---
id: "workshop:reading-challenges-revealed-ai-innovation-in-dyslexia-assessment"
type: "InnovateUS workshop"
title: "Reading Challenges Revealed: AI Innovation in Dyslexia Assessments"
url: "https://innovate-us.org/reading-challenges-revealed-ai-innovation-in-dyslexia-assessment"
date: "2025-03-25"
duration: "60 minutes"
language: "English"
series: "AI for Literacy Achievement"
instructors: ["Lizzie Jones"]
recording: "https://www.youtube.com/watch?v=mNRd-7efBI4"
transcript: "automatic, from the recording"
source: "InnovateUS (innovate-us.org)"
licence: "CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org"
---

# Reading Challenges Revealed: AI Innovation in Dyslexia Assessments

InnovateUS workshop · 25 March 2025 · 60 minutes

Discover how AI advancements are revolutionizing dyslexia screening and support in this interactive session. Learn about the latest advancements in identifying and addressing reading challenges with precision and equity. Through hands-on demonstrations of the latest tools and expert insights, workshop participants will explore practical ways to integrate these innovations into your literacy strategies and empower diverse learner.

## Learning goals

- Discover how AI advancements are revolutionizing dyslexia screening and support 

- Learn about the latest advancements in identifying and addressing reading challenges with precisions and equity

- Explore practical ways to integrate these innovations into your literacy strategies and empower diverse learners

## Speakers

- **Lizzie Jones**, Program Director, The Learning Agency

Part of the series [AI for Literacy Achievement](https://innovate-us.org/workshop-series/ai-for-literacy-achievement).

Recording: https://www.youtube.com/watch?v=mNRd-7efBI4

## Transcript

Automatic transcript of the recording: names and terms can be misheard, and the times can be a few seconds off. Each time links to that moment of the video. The transcript alone: https://innovate-us.org/reading-challenges-revealed-ai-innovation-in-dyslexia-assessment.txt (plain text) and https://innovate-us.org/reading-challenges-revealed-ai-innovation-in-dyslexia-assessment.json (every segment with its start and end second).

**David Fields**

[0:00](https://youtu.be/mNRd-7efBI4?t=0) workshop of Rethinking Reading, AI for Literacy Achievement. This is an Innovate US series sponsored by the Burns Center for Social Change's Equitable Literacy Initiative and led by our partners at the Learning Agency. In this series, we showcase how to innovate or how innovative leaders are using AI to make rapid gains in advancing literacy in today's classrooms. And you can find the full schedule and register for all the upcoming workshops at the initiative website. We'll drop that link in the chat so you have it. My name is David Fields. I'm a senior fellow here at the Burns Center for Social Change at Northeastern University. And the Burns Center designs and implements practical solutions to society's hardest problems. We develop innovative and participatory and equitable approaches to solving public problems using new technology. For those of you who are not yet familiar with Innovate US, this is a program out of the Burns Center. I encourage you to visit the Innovate US website, and we'll put that in the chat too, and check out hundreds of free 60-minute workshops and at-your-own-pace courses in AI and digital skills designed specifically for public sector professionals. So without further ado, let me introduce Lizzy. Lizzy has a range of experiences in education and research. In prior roles, she worked as a Spanish instructor at the University of North Carolina at Chapel Hill, and in several non-profit and governmental education programs with a focus on marginalized communities in the United States and in Latin America. Lizzy brings her passion for educational access and equity to the Learning Agency. So without further ado, I'm pleased to hand it over to Lizzy Jones, Program Director for the Learning Agency. Amazing. Thank you so much for that

**Lizzie Jones**

[2:00](https://youtu.be/mNRd-7efBI4?t=120) wonderful introduction, David. I really appreciate it. Before I really jump into the content, I have a very short poll for our audience today, because I'd love to know a little bit more about all of you. We got some locations, which is great, but if we can go ahead and launch that, the poll will give us a little bit more information about how familiar you all are. And if we can go ahead and make sure folks are muted too. I think we're getting a little background noise. There we go. How familiar folks are with dyslexia screening and identification, how familiar you are with AI powered tools that can support dyslexia screening, and then also your sector. I know we have a diverse audience and this likely doesn't capture all of you, so feel free to pick the one that best fits your work. And we'll just let those responses trickle in for a little bit. Amazing. So I see what I'm seeing so far, just to give some voiceover, is we have a range of familiarity levels with it being kind of concentrated in that somewhat to not very familiar. And then some folks who identify as not familiar at all.

[3:08](https://youtu.be/mNRd-7efBI4?t=188) So I think we have a real range of interest here, which will be really great and should lead to some good Q&A. And then familiarity with AI powered tools for dyslexia screening. So this is interesting. We had people who are very familiar with dyslexia screening who might be less familiar with some of these tools that now exist in terms of AI for supporting that dyslexia screening, which is great. And then in terms of the work that you all do, see classroom teachers, some of my favorite people to always have in an audience for something like this, district and state leaders, research, and then a good smattering of other. Amazing. Great. Well, to share, to kind of jump into the presentation, like David said, I'm a, my name is Lizzie Jones, and I'm a program director at the Learning Agency. I lost my notes for a second. Great. And I'll tell a little bit more about myself in a moment. But before we move there, I'd like to just talk about what we can expect in this call today. So this session, like David shared, is part of the AI and literacy workshop series that the Learning Agency is running in partnership with Innovate US and the Byrne Center for Social Change. And so this is the fifth workshop in the series. I'm doing our last one that will happen in April, and we have one more in a couple of weeks. And I know we have some repeat attendees in the audience, which is great. And then before we jump into our content, I wanted to take a moment and share a little bit more about me.

[4:43](https://youtu.be/mNRd-7efBI4?t=283) David gave a great overview, but just to kind of anchor me and what I bring to the session today. So like you know, I work at the Learning Agency. The Learning Agency is an innovation hub that is supporting tech powered advancements at all levels of education. And I work as a program director. In my work, I support a variety of our efforts, including scoping new projects that are really at the cutting edge of technology, and then also field building efforts to make sure the voices of educators and other stakeholders are included in the conversations we're having about tech development and deployment. Professionally, I started working directly with students for about a decade before coming to the Learning Agency. My first job out of college was working in a rural community in Costa Rica, offering wraparound support for students there. And then I eventually became a fifth to 12th grade Spanish teacher before going on to get my PhD and teach at UNC with college students. So I've really covered the spectrum from pre-K all the way to higher ed. And then I also periodically worked for the North Carolina Migrant Education Program, providing wraparound support and after school tutoring services for students and their families who came to the area in pursuit of agricultural work. Personally, I'm also the parent of young children who are just starting their literacy journeys. And one of my kids is also neurodivergent.

[6:04](https://youtu.be/mNRd-7efBI4?t=364) So I really get from a very personal perspective, the complex journey that families go on from noticing an area of support is needed to getting the evaluation for that support and then working with the school system to ensure everything is functioning together to give that student the ideal circumstances they need to be successful. And I combine all of these areas here in our conversation today. And so today's workshop will focus on exploring some interesting tools that caregivers, educators, district leaders, and school leaders can leverage for the early identification of dyslexia risk factors. I'll be spotlighting screening tools that enable more efficient, accessible, and scalable identification of learners who are struggling with reading. Without requiring someone to be specially trained in the assessment or long delays in evaluations, frequently schools can be backlogged and people go to private clinics, which creates issues with cost and timing and things like that. But to give a couple caveats with what you can expect here today before we dive in, while the tools that you're going to see are very compelling and can speed up and scale educators' ability to identify students at risk for dyslexia, further evaluation is still required to get to that point of a diagnosis. Think of these tools as accelerating the pipeline from initial screening to identification to intervention, rather than getting you straight to the finish line. An analogy I recently heard is that when frequently in conversations about AI, we talk about it like it's the equivalent of the loom during the industrial revolution, which replaced human labor. When instead, we should be talking about it as being like a crane, a piece of equipment that really amplifies what humans can do. So that's what a lot of these tools are doing. They are not replacing the support and intervention that is needed from our educators. Here we go. Oh, we're getting some, I think a little background shatter there again. But they are instead accelerating how many, how we can get students to that point

[8:11](https://youtu.be/mNRd-7efBI4?t=491) of getting the intervention they need. And some of these tools do still require adult supervision, particularly for our youngest learners. But still, these are promising innovations that can help speed up, particularly that initial screening process and scale school's ability to catch learners who are struggling earlier. I also think they serve the benefit of typically, when you're in the classroom, it's pretty easy to spot students who are at the polls. You see the really high performers and you see the students who are really struggling. But that big spectrum of students in between can be a lot harder to capture. And because these tools enable you to do universal screening with a large group of students, you can get data points for students who fall all across that range. And then we've been asking folks to please, please feel free to introduce yourself in the chat. So don't hesitate to do that as we get going into the presentation. Let's get started. So before I start talking about the intersection of AI and dyslexia screening, I think it's going to be helpful to spend a little time defining what dyslexia is.

[9:20](https://youtu.be/mNRd-7efBI4?t=560) The popular understanding of dyslexia is very oversimplified. I remember when I was young, thinking that dyslexia was just getting words flipped. I remember watching, I was watching a movie one time where a character had dyslexia and the movie portrayed it as the letters moving on the page. And I think that is a preconception that still exists in a lot of people's minds about dyslexia. But dyslexia is defined as a neurodevelopmental disorder, most commonly understood as creating difficulty with reading. But that definition really only scratches the surface of what dyslexia is. To understand how these screeners work, I think it's helpful to talk about those literacy challenges that learners with dyslexia may face during their time in school. So learners with dyslexia typically face challenges with decoding, fluency, and encoding. One of the most widely known indicators, though it's certainly not the only one, is a challenge with phonological awareness or the ability to understand and manipulate sounds in spoken language. This is a foundational skill that occurs before kids even start reading in school. That said, research for dyslexia has really skyrocketed in recent years. And we have a much, we're starting to get a much clearer picture of both the weaknesses but also the strengths that students with dyslexia bring to the classroom and to just life in general. But other indicators that are connected with dyslexia include word reading accuracy and speed. And researchers are increasingly pushing to broaden our definition of dyslexia to encompass causal factors beyond phonological awareness and studying other aspects like visual processing as well. Importantly, these factors do not look the same for any one person.

[11:07](https://youtu.be/mNRd-7efBI4?t=667) Given that variability, developing early, accurate screening tools that can provide insights into the specific subsets of skills learners are struggling with is really key. But traditional dyslexia screening methods typically require manual assessments that can be time-consuming, cost-intensive, and typically need to be administered by specialists. Tech-powered tools, including AI, offer a transformational opportunity because they enable faster and more scalable early identification of students struggling with reading. By analyzing skill-based data, these tools can quickly surface insights on specific areas where students need support. As an example of the speed and scalability here, ROAR, which is one of the tools I'll demo later in this presentation, and it uses some AI components, the whole suite is AI-powered, claims that their platform can assess an entire district of students in the time it would typically take to assess one student for those dyslexia risk factors.

[12:11](https://youtu.be/mNRd-7efBI4?t=731) The goal is that teachers can catch those students who are struggling, but maybe less obviously so, and to create more opportunities for screening within a school year. I mentioned my work in the North Carolina Migrant Education Program. I would not infrequently get students who, because they moved around every three or so months, could really fall through the cracks, and a screening like this, done periodically, could have helped capture some of those students. I once talked to a third-grade student who didn't know how to read, but our students are very smart, so she had gotten very adept at masking what was going on, at memorizing things, at getting good at imitating those around her, so she had been able to make it that far because there just wasn't, by no fault of any educator in her life, she just wasn't in a single location long enough to get identified by traditional means. And then AI models today can capture key literacy indicators like phonological awareness, decoding fluency, and reaction time, and sometimes draw on speech recordings, speech recognition, and gamified interactions with learners. In terms of AI and dyslexia screening, this is a little bit of a timeline of how these tools were developed over time. The field has advanced significantly in recent years, and the rapid increase of AI technologies in general is only accelerating that. So between the 1990s and early 2000s, researchers explored different ways to identify dyslexia, such as using brain scans, eye movement tracking, and handwriting analysis. These studies laid the foundation for future tools but didn't apply AI necessarily yet. In the late 2000s, there were early efforts to digitize traditional paper-based screeners to make them easier to administer and score. Around this time, researchers were also beginning to explore speech data for dyslexia identification. In the 2010s, machine learning and AI began to be applied for dyslexia detection. Examples include researchers using data

[14:18](https://youtu.be/mNRd-7efBI4?t=858) collected through eye tracking and MRIs to automate and predict dyslexia assessments. In recent years, beginning around 2020, and I have a theory as to why 2020 is the year, there was an increase in studies for harnessing AI for dyslexia screening. AI-powered tools are becoming more accurate, and tools for school and home use have emerged. In 2022 and 2023 specifically, there's a significant increase in the number of studies published on dyslexia detection, and I think that's because there's this kind of overlap in the Venn diagram between literacy getting a lot of attention post-COVID when there was so much COVID learning lost. I'm sure many of you recent saw the recent NAEP scores that came out. Combined with these rapid advancements in AI have made it so a lot more innovation has happened in a very short time around dyslexia and AI-powered tools. And there are several key innovators who are kind of leading the charge in this space.

[15:18](https://youtu.be/mNRd-7efBI4?t=918) These tools are designed to be fast, engaging, accessible, cost varies, but generally accessible, helping educators screen students without the need for intensive training or one-on-one administration. And I'll just review a few of these. DysTech is an Australia-based organization with an AI-powered screener that uses speech analysis to predict risk in under 10 minutes. ROAR, which stands for the Rapid Online Assessment of Reading, is a gamified assessment suite that provides real-time data to educators while screening foundational reading skills across pre-K-12. And it uses a mix of AI and non-AI components in the assessment suite. Early Bird Education offers a game-based screener designed for early learners ages four to eight. And it flags early literacy challenges and delivers tailored action plans for both educators and parents. It uses voice recognition in terms of AI. Amira Learning listens to students read out loud. You might sense a theme here. There's a lot of students reading out loud. Using AI to detect dyslexia risk and then match learners with micro-lessons and teachers with instructional support. And then Lexplore is a UK-based organization that applies eye tracking and AI for rapid reading assessments. While the platform doesn't explicitly screen for dyslexia, it does help identify students at risk for reading difficulties, so connect it.

[16:44](https://youtu.be/mNRd-7efBI4?t=1004) There are several research questions that are shaping the field and its future directions. As AI tools for dyslexia screening and support evolve, there's a growing need to ensure they're built on representative data. And in the case of tools that recognize that leverage speech recognition, for example, the models still require improvement so that their accuracy for children matches the near human levels we've achieved for adult learners. This is just because children's speech is different. If you've ever heard a kid try to give a command to an Alexa or a Siri, you've seen these difficulties in real time. And then with children's speech, there's also unique things that you want to make sure don't happen. For example, a child with a speech disorder could get accidentally flagged as having literacy challenges, even though they just talk differently. And this is same with English language learners and people with just non-standard accents. And then AI models require access to increased data across diverse settings, such as noisy classrooms, and among learners with variable speech and linguistic patterns to ensure that these tools are working accurately for learners of diverse backgrounds. This also ties to a question around the generalizability of screening tools. While assessments may demonstrate strong performance among specific subsets of populations, studies don't always demonstrate how well screeners perform across broader, more diverse populations. This is changing, though, over time. Creating more representative benchmark data sets to inform tool refinement is going to be critical for the future of these tools. And then another major area of opportunity is using screening data to not only flag risk, but to better inform personalized interventions. As I mentioned, some of the tools that I shared earlier in the presentation do provide teachers and parents with recommendations.

[18:35](https://youtu.be/mNRd-7efBI4?t=1115) But as we come to better understand the subset of skills different students struggle with, further research could explore how tools can tailor support to these specific themes. For students with dyslexia, ideally we would provide interventions tailored to specific dyslexia subtypes, but that's a practice that's currently harder to achieve in reality, but would be super beneficial to those learners. In the future, we might also be able to explore how tools can leverage more multimodal assessments to create a fuller picture of a learner's reading abilities. At the Learning Agency, we spotlight emerging technologies like automatic speech recognition to identify where increased attention and investment can drive advances in literacy proficiency. We also have a couple of other programs that we run and partners that we frequently collaborate with that kind of connect to this landscape. One of those is Renaissance Philanthropy. We partner with them to lead the tools competition, which is the world's largest tech competition, and several past winners, including DysTech and Roar, who I'll be demoing later in this presentation, are helping advance dyslexia screening. And then we do field building efforts like conducting educator surveys to better understand the real world classroom needs for AI-powered literacy tools. In our most recent survey of more than 650 educators, we found that educators ranked speech recognition tools among the top three most useful interventions for teachers out of the eight options that were provided. Okay, now that we have a sense of the landscape, I'd love to go into a bit of detail on a couple of the tools I mentioned earlier for you all to get a real grasp of how these work in practice.

[20:18](https://youtu.be/mNRd-7efBI4?t=1218) First, we're going to start with DysTech. And I've chosen DysTech as a first tool for a couple of reasons. One is that it's relatively affordable. You can go to their website, you can sign up for $11 US, United States, $9 Australian, it's an Australian-based company, will get you access to the platform and to 20 screenings. Until at like two days ago, it also had a free dyslexia screener on the website. Now the link is broken. We reached out but didn't hear back from the team about that. And then we're also at the learning agency, we're familiar with them because they were a tools competition winner. It's also very easy to get your own subscription. For many of these platforms, if you want a demo of the tool, you have to like sign up for an appointment and things like that. If you are just AI curious when it comes to dyslexia, this is an easy tool to be able to access on your own, check out with low cost. In the screener, it will prompt a student.

[21:19](https://youtu.be/mNRd-7efBI4?t=1279) So first, looking at this, this is I signed up for this platform and made myself a couple of test students. Test two and test test were the creative names I came up with. And once I made my students and gave them fake ages and things like that, I was able to assign them assessments. So this is what the student profile looks like. So this is test two, who's 10 years old. We don't know this student's dyslexia likelihood yet because they haven't done enough assessments to get that data. And then from the list below, I can start assessments to do with this student. The one that I'll show you screenshots of is the word reading assessment. So from this part of the platform in the back end, I can click start assessment and launch it for that student. And then there's a couple of options from there. You can do it in person, which assumes the student will be sharing your screen with you, or you can do it remotely as well. For this assessment, there are there are 48 words that the student has to read out loud. These are a mix of irregular words, real words, and pseudo words that the student reads out loud. And then there's also some additional educator instructions and quick tips and things like that. Like this test is going to be much more accurate if the student is in a quiet room. And then you can see it's very bare bones. You just have a word on a screen and you read it. During the assessment, there is a little button where you can click next word and things like that. But the student just reads all 48 words and then the results go immediately to the back end where the teacher can see how that student performed. And this is a good point of comparison to ROAR, which I'll demo here in just a moment, which has a very different user interface. But DISTEC, even though it's a pretty no-frills platform, the underlying algorithm used to predict the likelihood of dyslexia was trained on data from both standard readers and

[23:23](https://youtu.be/mNRd-7efBI4?t=1403) students with dyslexia to make it more accurate. And the team currently states that their tool achieves 90 percent accuracy in terms of identifying students with risk factors for dyslexia. And like I said, the screener provides immediate results on the likelihood of dyslexia and also offers feedback and recommendations for parents to seek additional support for their children. This next tool is ROAR, which we'll dig into in more depth because it's just kind of, it's a more engaging and comprehensive assessment suite. It's also based here in the United States. And so even though ROAR is not completely powered by AI, I've chosen their platform for the demo today for a few different reasons. One is that it's open source and free for users. And it also shows how these AI and non-AI components link up to create a more comprehensive assessment suite. I know we have at least one person who is from California here in the room, and ROAR was also recently approved by the state of California as a dyslexia screening tool for students in first and second grade. So in terms of state-level widespread implementation, ROAR is a really good example of that. In the suite of assessments in ROAR, which was developed at Stanford's Reading and Dyslexia Research Program, there are multiple short assessments. There's ROAR Word, which includes measures of sight word recognition, which is an important factor for reading fluency and comprehension. There's ROAR Phoneme, which measures reading efficiency. ROAR Sentence, which measures sentence reading efficiency, and ROAR Letter, which measures letter and sound knowledge.

[25:07](https://youtu.be/mNRd-7efBI4?t=1507) There are also additional assessments in the ROAR Dyslexia Suite, like ROAR Rapid Automized Naming is one example there. That is also the only component of the platform that requires the student to speak. Everything else the student can do on their own iPad or computer with headphones on to listen to the instructions. And there are also additional measures that are under research and haven't been validated yet, but are in process with the team. One of those is ROAR Rapid Visual Processing. Oh, ROAR is also on Ohio's approved dyslexia screener list, which is great. Then ROAR is also developing several math tools to help with math facts and things like that, and the Spanish version of the literacy screening suite is also currently being validated. ROAR is designed for a range of ages from pre-k to 12th grade, and it easily integrates with other popular classroom platforms like Clever and ClassLink, which can make it easier to scale for districts. And then within the platform, educators have access to real-time results through a dedicated dashboard that appear as soon as the student is done completing the assessment.

[26:20](https://youtu.be/mNRd-7efBI4?t=1580) And so, you know, the tool is being evaluated still in terms of all components as a dyslexia screener, but so far studies have shown very promising results. In a validation study comparing ROAR to widely used screeners like the Woodcock Johnson, ROAR Word achieved high accuracy and predicting dyslexia for 1st and 2nd graders. For earlier learners in kindergarten, the predictions were less accurate, but improved in combination with other assessments of phonological awareness and letter sound knowledge. And then this ROAR Spanish suite, which they're very explicit, is not just a translation. It is made specifically for Spanish language speakers. It's huge because if we wait for English language learners who speak Spanish as their first language to gain those English skills, we could really miss some early intervention time. So by being able to identify those students with literacy challenges earlier in their home language, we can go ahead and start providing that intervention early on. So now that we have a sense of this landscape, I'd love to just demo two of the assessments within ROAR and show a little bit how they work.

[27:31](https://youtu.be/mNRd-7efBI4?t=1651) And then after that, we're also going to go and look at what the reports look like, both for individual students, but then also groups. And there are demos available for all of the assessments that make up the ROAR suite, but we're going to focus on ROAR Word, which can be administered in just 5 to 10 minutes, and ROAR Sentence, which also only takes a few minutes to complete. And like I said before, not all components of ROAR use artificial intelligence, but they do use a variety of tech components to make the suite as comprehensive and useful as possible. ROAR uses AI within the sentence assessments to create test forms, which you'll get to see what those look like. And then other aspects of the assessment suite use approaches like computer adaptive testing, which you'll get to see in the demo, and I'll explain a little bit more of what that is. For the sake of our session today, though, I thought it was a compelling platform to demo because of its free open source nature, it's research-backed, and because the ROAR team just offers a lot of supports for teachers and people who are implementing this platform that are also for free and offered on their website. They do weekly office hours, they do PD, so there's a lot of support that goes beyond just the tool itself. And I like that you don't have to schedule a call with a salesperson to be able to access it. So with that said, we'll go over here to our demo of ROAR Word. Okay, great. So I'll press continue here. Welcome to the world of lexicality.

[29:06](https://youtu.be/mNRd-7efBI4?t=1746) You are a wizard searching for the gate that will return you to your home on Earth. To reach the gate, you must journey over lands ruled by magical guardians. In order for the guardians to let you pass through the land, you will have to tell the difference between made-up words and real words. As you can tell, very different interface than we saw in DISTECC. It's meant to be engaging, lightly gamified, so the student doesn't feel like they're just taking a test at the computer. A real or made-up word will flash very quickly at the center of the screen. The made-up words might look like English words, but they do not mean anything in English. For example, if you see a made-up word, press the left. So you see 3, 2, 1, the word will pop up. In this case, it's ZUP, fake word, right? So I'll sort that. You press. How, real word.

[30:07](https://youtu.be/mNRd-7efBI4?t=1807) And this is just the student who's doing this part of the assessment. After. You press the right arrow key. Holler, fake word. You press the left arrow key. Palm, fake word. You press the left arrow. Great work. You are ready to begin the journey. You will earn gold coins along the way. Importantly, word is a computer adaptive assessment, which is part of the reason. Oh. I think I just learned from someone in the chat that I was talking at the same time as the assessment. Sorry, it was muted on my end. So I thought it would be muted for you too. I'm so sorry. So to just, I'll give a little voiceover and then go back to the assessment for a second. But I was just trying to say thank you, Jen, for letting me know because I thought it was just all happening separately. But in the assessment, you see that initial section where you're just sorting words and the student gets to test run it.

[31:11](https://youtu.be/mNRd-7efBI4?t=1871) Then they get to earn gold coins along the way, right? And after that, word is a computer adaptive test, which means as a student is going through, it's going to adjust to their progress and performance. So if a student is getting a lot of words wrong, the platform will make it easier over time. And if they're getting a lot of things right, it will make it more difficult. This is supposed to make it so it's just less frustrating for the learner. So you saw generally the functionality there. So I will pull it up and let the friendly voice keep talking to you. But as you go along in the platform, you get to see, you collect coins, you meet different little characters along the way, and then you finally get to the end point of lexicality. And it only takes five to 10 minutes to get a good assessment of that student's capabilities. Great. And now I'm going to leave Roar Word and I'll pull up Roar Sentence in just a moment.

[32:16](https://youtu.be/mNRd-7efBI4?t=1936) Roar Sentence does use AI to generate infinite test forms. And many of you who are classroom educators might be surprised by the fact that it is an assessment of reading that is silent. So typically when we think about assessing a student's reading capabilities, we're thinking about the teacher being there with them and having the student read a sentence out loud and the teacher is able to hear in real time how the student is doing. Roar Sentence is completely silent. So the student is able to read to themselves and they indicate in the platform if a sentence is true or false. And some of the sentences are like just kind of funny. They're things like ducks wear hats on their feet and things like that, but they are intentionally designed to be widely known constructs and facts. So I'm sure many of you have heard stories of when they tried to create universal tests and like standardized tests back in the day, there were questions on like, what season do you eat clam chowder in?

[33:24](https://youtu.be/mNRd-7efBI4?t=2004) That was going to students in Florida who were like, I've never had clam chowder before. So they try to keep the content of these sentences as universal as possible. So with this, I'll just show you what the tool looks like really quickly. And then I'll talk a little bit about some of these computer adaptive components and things like that. Great, I'm gonna turn on my sound again so I can make sure you hear what you all are hearing. What grade are you in? To get started, press the number key on your keyboard of the grade you are in. For example, you would press four. Welcome, you are the new professor at Sentencia Academy. We have some bright students that need your help with their sentence reading homework. You will help these students decide if a sentence is true or not. Some of the sentences may be confusing, but do your best to label as many as you can. Press any key to continue.

**David Fields**

[34:26](https://youtu.be/mNRd-7efBI4?t=2066) Lizzie, I think we're seeing your notes and not the screen.

**Lizzie Jones**

[34:29](https://youtu.be/mNRd-7efBI4?t=2069) Oh, which are very lovely notes, but probably not what you need, so let me share that. So this is, you're a student at Sentencia Academy, and after this screen, you will get to pick your character then you get to be in the game. For students above grade five, they will remove some of those components that are a little bit more childlike, right? But I'll go ahead and show you what that looks like. Now let's meet your students. Press the number key on your keyboard to pick the student you would like to work with. Press a number key to pick a student. And so I'm gonna pick this nice green little bird here. Hi, professor, it's nice to meet you. My name is Jayden. I'm having trouble with my sentence reading homework. I need to finish it so I can hang out with my friends. Can you help me? Press any key to continue. And then I'll just show you a couple of the sentences and we'll move on from this part.

[35:29](https://youtu.be/mNRd-7efBI4?t=2129) In this activity, you're. So this is very similar to a ROAR word where left, you pick false, right, you pick true, and students are told part of this is the speed that you do it with. So an apple is blue. Correct. False. You pressed the. And it does this throughout. And then you get a couple of check-ins from your student along the way who tells you thank you so much for helping me with my homework and things like that to try and keep students motivated. And I think that the framing for students is certainly intentional with, especially for students to be put in that position where they get to pretend to be the teacher with the student is also kind of a unique take in the platform. So now that we've looked at the assessment a little bit, I want to share some of these score reports that are offered. And we're not going to watch this whole thing, but just to give you a sense of what this looks like.

[36:35](https://youtu.be/mNRd-7efBI4?t=2195) You through accessing and understanding the individual reports. This report provides an in-depth view of a single student's performance. To access an individual report, start by navigating to your district, school, or classroom score report. Find your administration, then click the dropdown button labeled show details. Depending on your permissions, you can view district, school, and classroom reports. This button will open the score report. Once you've opened your report, scroll down to the main data table. You'll see the first column titled report, followed by columns for first and last names. Find the student whose report you want to view and click on the hyperlinked graph symbol. And then with this, the colors, as you can imagine, match up with how well or struggling that student is doing with a certain skill. So pink indicates they need more support. Green, they got it.

[37:35](https://youtu.be/mNRd-7efBI4?t=2255) Yellow, something to keep an eye on. So to not spend too much time there, I also wanted to share what these reports look like, which can give you information for an entire school, district, or group of schools. Through the features, how to use them, and how to make the score report can be found in the video, logging into your educator account. First, let's take a look at the scores at a glance section. This section uses pie charts to give an overview of your student's performance in the core foundational measures, word, sentence, and phoneme. What we're viewing here is a visual summary of student performance. If you hover over each section of the pie chart, you'll be able to see the number of students that scored in that support category. The legend for the support categories is below the graphs. Green for achieved skill, yellow for developing skill, and pink for needs extra support.

[38:41](https://youtu.be/mNRd-7efBI4?t=2321) As we scroll down, we come to the main data table, which contains individual scores for all of the assessments. Use this scroll bar to see all of the data. You can customize your data table in several ways. First, you can filter the table by school and by grade. This dropdown menu allows you to select multiple schools. Similarly, this dropdown menu allows you to select the grades you are interested in viewing. Next, you can select which columns you would like to be visible in your data table. This dropdown can be used to add or remove columns. Lastly, you can freeze any columns to keep them in view as you scroll horizontally. Moving on to the header row, you can sort each column by clicking on the up and down arrows. You can also apply filters to an individual column by clicking on this filter icon and entering a value. I won't go into all of the details here, but there are other visualizations that you can also access to really see that data.

[39:49](https://youtu.be/mNRd-7efBI4?t=2389) So there's pie charts, there's numbers and percentages, then you can see individual scores. So you can really get as granular or as macro as you need to be for what you're trying to look for and the students that you are serving. First. So before I open things up to Q&A here in just a minute with David, I wanted to just summarize a couple of these key takeaways that I hope you all have been able to absorb from some of this, but just to recap. So in the past several years, we've seen major advances in the application of AI for literacy screening, particularly in the development of tools that can screen for the risk of dyslexia. These include tools that analyze real-time speech for risk factors, as well as others that assess reading data and track eye movements. Screeners like Roar and DysTech, but also others are making it fast and easier for teachers to screen students for these risk factors while achieving high levels of accuracy. So they can allow students to screen more students more often without the heavy time that is typically required for traditional assessments.

[40:54](https://youtu.be/mNRd-7efBI4?t=2454) These tools also collect a lot of data that can be super rich and really empowering for teachers as they give a snapshot of students' literacy skills, how they're developing, and where they may need extra support. And then in terms of where these developments need to go next and where researchers are focusing on making major advancements, while these tools are reaching compelling levels of accuracy, which is great, innovations can always do better. So especially for thinking about one, that leveraged speech recognition for children, for example, developing those further to reach that same level of accuracy as we have for adults is one of those big next steps and is a real kind of opportunity space. There's ongoing work to improve accuracy across learners of diverse backgrounds, and especially students with different language backgrounds. And then with more representative data and research into the best ways to link assessment findings to personalized interventions so that teachers can better target the specific skills students need. Okay, and with that, I'll pull the deck down and can be here with David to answer any questions.

**David Fields**

[42:04](https://youtu.be/mNRd-7efBI4?t=2524) Great, thank you, Lizzie. That was great. Learned a ton. And before we dive in, I just wanted to acknowledge Rick Burns, who I saw is on the call with us, and I'm so excited to see him here. We're so thankful for his support in all of this. And so let's dive in. We've got some great questions here. So just to start off, you mentioned that literacy levels declined after COVID-19. Could you elaborate on this? What do you see as the main reasons behind the drop?

**Lizzie Jones**

[42:39](https://youtu.be/mNRd-7efBI4?t=2559) I mean, I think many of us who had young kids during COVID, you know, it was just, it was a time when I think we saw some of our prior differences between student groups get really exacerbated. Students who had parents who were able to at least work from home and help them, students who, the tech divide was a really big gap. You know, and it was, everyone was, I think, working extremely hard to take care of their students in the best way possible. But the part of living through unprecedented times is that we didn't have a full toolbox and we're making, you know, so much has happened in the past few years. It really feels like a whole different lifetime ago. And recent NAEP scores have shown that we just haven't bounced back and some of those key skills and the way that all of us would like. And so I think giving everyone who is serving our young learners from parents to teachers and district and state leaders, as many tools as possible to close as many of those gaps as possible is only going to be a benefit.

**David Fields**

[43:46](https://youtu.be/mNRd-7efBI4?t=2626) That's great. That's great. I was looking at our questions here. So I've got one in that is, says ROAR's reporting features seem like a powerful tool for understanding how a school or district is performing. Do you know if Department of Education, such as in Ohio or California, are using it to monitor reading performance at a broader level?

**Lizzie Jones**

[44:11](https://youtu.be/mNRd-7efBI4?t=2651) I know that, so ROAR has worked with, I think there are 23 states maybe now at this point. And I do think one of the goals is to be able to monitor reading progress at the state level, in district level, school level. So really comprehensive. Even though I'm talking specifically about dyslexia here today, it is not just made for learners with dyslexia. Even students who don't have dyslexia might struggle with many of these components and ROAR is made to be able to spot that. So for example, it's able to be used by students from pre-K to 12th grade. We would hope, though it doesn't always happen, that by the time a student reaches 12th grade, they've been caught already and identified if they have dyslexia. But we know that we're getting students in middle school and high school with reading. gaps in some of these foundational skills, and being able to really zone in on those, even for older learners, is super necessary. There's been some kind of concerning research and data that for students, there's states where they've caught students up and have really boosted literacy rates for students up to third grade, that frequently goes down by the time students reach middle school. And they're required to do some of that more complex decoding to really understand what they're reading. And so Roar is able to support all of that. And then I also just think they're unique in having a research practice partnership model. If you're interested in working with them, or just learning more, you can go to a free office hours and talk to them. Another thing that I listened to as I was preparing for this, I listened to a recent webinar that they did.

[45:55](https://youtu.be/mNRd-7efBI4?t=2755) And one of the other things they were talking about is you cannot currently use Roar to disaggregate data for certain student groups. So you can't pull it out and see how are all of your English language learners doing, for example. And that is something where, and that's because for a variety of reasons, that could be problematic, too, of zoning in on certain groups. And so that was something that they talked about as like another functionality of the data that the team is exploring, how they could do something like that, but implement it the way it needs to be implemented. So it's ethical and protect students. That's great. And is there an age

**David Fields**

[46:34](https://youtu.be/mNRd-7efBI4?t=2794) that's kind of most effective to test children for dyslexia?

**Lizzie Jones**

[46:39](https://youtu.be/mNRd-7efBI4?t=2799) The earlier the better. There's some pretty compelling research that if students are not, don't have good foundational skills by the time they reach third grade, it's really hard to catch them up after that. Some of the screening tools that Roar is currently working on and that others are working on are identifying some of these indicators of dyslexia that are pre-reading. People used to think and researchers used to think that dyslexia was something that appeared in elementary school. And we now know that's not the case. Dyslexia is hereditary. It's neurodevelopmental. You're born with it. And so what are the signs that we can pick up as early as pre-K when students don't have that letter naming and recognition necessarily that we would test for in one of the currently available assessment tools, but would be one of those dyslexia indicators that would at least signal to people that it's something we need to keep an eye on. But I also was listening, the brain is amazing. So even learners that are getting diagnosed as adults, there's a lot of adults who slipped through the cracks when they were young learners for a variety of reasons. Reading is incredibly complicated, but it's never too late to retrain your brain and get some of those foundational skills that you may be lacking.

**David Fields**

[47:59](https://youtu.be/mNRd-7efBI4?t=2879) That's great. And Roar has it, I think you mentioned it's Australian. Is it being used kind of widely outside the U.S.? So there's DysTech is Australian. Roar is based out

**Lizzie Jones**

[48:11](https://youtu.be/mNRd-7efBI4?t=2891) of Stanford. So I know DysTech has been used in the UK and things like that. I do not think it is as nearly as widespread as a platform as Roar is. Roar is also unique because it's just incubated at a research institution. So the research that gets to be around it and things like that, it just really gives it some good meat and backing to it that I think is harder to find in some of the

**David Fields**

[48:34](https://youtu.be/mNRd-7efBI4?t=2914) other commercially available tools. Roar is available outside the U.S.?

**Lizzie Jones**

[48:43](https://youtu.be/mNRd-7efBI4?t=2923) They only released stats on how it's being used within the United States. I think that's part of their mission. But the fact that anyone can access the platform, so it wouldn't surprise

**David Fields**

[48:55](https://youtu.be/mNRd-7efBI4?t=2935) me if they were also getting global reach. Gotcha. And we had a question about where folks can sign

**Lizzie Jones**

[49:01](https://youtu.be/mNRd-7efBI4?t=2941) up for the Roar office hours. Great. I have a link sheet where I can drop access to the Roar website. Here, let me put this in the chat. So if you just go to the Roar website up at the top, you'll see, I think they're doing, um, one of their office hours is starting in an hour, East Coast time, three to six Pacific time. They also have some PD sessions that are coming up. And then if anybody wants to get into like the technical nuts and bolts and you're curious about that, they have a really interesting technical memo, technical guide that goes through the validation studies and really just like everything that they have done to show the efficacy of their tool. And I'll also drop that into the chat.

**David Fields**

[50:07](https://youtu.be/mNRd-7efBI4?t=3007) Um, let's see, we've got, um, another question here. Uh, let's see, Jen, do you want to ask your question directly or I can try and frame this? Okay. She's there. Um, she said there's really great tools and assessments to help students. Um, you mentioned the top performing tools are data rich, um, but the data visuals of the assessment don't appear to be accessible for educators who require the usage of AT to decipher insights from reporting results. How accessible are the assessment tools themselves?

**Lizzie Jones**

[50:49](https://youtu.be/mNRd-7efBI4?t=3049) So I think you, from all of these, you would have to sign up for an educator account and then you should be able to get those visuals. I know that's the case with Roar. Um, with DISTEC, it's the same thing. Once you have an account, you can get it. I do think it's just one of the difficulties with some of the other commercially available platforms that are really great, but it feels like it's hard to get into the backend and see what it looks like because you have to set up an appointment and things like that in order to access it. Um, and then depending on your LMS, Roar is able to integrate with some of those, but not all of them. And this is just kind of one of the difficulties of the United States education system is things can vary so widely from state to state.

* * *

Original page: https://innovate-us.org/reading-challenges-revealed-ai-innovation-in-dyslexia-assessment · Licence: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org