Using AI to Make Math More Accessible

Welcome to the future of Education. And now here's your host, Michael Horne.

Hey, Michael, here. What you're about to hear is a webinar that I hosted for a company, M7E, that full disclosure, I'm an advisor to. It's two of my former students that founded it. And it's a very cool AI tool that does something different from a lot of the tools out there on the market. It's not student facing, it's not teacher facing. What it does is it works with curriculum providers to take their math content specifically and use the AI with a set of clear rules to reduce the language complexity so that the curriculum is actually teaching and assessing on the math skills rather than some of the language things that might run interference for multilingual learners. In particular, I hope you enjoy the webinar that we recorded, find it interesting, informative, and that it sparked some questions for you about how else might we use AI that sort of steps out of the typical notion of just, hey, it's a chatbot, and where are the applications that might take off that could make an impact in education? Let me introduce the two folks first who have been digging into this problem from both the research and product perspective. First of all, we have Abdi, I'm looking for you on my screen. There you are, Abdi Ghalid. He's a Harvard Education Entrepreneurship fellow and the co founder of M7E AI. And we also have Kadar Sridhar, also a Harvard Education Entrepreneurship fellow and also the co founder of M7E AI both, as I said, former students in my class. And together they've built this company and product that really evaluates these existing math problems and tasks for linguistic clarity and accessibility. They flag hidden barriers that can trip up students and then they suggest, I think, importantly, revisions to keep the mathematical rigor intact, but while making the language and design more equitable. So I'm excited to bring them in. And Abdi Kadar, welcome. I want to get into it. The way we'll do this is I have a couple questions for you guys up front and then I'm going to sort of give you the stage, if you will, to maybe show what you guys have developed and how you've been using it with some curriculum companies. But I think your own personal stories to this, I got to watch it a little bit up close. But for those that don't know, you tell us your own personal stories about how you came to build this tool. Why did you see it as a big problem worth solving? Because I know a lot of folks don't even tend to think about these challenges a lot of times. So Kadar, Abdi, whoever of you wants to take it first.

Thank you so much, Michael. And thank you everyone for joining us today. My name is Abdi. I grew up in Norway and like Michael said earlier as well. But I remember very clearly how Mad content was created and how that shaped my experience as a learner. And a lot of that stuck with me as I was growing up. And then I came to the US as a student athlete in track and field and built a career always around the core theme of using technology to make learning and organizational process more effective. I spent years working with data and AI, especially my master's program with Kedar, where most of our work was focused on how AI can streamline complex manual processes. So M7 education is a natural extension of that work. And curriculum creation is one of the most complex and time consuming processes in education. We'll talk a little bit more about that. And we think AI can dramatically improve both the speed and the quality. And because I personally the impact of curriculum design, this mission is really, really important to me as well as is to Kedar. So over to you, Kedar.

Awesome. Thanks for that, Abdi. And thank you, Michael. And thank you everyone for being here today. We're excited to chat about what we've been working on. So I'm Kedar Samit Abdi, also international student, grew up in Oman in the Middle east. And I sucked at math like it was. I just wasn't, I just wasn't doing well. And turns out a lot of the things and the questions didn't make sense from a lot of these big publishers, questions about lacrosse, which I had never known or played, questions about skiing. And I live in, you know, in a very hot country. And like all of these different contextual clues and things that were actually distracting me as well from what I was actually trying to do. Fast forward to, you know, ended up coming to the US and doing my undergrad at UCLA in computer science. And I ended up in the STEM field. And I think a big focus for me is how can I improve access to that field and to those oriented careers. My career is in product and tech, but specifically I worked as well at an undergrad admissions for ucla, focus on improving access to higher education, and then ended up at a nonprofit focused on STEM literacy, digital literacy, AI literacy, helping people bring into more technical careers as well. And so this passion, this through line for both of us has been how do we now provide a way for students, for people to Access content when the thing that is being tested for is the wrong thing being sort of expressed and shown. You know, people spend more time decoding language instead of actually testing their own math ability in this specific case. And so with Abdi, you know, our master's was in learning, design, innovation, and technology here at the School of Education, and. And we worked across the board, you know, across Harvard, across MIT and rest of the Cambridge schools as well, to continue diving into the space, doing research, and figuring out how do we best tackle this problem. That was very close to both of us.

So I think it's perfect. Lift off. And I love that. Well, you probably didn't enjoy it, that personal story of a kid struggling with math. But when I think a lot of people think about AI in education right now, I think a lot of people, like, the thing that comes to mind is chatbot, right. And I think a lot of people are fearful of, like, the student facing chat bot in particular. What I think is so interesting is that you all have built a tool, actually, that is like a couple layers before the student, right. To make sure that the curriculum getting to them. What I think is so interesting, though, right, Is before we get to solution and what. How AI you've been able to use it to help districts and so forth, let's focus on problem first. And where do you see districts, schools struggling most in their current evaluation processes, especially when they're comparing multiple math publishers or frankly, like the homegrown materials that we see or materials that teachers are taking from other teachers in all of this. Kadar, why don't you start off on this one?

What we've heard and what we've also sort of experienced while, you know, speaking to district leaders and just speaking to all the, you know, all the people in the system, right? Whether you're a teacher, district leader, or part of assessment teams, researchers, editorial teams, or, you know, publishers themselves. There is a. There is the. The core problem of I am a district leader or I'm an instructional specialist or a curriculum manager, and my district has its own needs, right? My district, I have a specific type of students. Maybe there's more bilingual students here, or maybe there's more students that, you know, with. With a lower average literacy rate or a lot of these other things. But every district has such a different profile, and yet content is sort of dispersed equally to everyone. And so for something that we've been trying to tackle as we've got into more of these conversations, is how can we help district leaders specifically and, you know, districts themselves look have visibility into all the publishers that come forth, you know, come to them, be able to see which story aligns with their populations, which publisher and material Best speaks to, you know, and Best has, you know, shares that voice with the. With the students themselves and the educators themselves. And in general. And, you know, and Abdi will expand on this as well. There is just right now a very manual process and there's limited bandwidth in general when looking at how this is best aligned not only with my district and our goals, but also coverage in terms of the broader, you know, the broader ecosystem as well.

What we've seen, just to add to that is we've been talking a lot with district leaders, a lot with Kirkham developers and editorial teams, with teachers, with principals across the board. And there's the idea when it comes to district leaders and schools mainly is there's a heavy reliance on that idea of these curriculum developers know, kind of Best. But then there's this almost sort of like a little bit disconnect when it comes to the. The diversity of students in the classrooms that is changing dramatically. And we've been looking at it for multiple aspects of how can we help build something that could help the district leaders and others to evaluate the content, to evaluate the curriculum and be part of the design process before it reaches their specific district. And Again, there are 50 million K12 students. America is very, very large. So different districts have different needs. And like Idar was alluding to, how can you make sure that my district and my students and the community that I'm supporting, how can we make sure that they have what they need in order to excel in their path as learners? So we'll cover more on that in few minutes.

No, no, no. Perfect. So I think that's a good framing the problem or challenge, if you will, that schools are facing. And with that, I'm going to give you guys the MET virtual stage. I guess these days it is in zoom, but like three sort of central questions that we'll have you answer. One is briefly framing what you've learned from the research and from working with publishers and districts. So one, briefly frame what you've learned from the research and from working with publishers and districts. Second, then let's go to show, but not show and tell. Show, not tell. I want you to show a concrete demo of M70E on real math problems, what the platform sees that humans might miss in sort of a quick review. And then I think, obviously give us the. This is a tool, a free tool at the moment that schools and districts can use during adoption cycles, RFPs, internal reviews and so forth. So show us how that actually can be used. So I'll kick it to you with that framing.

Awesome. Thank you so much. Yeah, so I'll walk through a little bit about the backstory. We touched upon that already, but give you a little bit background on where we got started, the problem that we're seeing. And then I'll hand over to Kedar who can show you how the platform works and how you can take part of it. So as I said earlier, there are. The first problem that we see here is the comprehension crisis. There are 50 million K12 students and 61% of them are below grade level in math. We're seeing that one in four students are bilinguals. The thing that we're seeing is that when you're looking at math comprehension, it's not just the bilingual students, but also especially in the United States, where you have zonings and you can have a specific curriculum developer provide the curriculum and content to a community. And in that community could be. Half of it could go to a public school that has less resources. And then the other half could be an affluent community that have more resources, but they're both getting the same curriculum. What happens is the affluent one will get probably better scores than the other one, even though they are could be bilinguals, but you also have a lot of native speakers that live in that community, which then adds to a lot more, many more, millions of the students that struggle with math because of linguistic barriers, because of the exposure they're getting to their personal experiences. Like Kedar mentioned earlier, lacrosse, if that's one thing, or badminton or whatever it is. If you don't have been exposed or don't have experience in that, it becomes sort of barrier to you to solve the math war problems. When we did a lot of research, we found out that this also has been done before us. But we highlighted and we improved upon it, which is the problem is not so much about the students mathematical ability, but more so on the content and the way it's presented to them. Then when we had conversations with teachers across the United States, we also realized that teachers don't have a lot of time to scaffold every single student. We're talking 20 or more students in the classroom. So what they do to try to make things work is simplify the word, the problems in a simple way to the students, use Google Translator or other methods to help the kids to get through the problem. Again, there's not enough time for that level of scaffolding but then this creates more teacher burden, extra work and inconsistent instruction across the classroom. And then when we had conversation with the publishers, we see they spend a lot of time with the editorial team to build the curriculum for that specific state and that specific district. But at the same time, the editorial team is also. There aren't enough people in the editorial team who have the experience from these different classrooms, in this diversity of classroom that is growing in the United States. So there's a lack of scalable workflows to evaluate and revise that content for every single student. So we've seen that as a problem in conversation with the publishers and then the districts. One of the most important piece there is how can you as a district leader or person, part of the district evaluation team or procurement team, how can you evaluate multiple different publishers when they're coming with you with curriculum materials in order to make sure that this content actually fits with your student cohort? And of course you all have your own rubrics. But how can we make, how can we elevate that and try to figure out ways to stress test it before it even hits adoption or any decision making? Our journey started from a product practicum here at the Harvard Ed School. A publisher came with a problem. They have 15 million plus students and they looked at, we had this growing 25% bilinguals and a lot more many students that are in the classroom. How can we make sure that the language is not a barrier and are there things that we can do to build and develop so the workflows of every area becomes simplified and more comprehensible? Like I said, we spoke with educators, researchers, editorial team, district leaders to to confirm that it's a systemic issue. And then we went through 300 plus research papers as well as publishing our own research that came out a few months ago. And essentially the mission for us as we touched upon multiple times is to make sure that we can remove the unintentional linguistic barriers without simplifying the mathematical rigorous. So when we're saying making it easier or more comprehensible, we're still making sure that the mathematical rigor is there, but we can use a different language that is maybe more universal and not cultural. So research highlighted that precise vocabulary plus clear syntax equals better comprehension. So our solution, which Kedar will show you more about, is to evaluate and to revise curriculum language before even reaches the classroom. These are some of the feedback that we've been hearing from the teachers across the U.S. again, every student struggling with math war problems, even native speakers, some of the teachers because of the not enough capacity they ask other students to help their friends to see if there are ways to help everyone in the class when they're not when there's not enough time. Curriculum developers don't have that insight. That's what I touched upon earlier in terms of the editorial team could consist of few people, but not everyone has that experience from every single classroom. So what you're seeing with M7e and we'll explain what the seven criteria in the M7 is is to strengthen the math content before it reaches the classroom. So it's an AI powered curriculum intelligence platform to evaluate and revise based on the seven factor framework that we built through this research. And then making sure like I said earlier, we keep the mathematical rigor and without altering the math, maintaining standard alignment, original cognitive demand and rigor, helping the editorial team from the publisher aspect of it to streamline their workflows but also offering a clear and reliable signal of student content comprehensibility to support adoption and HK I'm aligned procurement decision for the districts and review teams. And then of course along the way they'll also help reduce the teacher burn and then create something that can scale across the systems upstream. And we'll see. We'll show you how it works pretty easy with publisher schools, district and others. And I'll give you over to you Takedar.

Awesome. Thanks for that. For that Abdi, for that framing. To kick off how we're sort of thinking about this from a system level. So the actual seven factors that come in and you can see them on the screen here are what we've been able to synthesize through all of our research and our interviews and looking across the fields of translanguaging to linguistics to mathematical language routines to best pedagogy and learning sciences. And a lot of it as part of what we've been doing in classes and research labs and sort of associations involved with the Harvard Education School and beyond as well, we figure out that sort of synthesize these seven criteria in order to create comprehensible math language. And it's all about like Abdi touched upon, removing the unintentional barriers or the hidden barriers, specifically things that were not intentional pedagogical choices. Right. If something is meant to be there, that's great. But if something sort of slips through and then creates this hidden barrier that can impede student comprehension and learning, that is what these criteria are working on sort of preventing to happen in the first place. These criteria are then further broken down into 60 to 80 different sub criteria as well, all aligned with sort of the latest research and our own findings and interviews in the field as well. To dive a bit deeper, we've actually built our own linguistic comprehension model. So this is where, you know, where we see the value of AI. How can we build something that is scalable but is fully based on the research and how we're able to fine tune this with not just our criteria but also the Common Core standard guidelines, different state guidelines specifically as we're trying to make sure that we have that alignment there. Being able to get live classroom feedback. So from our testing on the ground, along with the research that we're working with, taking in teacher observations, taking in student comprehension, and then being able to sort of iterate upon those outputs as well, getting editorial teams to review our outputs as well, and constantly evaluating if that what we're producing has those benefits and has those sort of increases in comprehension that where you set out to achieve. As we sort of train this, the goal is can this LCM be something that evaluates math content for comprehension and those barriers and identifies misalignments in general with what the actual guidelines and whether it's grade guidelines, state guidelines and those factors as well there. And then being able to produce that revised content that is optimized for this comprehension, but specifically focus on reducing cognitive load when it comes to things that don't concern the math part of it or the math comprehension part of it. So I'll stop sharing the slides here and actually just dive into what we built and the actual platform itself, which is live, which is something that we have educators and districts already employing here. So this is M7E AI. This is the curriculum evaluation platform that we've built here. And the goal is again as we've said before, how can we now provide an easy way for districts to upload sample units, word problems assessments, any sort of curriculum content, sample items that they're getting from publishers as publishers are making their claims or sharing content as well, and then being able to get a district summary and sort of the high level features, the strengths and the weaknesses of whatever has been uploaded, getting deep evaluation based on the criteria and then even getting revisions that can help support student learning at a broad level as well. So we've built this for superintendents, curriculum directors, review committees, you know, sort of school boards as well, anyone that is making those adoption, procurement, an evaluation oriented decision specifically. So just as an example, we've uploaded an open source math curriculum. It's great for fractions and something that you're able to do, whether it's uploading a bulk set of files or different sorts of components that I highlighted earlier, you're able to pick the specific grade level From K to 12 and then even the state standard. So whether it's California, Common Core, Florida brought up best, Texas, teks, and we're working on getting more and more specific state guidelines in there as well. But most of the country uses the Common Core. And what we found important is actually the learning goals and standards not just for the curriculum, but what are my goals as a district or what are my goals that I'm trying to sort of enact across this curriculum as well. And this is helpful to share more context about the profile of the district or things that we're looking to achieve in general. So for this great form fractions content, we have a couple of learning goals as well as a couple of standards that, that align to sort of the Common Core there. And what we're able to get is a summary report. And I won't go fully, you know, I won't read this out to you, but at a high level it's giving you a snapshot of what is actually makes sense in the content you've uploaded. What is that linguistic structure look like? What are the main issues and key strengths as well? What I will say, and I can scroll down just for some context here, is when we revise it, we take an original problem and then revise it based on all of the factors that you saw in that linguistic comprehension model. So in this specific case, here's an example of two kids that are planning on going for a run, right? And immediately you're able to see soccer practice, after school activities, these sort of different names as well. And what our goal is is with testing with students that we were, you know, working this on with, they got stuck at, oh, soccer practice, I don't do that. And also after school activities, my school doesn't offer that. Right. So being able to even again universalize what we're doing across the field here and then making it very simple in terms of they just want to measure these distances, we're still keeping the students and we're able to get to the core of what the math problem is asking specifically. And now there is no, you know, it's still a three part math problem. Comprehension is not changing, but we've already removed a lot of those contextual barriers that actually prevent them from comparing a lot of those fractions there. So if we jump back to the district report, we're now Able to see, you know, there are strengths in terms of cohesion and labeling, but really gaps in terms of those contextual references, different formats as well. And you know, any exclusionary measures that might be taking place as well. We follow up with implications for district adoption and review in terms of how this matters for the guide, how this sort of aligns with the Common Core guidelines for that specific state and what might be helpful in terms of revisions, as if this is something that you're choosing to adopt as a district. And then finally recommendations that districts can sort of follow in terms of how to revise what parts of it to adopt. Maybe there's parts of it to adopt and parts of it to revise. A lot of factors like that as well. And then that's sort of the high level overview. And then it's sort of a deeper evaluation based on the criteria as well that, you know, the seven criteria that we've. That we identified earlier in terms of why or why not it passes each criteria and sort of general notes in terms of the, the main issues that were found on our. So this is more the instructional design, you know, curriculum developer oriented level instructional issues specifically and then recommendations for further improving it as well. So we wanted to give all the context as district leaders, you know, and as curriculum managers in general. They want to be able. We are trying to create a transparent model that is clearly showing why we made revisions or why we made changes or how we're doing a lot of this evaluation specifically. And that's why we're able to explain all of this. And then these reports can be exported. So at a high level, this is what we've been able to accomplish so far with M70AI. And we just want to be able to help districts make better decisions and give them, empower them with the tools that can help them find success for their students.

Yeah. And Michael, before you jump in, you asked a question. How can school industry leaders use this kind of tool? So essentially, I think I said in the email as well, we're offering for free to district leaders and schools to use the platform. The way you would do it essentially is just contact us and then we'll help you set up and get started and then, and then work with you from there.

Perfect. So they can just reach out directly over email from. From that response. Perfect. Okay, let's go to Q and A Elmira. I think that they may have answered the role of AI in evaluating this, but if so I'm going to go to the second question. But if I have that wrong I'm going to bring you back in, but we have a question around. So the tools aren't evaluating curricula in terms of pedagogy, like linguistic scaffolds, math language routines, teacher moves, and so forth. Is that correct? And so just I think tighter definition on what, where it is reviewing and where it isn't.

Okay, I can tackle that. Perfect. So we actually. And again, this is just an example here, but in terms of the recommendations, it's not just the language recommendation, but it's also pedagogical recommendations and then also the formatting of content. So we're providing feedback on even the images used or the diagrams or anything in that sort of context There. A lot of times our revisions actually end up providing more scaffolding because again, if something's an intentional pedagogical choice, we don't want to be eliminating it. We want to be further explaining it at a high level as well. So whether it's providing visual supports, clarifying instructions, providing more scaffolding, those are the things that we're also looking for in addition to the linguistics side of things and also checking for math correctness and the math principles there as well. And sort of what, you know, we want to be very clear about what each revision is supporting, whether it matches the learning goals that were put in and the standards, but being also focused on the linguistic side of things, comprehension wise. Hopefully that answers that question. But we are definitely working to provide more visibility as well into the pedagogical side of things.

Perfect. Let me ask this next question that is in the Q and A. And then, Christian, I'll get to you after that, which is this question has to do with AI mediated personalization. As tools like this become more common, do we risk siloing students learning experiences? So take the point about content delivery being responsive to context. Right. Like Kadar, your observation around, you know, skiing while living in the Middle East. But if most students mediate their learning in this way. You mentioned scaling this technology. Do we risk opportunities for negotiated understanding? Do we lose something important if 30 students in class engage with a given math problem by interpreting 30 different personalized descriptions of the problem? Would it not be beneficial to everyone for students in Oman to learn something about skiing while North American students gain exposure to Omani? You know, racing or falconry or something like that? So I will let you both answer that question or whichever one wants to take it.

And just to be very, very. Just to clarify what, what we're doing here, we aren't personalizing for every student Right. The goal is how do we make content universal across various groups of students? Because this all started with the static textbooks that we were, you know, the publisher came in and they, you know, these textbooks are published and they, you know, they don't have review cycles for, you know, maybe it's four or five years, different cases for different publishers. But how do you turn content that is static, printed, even supplemental digital content to, to some extent as well into something that could be understood by various groups of students? Because like Abdi mentioned, an affluent, you know, an affluent area will be getting the same sort of curriculum as a less affluent area. So how do we sort of equal the playing field on that aspect? Abdi?

Yeah, I think there's so many products that now are hitting the classroom for teachers, for students. When we looked at the ecosystem, we're thinking more of if the EcoSystem consists of four players, you have the students, you have the teachers, you have the curriculum developers and district leaders. Where can we make a system level change? We didn't get out there to say we'll create a personalized product for students or for the teachers. It was more so when we started with this at the Harvard School, the practicum the publisher was doing, what Kedar just said is textbooks, depending on what publisher it is, they review them every two years, every four years. And those are the core programs that students learn. So we're looking at how can we make the content, these textbooks, more universal so every student can bring their own, leverage, their translanguage in that touch upon, in order to understand the content in front of them and not let that be a barrier. And if we, if we, we spend a lot of time on figuring out what are these seven criteria. It was actually five in the beginning and then it became six and then became seven and then we changed what the seven and the six was based on the feedback from the researchers across UC Berkeley, UCLA, MIT, Stanford, etc. And we, we came to understanding that this is something that these seven criteria can be helping across the board, every student to understand any content in front of them. But yeah, not nothing very personalized to every student. It's just across the board. Every student should be able to understand any type of math if they leverage the M7 framework.

One other quick thought and then I want to go in. So Christian, I'm going to come back to you in a. I want to go to one other question in the Q and A because it's, it streams perfectly. And then I haven't forgotten about you. Just one other observation which Is this is always relative to the content or standards objective that is trying to be taught. Right. So skiing for Omani students might be something worth teaching, but it might not be part of the math purpose. Right. It might show up in social studies or something else. And so it's not to take that out, but it's to make sure that relative to the objective that is trying to be taught, that we are making sure that we are in fact teaching that. And then when there's intentional interference or complexity, you can say, yes, we're keeping that because there's an instructional purpose here, as I understand it. I think this goes into the next thing, because this is actually already done a lot with people who write standardized assessments. They have a whole teams that are trying to make sure questions are not biased or unintentionally asking something, you know, testing for a different objective than the one that they hope to. And so it goes to Rebecca's question, can this resource be used to review assessments? So, Kadar Abdi.

Yeah, yeah.

Short answer is yes.

Yeah, it can be used like for an iceberg. It's like anything from overall curriculum to teacher notes to any instructional external scaffolds, worksheets, problem sets, anything that is content based, that is student facing, that is in those different categories as well. And we take a lot of account into specifically looking at an assessment versus an instructional material, because an assessment is more in the silo of the student, whereas an instructional material is in the context of a teacher and being able to understand what teacher's role in the classroom is as well. And so these are all things that we're continuously building into the platform as well.

Yeah, and we'll share, we'll share with the group here. We tested grade four. Grade four. 17 units. 17. 17 lessons plus four assessments. And we'll share that after the call. It's 200 pages long, but quickly you can review and see how the original was and how we revised it and the context around that. But yeah, assessment can be also used. Absolutely perfect.

Let's do lightning round. We've got three questions that I think all connect and then we have two questions that are somewhat different. So the three questions that connect are who's your target audience? Right. Sort of. There's publishers, there's schools and districts. How do you think about them? The related one is has or would M70AI ever partner with another established organization with a similar mission? Are y' all a for profit company? How would that work? So still. And then Julie asks, how has this been received by the Publishers. So maybe you can sort of walk through each of those relationships because I think you have precedent on all of them.

Yeah, I can jump in. Killer. So it's been received very well by the publishers. It's been more like almost very surprising that they've been very. They essentially love the way we approach the problem because there's stuff that they're already working on. And if you go back to the whole story, it started with a publisher coming to us and sharing the issue that they are facing, the challenge that they're facing. And now we've been across multiple publishers in this space and it's been just received really well. And we continue continuing to work with each and one of them.

Let me give my answer of what I think it is for you guys and then you can correct me since it's your all company. Yes, for profit, but yes. You have partnered with mission aligned nonprofits in particular as a tool to help them evaluate curriculum. You can give some specifics in a second, but I think it's a commercial relationship with the publishers and right now it's giving this tool away to schools and districts to sort of uplevel the field and help them ask the question of publishers more, more, more to bring it into the criteria. Is that. How did I do guys?

Yeah, perfect. So yeah, essentially going back to the system level of change, I don't think that would just us working with the publishers. I think it's like a whole community drive. So working with schools, doing the RCTs and then also working with district leaders, seeing the value of it. Work with nonprofit, working with other organizations. You asked the question about other organization that are similar. We were absolutely open to working with them as well. It's more of like anyone that's helping with making the change and improving the comprehension level for students. We're were for that publishers. Yes, there is a, there is a commercial piece there, but for schools and districts there's. It's a free product.

Okay, let me ask this question. I'm going to switch gears. So there's a question. Could a homeschool parent or a small micro school use this tool doing math with their child or would it be too time intensive?

Definitely. I mean this is the, the goal is how do we now reduce the time that it takes to figure out if content is good and if it, if it isn't, you know, best serving my child or best serving the students, how can we now revise it to make sure that it has a lot of those, those factors in there given my child's you know, age and developmental sort of Persona and you know, the, the main, the key learning goals I want to get there. So whether you're a micro school or a parent, there's definitely scope for, you know, uploading, you know, content that you would be teaching anyways or being able to revise it to best meet your child's needs in that sort of aspect as well.

Okay, perfect. So then Peter has an interesting question. Can the technology go the other way? For example, could content become very specific for an art school or a ski academy or a tennis academy where young Olympians or or similar athletes are being taught? Which is perhaps not something you guys have thought about, but I don't know, maybe you have.

Yeah, it's, I mean, yeah, it's not something we've necessarily thought about, but it's definitely very doable. Right. I think it's very making, you know, it's. The goal is how can we best transform content to best fit the needs of the audience. Right. Or the needs of the learners themselves and cater to them. So I think in that respect, if the goal is, oh wow, we actually have a focus now, let's now shift language to best serve that focus or that content area, that domain sort of area. Yeah, very, very doable.

I'll just add to that. I think it's very doable. Peter. I think if you. In what Kedar was showing earlier, we have this specific area that can essentially help the output, the evaluation and the revision. So in the learning space, if you can focus on and say I am very specifically focused on art school or ski Academy or etc. And these are the learning goals that we have for them, then we'll take into account that with the standards and then and give you the evaluation in a series. So we haven't done it, but it's absolutely doable.

Okay, let me ask the last question and then we're going to let you guys have the final word as we wrap which is, is M70 intended to be used by students with dyslexia, ells, other learners with specifically defined needs, or do you see applicability for all students?

So currently the way we've been thinking about M70 is to be used by district leaders to evaluate the, the con. Their material that they're thinking of in, in their district be used by curriculum developers as they're thinking of creating new material. But also in the study round that we've been doing and still doing is teachers can use it, but it's not a student facing product and it's also not as teacher facing product, even though teachers are using it to give us feedback. If you guys remember, very early on we were showing you the linguistic comprehension model that we're building and on one side of his classroom, feedback that feeds into the model to give us feedback on how to improve it, how to continue develop something that is really relevant to the classroom, but it's not intended to be used by students or teachers. It's more for the ones creating the product.

And what I mean, what I'll just add there, in terms of the beneficiary, it is all students, right? The content is the content that we revise, evaluate is best, is intended to serve all students, regardless of their linguistic background, regardless they're coming from. Because if we can improve it for some students in terms of these buyers, it actually helps. And you see improvements across the board for all students. So that's the goal.

Perfect. All right, so we'll wrap there. Huge thanks for a series of really interesting, thoughtful questions. Huge thanks to Abdi and Kadar both for sharing the original research that went into this. And then the practical side of this work as we explored how to make sure that math materials aren't just aligned to standards where there's a lot of good work already, but also that students can be able to understand them by getting the linguistic complexity correct and getting away from these hidden linguistic barriers, especially for multilingual learners. But as Kadar and Abdi just said, for all.