actually show this here and let's make it a little bit bigger. Mandy. And so what we have here is this is the really the dig deep, dive into these, these kind of high quality instructional practices and content. And then what we've done is you may want to walk through the structure of these so that people can understand how to use the guide.

Mandy DePriest
Yes. So we have an example of the, those pillars that Carrie and Jasmine were discussing, the foundational knowledge, the, you know, building knowledge, et cetera. And then we have identified a use case. There could be definitely more. Use your creativity as you think of, you know, other ways that you might use AI, get those wheels turning. And then we have the worked example of that particular use case. But we also have more broadly, like, what would best practice look like in this foundational pillar and what would practice we want to avoid look like? So it describes what the teacher is doing, what Genai is doing and what the student is doing. And the hope there is to really articulate how the teacher needs to stay in the loop and how the student needs to be engaging. Like, everyone's a part of this. Right?

Mandy DePriest
The students aren't just passively sitting there and letting the instruction wash over them. We need to communicate to them our expectation for how you're interacting with Genai and what you're responsible for doing. And then we have the negative example, the avoid, because we want to make sure people understand. Like, yeah, it may be tempting to just kind of knee jerk, okay, I'm just going to do this, generate a text, whatever, without applying my professional judgment to it. Without, you know, situating it in the larger instructional context. We know you're pressed for time and it's very easy to do that and there may be times when you just need to do that on days when you need a sub or something and you just need some materials out there.

Mandy DePriest
But just so that you're aware of that and you're able to be strategic about how you're using Gen AI. So, but yeah, click here is where you would access the worked example for each of the pillars. We have foundational skills and fluency, building knowledge, leveraging complex texts, all the ones from the slide. So are we ready to click in?

Amanda Bickerstaff
I'm on mute. Yes, let's go for it. Okay, so just really quickly for those in the chat though, and everyone, I love how this is opening up like the bigger instructional kind of quality pieces. And so while this is very literacy focused, hopefully some of this you would be able to like abstract a bit as well. But yeah, I think that this is really interesting how quickly something like this we can start seeing how it could have bigger impacts on just teaching and learning in general. Did that work? Mandy, did the worked example come up?

Mandy DePriest
I clicked on the wrong one, but I wanted. I wanted to talk about the text set one because I think this is a really helpful use case. When I was in the classroom, I always needed text, like every day it was like grist for the mill. And I had, you know, certain texts that I could access through my curriculum and I had certain places online where I knew I could go to access texts. But then like, it was just constantly needing more and oftentimes they weren't exactly what I needed. Sometimes I would even write my own texts. Like if I we did a novel study where there was a very specific historical event and I couldn't find anything exactly what I needed. And so I like literally did research and wrote a text for my students myself.

Mandy DePriest
And that was before, you know, we had access to Goodgenai that could do that for you. And I loved, when I was looking at SAP's resources, that you have a lot of good stuff about text sets specifically and building up that level of complexity. Because the difference between generating a text set versus different generating just a level text is that with the text set, we have this kind of connection around a topic, but also this gradual increase in complexity. So the way we structured this example was to start out with kind of a very basic text that is meant to build foundational knowledge at really accessible reading level. And then each text is designed to build on the one that came before it to increase the level of vocabulary, the complexity of the sentences.

Mandy DePriest
As we have built our background knowledge, we're getting into more complex concepts. So they all work together to create a cohesive thing. So let me pull up the worked example.

Amanda Bickerstaff
And while you're doing that, Mandy, Carrie or Jasmine, do you want to talk a little bit about, like, the research behind that?

Speaker 2
Yeah, it comes from a couple. A couple of places. So we know there's lots of research around text complexity, and that was kind of a foundation, part of the Common Core Standards in the United States. I know there's folks here from all over, but kind of really outlining grade level complexity that can lead to college or career. At the same time, we know just putting hard texts in front of kids isn't the move actually that creates a lot of counterproductive struggle where kids are just kind of facing barriers that might be about decoding, but might also be about content that's referenced there. And that really, you know, those of you who are secondary teachers probably extends into every subject.

Speaker 2
So there's also a lot that we know about how supporting vocabulary and knowledge development through engagement in multiple texts on the same topic or resources can really help to scaffold that vocabulary development and that contextual understanding. If this is newer to you, like a really like small grain size example is the novel the Watsons Go to Birmingham takes place in the civil rights era in the 1960s. And while it's fiction, there's lots of historical references that if you know those already, you're going to read that text differently. Text sets can really help kind of create that same kind of access for all students and across all subject areas.

Mandy DePriest
Yeah. And in this worked exam, I'll show you how these worked examples are set up. We have kind of an introductory piece explaining the pedagogical value. There is a link to the original chat in each one. So you can see my actual conversation I had with the chatbot. And then in this particular instance, I ended up creating a supplementary resource with all the text that it generated for the set, which I'll show you. That'll be the grand finale, I think. But you can see in my first prompt. Here, let me make this a little bigger. There we go. I started out with this. I explained I'm a seventh grade language arts teacher. I'm working with a seventh grade science teacher to design an interdisciplinary unit on energy transfer and conservation. This is what my seventh grade son is studying currently.

Mandy DePriest
And I told it I would like to create a text set designed to increase students Vocabulary knowledge and ability to read complex grade level texts on the topic of energy types. And I kind of give it some more information. I want no more than six text. I get really specific here. The more you set up in the prompt, the more likely you are to get a relevant and accurate response. And then I tell it to determine complexity. Consider things like length, density of information, level of academic vocabulary and sentence variety. And each text should sequence to support students ability to read the next passage independently by building their vocabulary and background knowledge. So I'm being really, I'm not just asking it generate texts about energy transfer.

Mandy DePriest
I specifically want these texts to build on each other and then notice my last sentence in the prompt after I've given it all the context is ask me any questions you have one at a time to gather further information that will help you create the text set. So that way if I missed anything, the chatbot can identify that and pull it in and use it to inform its response before we get into the serious text. So we love this technique to help the bot build its understanding to make sure you get a good result. So it responds. It's very complimentary. This sounds like a well scaffolded approach. And then it starts asking me those questions. It's important to tell it to ask the questions one at a time.

Mandy DePriest
Otherwise you will get a big list of like 10 questions that you know it can't answer all at once. So this way it's more back and forth and conversational. So it asks me, would you prefer the text to focus mainly on scientific explanations or do you want like some narrative, some historical perspective? And then I drop in an explanation like this is why I did the prompt the way I did. This is what I think about the chatbot's response. In this case, the teacher prefers to keep the text focused on scientific explanation. It's not that narratives and historical perspectives aren't valuable. I just don't want them for this particular text that maybe I'm going to generate another text set around that later.

Mandy DePriest
So just so you can kind of see the thinking that goes behind it and really understand like this process so that you can learn to replicate these yourself. So then I put into the bot, let's just stick to informational texts. It continues. Got it. Do you want it to include visuals or diagrams? I did tell it. That would be great. Spoiler alert. Ultimately it forgot to include those. So someone like a chatbot like Claude might better at that than Chat GPT anyway. But it's it asked and then we just kind of keep going. It asks me a few more questions. I tell it no, assume no background knowledge for the initial text. We're going to keep everything approximately to seventh grade reading level two. It didn't ask me that, but that's something I want to make sure to include.

Mandy DePriest
So we kind of keep going with this back and forth. I tell it exactly what I want, and then it kind of stopped. It said, if you have any other specifics you'd like included, please feel free to share. I think it's. That's it for now. So I had to tell it to pull the trigger and generate the text set. Otherwise it would have just sat there indefinitely. We're working with a tool guy. It's just a program. It's not a person. So I did have to tell it, generate that text set now. Okay. And then it gives me an outline. And this is not really what I want. I want the actual texts. So I tell it, I'd like you to generate the actual text for number one because it's already gone ahead and outlined six.

Mandy DePriest
So let's take it one at a time, just focusing on text number one. And it gives me what is energy. Remember, with this one, I wanted it to be super foundational, building that basic vocabulary. I wanted it at a fairly accessible reading level approximately seventh grade later, I could go in and generate like, maybe some lower reading levels if I wanted to, if I need that. But right now we're just establishing a baseline. So it talks about potential energy, thermal energy. I would confirm all of this with the science teacher I was working with to make sure that the information is accurate. From what I know from working with my son's class, helping him with his homework, this is all on target. And then I get to thinking, you know what, it might be good to have a glossary of some of these terms.

Mandy DePriest
So I tell it include a glossary that allows students to capture and express their learning as a way to work through the text. And so it revises and it includes the glossary. At the end. We have this glossary of terms. So from there we just kind of keep going. I did ask it to suggest some activities for learning and expression, but like, ultimately I kind of abandoned that. We keep going and then it generates. I ask it for the next text and the next. And from that we got. Hang on. I ended up with. I ended up with 5 because I felt like the 6 text was not especially. Oops, that's not adding to anything. So I ended up with a text set and I wanted to show what you could do with this once you created that. Here we go.

Amanda Bickerstaff
So, Mandy, I'm. I'm just going to jump in just for a second, because we do have some questions in the chat about, like, the way that you use the bot. And so I know we. Can you. Can you actually pull up the chat, though, Mandy, just so people can see that this actually did happen? Because I don't want. I don't. Sometimes it feels like we're baking a cake on tv and we're like. And then. But what you will notice is a couple of things. Number one is that Mandy was using great prompting technique, and she was also being incredibly resilient because, like, oh, sure, visuals. And then it just straight up didn't do it. So, you know, or when the question stopped, she knew to ask the question. So those prompting techniques are things that.

Amanda Bickerstaff
That's why were saying that building your own AI literacy first and using the tools will make this better. But you also can start to see, like, this is why our prompt library exists, is that you can see those best practices. Mandy starts with a really fulsome prompt that flips the script to the bot, asking her questions, which is a. Is a. A really great way and great strategy. But then you can see that resilience, those questionings, those reframing, and that's going to be where we get to this place. And then now, if humanity can show those tech sets, it didn't mean that she hit a button and all of a sudden she had five texts. The cognitive load and the protective struggle. We're not just talking about students. I know. I. We talk about this all the time.

Amanda Bickerstaff
We do want you to save time. We do want you to be more efficient, but not to that extent where it's actually going to remove the value of these tools of actually not just reinforcing your existing practices, but going further and farther for your students. And so, Mandy, how long do you think this took you? And I know you were trying to make it perfect for this, but, like, how long do you think it took you to get to this level altogether?

Mandy DePriest
It was about an hour from when I started prompting to formatting this document. I was just copying the responses of the final text and dropping them into this Google Doc. I put the fun little header in. I bowled it out, the titles and I made sure each text was on its own, like, started on its own page so that I could print them separately. So I can assign these to students digitally as a complete set. I can print out physically. I could break it up into separate documents and just give them one at a time using my learning management system digitally, I have a lot of flexibility into how I want to get these two students. So, yeah, altogether about an hour to end up with five texts that I can continue to reuse, you know, through subsequent years and units.

Speaker 2
So, yeah, I'm really glad you brought that up, Amanda and Mandy, both of you, because I think the point is also thinking about instructional purpose in an ideal world. You're going to have a lot of these resources already in the instructional materials that you have in your classroom. But we know that we don't live in an ideal world. And how true that is for all of you who are here today is probably variable. I had seen a question earlier around how we recommend pairing these kind of suggestions with high quality instructional materials, and I would say only to fill in what the materials don't already offer that really get at the instructional need of your students. So in this case, this was really about access to some very concrete content knowledge that will help them then engage in the rest of the lesson.

Speaker 2
It's not meant to replace the rest of that lesson. It's actually meant to kind of lift up that shared understanding that all students need. So that instructional purpose piece comes from you. However, I think that like, one hour of Mandy's time is so much less time than what she had described that she used to do, which is kind of googling and trying to find or looking for resources and kind of trying to find those kind of references. So really thinking about that instructional purpose is important and that's great.

Amanda Bickerstaff
And also, I just want to kind of point out a couple things, like there are tools that do this for you, like there's in the chat you have, and we always are going to focus on showing you what's under the hood, because while you might use another tool or you could create a GPT, this is what's happening. This is how prompting happens, whether you see it or not. And we're just very big on foundational skill building. If you do this once or twice and then you go off to another tool, you're going to better at using that then. But also, Mani, there was a great question about how you would share this and cite this to your students that they were AI generated.

Amanda Bickerstaff
I think it might be worth actually showing also the disclaimer that we have on the guide too, if you want to talk about, like, the way you would cite it. And then we can kind of show you an example.

Mandy DePriest
Yes, let me see. I'm lost in My tabs, as per usual. All right, so we have at the very end of our guide an AI use statement. There it is. We did develop it with the support of Claude, is my favorite chatbot. We, and we use ChatGPT, obviously for the interactions. We wrote most of it ourselves, but we would sometimes go to Claude and say we're having trouble with how do we word this particular concept. And it would help us refine ideas a little bit. And actually 90% of the AI contribution was the worked examples that I was acknowledging. But, yeah, Claude did help us refine a few ideas. All of it was vetted through our professional judgment. And gosh, we did multiple passes over this to make sure that everything aligned with best practice and what we really wanted to say.

Mandy DePriest
So this kind of statement, and I may have put it in the text set document, I can't remember, is the kind of thing that you could disclose to students at the end just to model being transparent about AI use.

Amanda Bickerstaff
Yeah. And I think that if we did it, Mandy, we'll definitely update. You're the everyone, you're our 435 closest friends. So you're the first people to ever see this document externally. So we will probably make a couple little tweaks based on this. We did move pretty quickly to do this, but I think that there's some really good questions, though, Mandy, too, about citations. If you're using the paid version of ChatGPT or even using Google Gemini, you can ask for any citations potentially that could be supportive if you wanted to have that be a part as well. I think that what we see is that, you know, it is going to be something where you kind of lead into the, like, lead into what you want to be a perfect text set. I mean, that's the kind of thing that's a value here.

Amanda Bickerstaff
Like, if you want it to have just a text set, you want to have references. If you want it to have, like those links out that can lead to greater learning for those kids that have that, like, that's your. That that is your expertise, your pedagogical framework that can be very successful. So I think that's where these are meant to be, the starting place. These are meant to show you what can be. But ultimately, you are the expert, right? You're the expert about your students. You are going to be building your capacity, your mastery. And that's really what we want you to see, that this is a choose your own adventure. We're giving you the basics, we're giving you the foundations, but then you use those best practices of evidence bases in your expertise to make something that really works for your students.

Amanda Bickerstaff
So we want to go back to the presentation, Manny, and then we could talk just a little bit. We have some great questions in the chat and so. But it would be great to kind of just bring us back to like, what this can mean for you all and then we'll do like a last word before we go into the questions. But I want to reestablish that this is a work in progress. This will be as good as you all help us make it. Like I said, you're our closest friends here. We're all friends. Like, the idea is that this guide is actually developed to be tried out and you might think that some things are crazy or you love one or two of them, or you have a better idea, we want to hear it.

Amanda Bickerstaff
So there's actually a forum at the very end where we are asking for your feedback. We would love you to, if you're up for it, if you're a literacy educator or if you're, you know, K5, wherever you are, you want to model, we want to see what that means. And then if you provide that feedback, we will be actively looking at that feedback to improve this document because one of the reasons why it exists is that there's just this dearth of actual evidence about what's really happening and the opportunities, but also the potential risks in our literacy classroom. So what I want to do is we have a bunch of great questions, but I kind of want to go around the horn a bit about kind of a lot like a wrap up thought.

Amanda Bickerstaff
But like by doing this process, Carrie, Jasmine, and Mandy, like, what is the thing you learned and what would you want the people in this webinar to kind of take with them to when you think about actually applying this guide, maybe. Carrie, start.

Speaker 2
Yeah, I'm like, I'm learning in real time. I'm trying to, you know, see. I'm so excited about the chance to kind of see folks experiences which are already coming through in the chat. One of the things I learned, I feel like there's so many options right now. It can be a little bit overwhelming. And I think even in thinking about what options to develop, so there's some benefit in really thinking about what is going to work best for me right now with the technology that exists as it is that I can kind of learn from. I was looking, there was a question in the Q and A around thinking about explicit instruction versus deliberate practice versus precise Feedback and kind of where AI can really be supportive.

Speaker 2
And I think having a learner mentality about, I'm going to try it for one specific purpose first and see and make sure that it's a tool that's actually furthering the goals I already have instructionally. I think a mistake right now is to act like the AI is the thing, the instruction is the thing, the literacy learning is the thing. And some tools help us learn more and learn better and go further, faster. And some are distractions. We kind of already know that about so many things that teachers are always juggling. So really just being able to focus on that kind of connection to your instructional goals already and find new ways, it's. For me, I got to learn from Mandy and the way she worked these examples. And I'm already thinking, ooh, I want to try that myself in lots of different ways.

Speaker 2
How about you, Jasmine?

Amanda Bickerstaff
Why don't we also come off share so everyone can see everybody as we start to wrap up? Great.

Speaker 3
So I've been thinking about this work from more of a product lens, right. As that's the role that I play and really seeing the critical role that educators play in shaping better products that then can support teaching and learning.

Amanda Bickerstaff
Right.

Speaker 3
Like these, any kind of like edtech, but particularly with the introduction of AI products and tools, they have so much potential. And if they're absent of that instructional expertise, that educator background, the use cases, and that knowledge being infused into how the tools are continuing to evolve and develop, then we're, you know, it's a missed opportunity. So I'm like product developers, you need to listen to the educators and how can we like really center the voices of educators and practitioners and that knowledge and expertise of high quality instruction and learning to inform the continued development of these tools?

Amanda Bickerstaff
Hear, hear. Absolutely. We want educators to be the center of this conversation. And then Mandy.

Mandy DePriest
Yeah, I mean, I'll echo what Jasmine said about the importance of educator voice. And that's one of the things that I like the most about using a foundational model like ChatGPT or Claude is because it puts me in the most direct contact with the tool. I know there are so many tools that will do this for you and you can just do a click and click a button and get a thing, and some of them are great, you know, but also I would always question, like, is this tool cutting me out? Is there anything where I'm pressing a button and I push it right out to students? Does it have an opportunity for me to incorporate my Voice and expertise. And so that's the thing I like to emphasize is, like, you're the important piece.

Mandy DePriest
And there's so much concern about, will AI replace teachers? Let's not make it easier for them, right? Let's not, like, just be the button pusher. Let's be able to demonstrate, well, look, here's how I interact with this chat bot and apply my expertise to make sure I have exactly what this student needs. The passage about slimy unicorns, because AI is not going to get there on its own. So, yeah, the importance of centering practitioner voice and student needs is a crucial takeaway for me.

Amanda Bickerstaff
And get ready for the new AI for Education mascot, the Slimy Unicorn, everybody. But the thing that I think I learned was actually, I'm going to take a step back. Like, we've been doing this work for the last year and a half, and it was a really big deal for us to kind of dive into something so pedagogical, because the technology, you know, the hardest thing that we find is not being able to go into a school. And even the tools that you set, like the ones that were shared, there's no perfect tool yet. There's no pedagogical tool. And that's to Mandy's point, and that's even to Jasmine and Carrie's point. Like, that level of expertise is so important.

Amanda Bickerstaff
But we do believe that if we start to create some structure around evidence, around applications, and allow for you all to do the work and have you go and actually test this out, that what we can do is we can start to build better things for our students. And while it's not going perfect and it's definitely we're going to learn so much in this way, we're really excited about this because with us, like, being able to partner. And I just want to say thank you to everyone that helped as well. Like, it was a huge troop effort. We did this in. How much time did it take us to do this? A couple months, everybody. Like, like two months, maybe, max. Including the design. That's very rare. So we're very lucky to be aligned so closely with SAP.

Amanda Bickerstaff
And we're hoping that we'll also have a math tool guide as well in the spring. So maybe this is a series that we'll have. For those of us that are STEM teachers like myself, I've got my literacy people here, and of course, I am the former science teacher, but just really excited about this work. And I just want to say, if you have to run before the end, we're going to stick on for some questions. But we want you to take this as a galvanizing action of we can go and try as long as we're centering evidence, we're centering our expertise and we're centering our students actual like experience that we're making this better for them. And I think that's what we ultimately want to do with our two organizations.

Mandy DePriest
So.

Amanda Bickerstaff
Okay, let's leave some questions we're going to start with. Actually, there was a question for Carrie and Jasmine from Lauren and she was asking about how could you like, kind of come through, like, think of science of reading is taking up so much space.

Mandy DePriest
Right?

Amanda Bickerstaff
So much space. So how would you connect our guide with the idea of the science of reading and like how the two overlap or can support each other?

Speaker 2
Yeah, yeah. I mean, I think that comes a little bit into effect when we're talking about actually those content area buckets. Science of Reading Sor. Like is like a literacy buzz term that we're hearing all over the place. What it means is research based best practices that are rooted in how to teach students how to read. And so when I said before, like, how can AI enhance that purpose? That's what we're thinking about. So there are lots of tools that can help you do that better. And there are lots of practices that aren't aligned to science of reading that you. There may be tools for those too. We want to kind of weed through that clutter.

Speaker 2
One of the things that I think a lot about is when you are using those AI tools that aren't kind of you at the steering wheel, like Mandy talked about being able to ask like, what is the research? What is the pedagogy of this tool? Is there someone who can explain that to you? Can go a long way. And then it should really connect with what you're learning about evidence based best practices in the classroom.

Amanda Bickerstaff
Absolutely. And I think that this is an opportunity, like this is not going to be the most fulsome document because someone also is saying, like, the idea for more writing, we did kind of start from a space of where we felt more confident about where the technology is today, where, you know, there are a lot more questions about student use of AI, especially in ways that are not deeply structured and transparent. And so. But it is a good idea of like when you start thinking about how these can come into play. So thank you for that. And then I'm gonna go to Mandy. So Brian asked a really kind of like really interesting question. We don't have answer for this, by the way, but it is like, interesting.

Amanda Bickerstaff
Like, like, how are we, you know, the idea of like thinking through the steps of good teaching, right? Like explicit instruction, deliberate practice, feedback and retrieval. Like, what are we thinking? Like, we are starting to do that work right internally of like, how we start to think about this, especially around assessment instruction.

Mandy DePriest
What are you.

Amanda Bickerstaff
Do you think there's anything that we can pull out from this guide that's going to be a little bit easier for like, anyone to use around those practices?

Mandy DePriest
I don't know if it's like explicitly named in the guide, but it's kind of implied in there. Is that treating a chatbot as like a thought partner like you would someone in a plc, we say all the time that they're designed to be creative. And so imagine you had a PLC partner who was well versed in the science of learning that you could bounce ideas off of and they could give you endless suggestions and possibilities and they wouldn't all be good, but that would be where you as the expert would come in and apply your judgment and make sure that they were aligned to what you have going on in your classroom with Gen AI and otherwise.

Mandy DePriest
So, yeah, I think using it as kind of a thought partner and a subject matter expert to enhance and work with yours is probably the best way to align that at this time.

Amanda Bickerstaff
And so no, and I agree completely. And I think that what we hope to see is that we can, if this goes well, everyone, if we can continue, we plan on continuing this and we do this already through our professional development. And so I'm going to do one more question for each for Jasmine and Carrie and then we're going to kind of one more question for Mandy and then we're going to wrap. There was a really good question around that kind of the high quality instructional material piece. Like, what do you think about how that like you can take existing high quality instructional material and then like go and use that as a base? Like when you were, when you. Now you've learned so much about generative AI, like, what are you thinking about that kind of hq? I am support.

Speaker 2
Yeah, I mean, we're big proponents of hq. I am in sort of two ways really increasing kind of the quality on the publisher side and increasing the use in classrooms. Because again, high quality, strong instructional materials are a lever for equitable instruction. At the same time, I feel like a lot of these, you know, the example that Mandy illustrated for us was really about like the physical text that you might add in. But a lot of these examples are about kind of ways of interacting with the content. And I think that, you know, the sky is the limit in terms of how you might be using some of those interactive tools to support implementation of the materials.

Speaker 2
I would say the other thing is that we've got imperfect instructional materials all over the place and being able to make just slight adaptations that help with quality without kind of throwing. Throwing everything out the window is a skill that all teachers understand they need to have. And there's lots of, I think, uses here or uses folks are going to be exploring that can help with that adaptation piece.

Amanda Bickerstaff
Absolutely. And I think that, I mean, we have to get real with each other too, a bit. Right. When we say hq, I am, it's kind of a buzzword, but, like, how much of us do have HQ I am at our fingertips. I mean, it is. It is rare to see that for an entire curriculum, especially one that can meet the needs of our actual students at the levels of which and their levels of slimy unicorn interest. So, last question for Mandy. I think this is really, like, it gets really interesting in terms of, like, how we do what we do. And so we love Claude and we. But we use ChatGPT as well. Do you think that, you know, if someone's really getting involved here, that it's worth moving to, like, the paid chatbot version, Mandy, of like ChatGPT or Claude Unlimited?

Mandy DePriest
Well, as you know, I prefer Claude. If you're on the free version of Claude, you're going to run out of interactions fairly quickly, especially for those sustained conversations, like I showed in the worked example here. So, I mean, I would say that the free version of CLAUDE is still a strong enough writer, but I think you're going to run into capacity limits, so it's worth having a conversation with your district. I know purchasing at the enterprise level can be prohibitively expensive, but if there's an opportunity for individual reimbursement, like if you're a teacher that knows you're going to use it quite a bit, it's worth asking. That said, I, I think if you run out of tokens and Claude you or interactions and Claude, you can jump over to ChatGPT or something similar and kind of make it work.

Mandy DePriest
If you're not able to get access to the paid. But again, the main important thing is your expertise anyway. So any chatbot can get you like maybe 80% of the way there, and then you can come in and whether you're doing the last 10% or the last 20%, that's what I would focus on the most is making sure that you have what you need to inject your expertise to really take it to the next level.

Amanda Bickerstaff
And this is why, I mean if you use the free tools out there too that you love, then you know that are more applications then we would suggest that. But there is a reason why things are paid for, Claude. It's just access. You always get access to the highest model but you do get limited how much you can use it. But ChatGPT, the 4.0 model is actually going to better. Not only is a model going to have higher quality outputs, but it also has modalities like research, like can go to the web, it can do code analysis, has canvas where you can actually like edit within the documents. So it's something to consider. I just want to answer one question from Dan that came in a little bit late.

Amanda Bickerstaff
This idea of like chatbots as thought partners or content experts in certain things we say thought partner, less content expert. But because the technology is so new, you cannot just fully trust it. So even if it's a seventh grade reading level on a science content, it could make a mistake. It could use an outdated, let's say that something new happened outside of the knowledge cutoff. It might not be there. It could have a mistake or a conflation or hallucination. You can't take the eye off the ball. The same way that you would read the like and evaluate even though you're producing it doesn't mean you could take away the reading and evaluating component of it even if you liked it where it came from. Because like you can see in Mandy's worked example, she asked for a visual and it was like no.

Amanda Bickerstaff
And so you need to keep that like level of like. Like you can't get so comfortable with any of this. And it doesn't mean that it won't get better and better. But ChatGPT will be 2 and 9 days. Think about your little toddlers, if you have one at home. Think about the Internet when it first started. These technologies are not great yet. So your expertise is going to be really necessary. Okay everyone, thank you guys so much. We are absolutely so thrilled to have you. This is our. We had so many amazing people and some friends, new friends, old friends. But I want to thank you to everyone that helped us.

Amanda Bickerstaff
So Carrie and Jasmine being the best partners on this work that we could possibly ask for, the work that was done behind the scene by both of our organizations and the work of Mandy and Corey on our team, really leading the work from the Generative AI section. And I just hope that all of you enjoyed this. Share it, try it. Give us feedback. And what we will do is we'll keep working together to responsibly adopt Generative AI. So thanks everybody. I hope you have a wonderful morning, evening or bedtime. Bye, everyone.

Speaker 2
Thank you so much.

Mandy DePriest
Bye.