Modal Education

In the United States alone, an estimated 7.3 Million students receive special education services. That equates to about 15% of all public school students in the US, and that number has only grown over the past few decades. Unfortunately, by the time many of these students graduate, they are years behind academically across all subjects. The Individuals with Disabilities Education Act (IDEA) provides government funding and mandates that schools provide support for students on IEPs, but this support is often limited by a number of factors, such as population or geographical constraints. And while supplementary EdTech solutions do exist, the majority are not designed for students with special learning needs, and the few solutions designed with learning needs in mind are often very narrow in scope. This means that teachers often lack the resources to effectively teach students with special learning needs, which in turn often equates to poorer than average academic, financial, and career outcomes (55% of special education students continue to postsecondary education versus 62% of their peers in the general education population, per the National Center for Special Education Research).

Modal Math seeks to leverage machine learning and AI to create Modal Education: accessible, deeply-personalized learning experiences designed to maximize engagement and information retention for PreK-5th grade special education students.

Our goal is to develop a unique, novel machine learning model that will not only allow for adaptive learning paths, but that will also allow fully accessible differentiation based on the needs of each student—a "tutor in a box." In other words, one student's learning path may be full of gamified learning designed to deliver audiovisual content at a pace 1.1x that of standard classroom instruction, while another student's may instead focus on mastery of word problems and spatial reasoning. We will construct predictive models that assess student performance and progress across various modalities, such as American Sign Language, videos, audio, and text. This approach will enable us to tailor content delivery and ultimately improve outcomes for students with special learning needs, and represents a huge leap forward in the realm of adaptive learning and assessment.

A platform like this doesn't currently exist—in fact, Modal Math itself was created by a teacher of the Deaf and Hard of Hearing specifically because resources for special education students were so limited. Adaptive resources designed specifically for students with special learning needs simply aren't yet available. However, the need is great, and growing each year.

The sole purpose of Modal Education is to offer adaptive, personalized education to some of our most disadvantaged and disenfranchised students. There are several ways we'll endeavor to do this:

Personalized Learning Paths: Our primary goal for the next 1-2 years is to develop machine learning models that will enable our platform to analyze student performance data across dozens of indexes (including performance, engagement, and interaction data) to create personalized learning paths tailored to each learner's strengths, weaknesses, and individual learning styles. We want to ensure that students receive targeted support where and when they need it most, with the end goal of overcoming learning barriers. 

Accessibility and Inclusivity: By prioritizing multimodal content delivery—including support for American Sign Language, video, audio, and text—we'll be able to dramatically improve accessibility, equity, and inclusivity for diverse learners. This especially includes students with special learning needs, ESL learners, and students who lacked early developmental resources due to poverty or other reasons.

Closing the Opportunity Gap: Millions of children across the US face deeply-rooted systemic challenges. These challenges disproportionally impact students facing poverty, Black and Latino learners, and students with special learning needs. Modal Education's goal is to ensure that all learners have access to high-quality education resources and opportunities, regardless of their background or socioeconomic status—and we want to be able to offer many of these resources for free.

Early Intervention and Support: Our AI-driven assessment tools are intended to help enable early identification of learning gaps, and to help parents and educators also identify potential intervention strategies. This is critical at all ages, but especially during PreK-5th grade, when students are meant to be forming the educational foundations that they'll need in both academia and the workforce.

Empowering Educators: We want to be able to provide educators with insights, actionable data, and real-time feedback into student progress and learning. This will allow them to better tailor their teaching strategies to meet the needs of their individual students. 

While we want to bake inclusion and equity in our product itself, we also believe overwhelmingly in the importance of inclusion and equity in our distribution and business processes—after all, some schools and families simply don't have the resources to purchase supplemental tools, and many can't afford the laptops or mobile phones they'd require to access the platform at all. To that end, we also intend to:

• run free pilot programs with public schools and community organizations in underrepresented and economically disadvantaged areas to ensure wide access to the program,
• offer free online resources,
• eventually offer offline resources as well

Modal Education was founded Nadia Iftekhar, a certified teacher of the Deaf and Hard of Hearing (HoH). Nadia spent years searching for—and creating—accessible, multimodal resources for her students before she eventually launched Modal Education's MVP: Modal Math. Her educational background is in Educational Psychology, Educational Technology, and Elementary Inclusive Education, and that background has been incredibly valuable as we've grown Modal Math. More valuable, however, are the years she's spent working as a teacher: for the last 8 years she's worked with students in both classroom and virtual settings, one-on-one and in large groups, nearly all of whom have had a diverse range of special learning needs. 

Maria Ada Santos is Modal Educations COO, and is both a former PreK-1st grade educator and a veteran of the EdTech industry. She's spent the majority of her career either working directly with students or managing projects designed to improve access to quality education. Her passion for education is intrinsic, but also stems from her own background as a Latina who grew up attending underfunded public schools. The educational challenges that face Black and Brown students, especially those living in poverty, aren't academic to her; they make up her lived experiences. Maria also grew up alongside a younger sister born with severe developmental disabilities, and she therefore also has a personal connection with and passion for special education. 

The team's areas of expertise include pedagogy, operations, automation, and computer science.

Modal Education's initial, non-adaptive MVP—Modal Math—was launched in 2020. The platform is multimodal and post-revenue, with 100+ active students. We've seen enough success to be reasonably confident in product-market fit, and several teachers have continued to use the platform for years. However, while the current platform is accessible, it's neither adaptive or personalizable. 

We've spent several months conducting research into machine learning models, and have identified potential school partners who may be interested in piloting the program once a prototype is developed. However, because we don't yet have a functioning adaptive prototype, we believe our solution is still in Concept.

There are several aspects to Modal Education that we believe make us innovative, notable among them being that accessible, adaptive solutions designed specifically for students with learning disabilities simply don't exist. However, while that gives us a first-mover advantage, it's only a baseline. We believe that our greatest advantages and innovations are as follows:

Multimodal Data: Modal Education was founded on the belief that most students, especially those with special learning needs, benefit from multimodal instruction. We plan to leverage multimodal data—textual, audio, visual (including American Sign Language), and interactive inputs—to better tailor our machine learning models and each student's individual learning experiences. This approach will not only allow us to accommodate different learning preferences, but will also provide a richer dataset for AI to analyze and adapt to individual learner needs. We believe this will result in personalization and adaptiveness on an unprecedented scale.

AI-Driven Predictive and Adaptive Learning Paths: Our primary goal within the next year is to create learning paths that adapt to students' current learning needs. Once we've reached that goal with a fair amount of confidence, we then plan to adjust our models to also include predictive learning paths: in other words, building adaptive learning paths that anticipate and address each student's potential learning challenges in advance. Nearly every existing EdTech product is reactive, rather than predictive. We want to change that.

Dynamic Feedback and Recommendations: Like many existing EdTech platforms, Modal Education will offer real-time insights into student progress. Unlike most other solutions, however, Modal Education will also be able to provide predictive, comprehensive recommendations for parents and educators based on dozens of student datapoints. The goal is to provide a supplemental tool that will not only make the lives of educators easier, but that will also offer substantive data that educators and administrators will be able to use when creating school policy decisions. We believe this will result in a push to further integrate data-driven decision-making tools across educational settings, and we hope this will in turn lead to a paradigm shift in how educational content and interventions are designed, prioritized, and delivered.

Promoting Equity in Education: As current and former educators, we are extremely familiar with the opportunity gaps that currently face disadvantaged students across the United States—especially students of color, students facing poverty, and students with special learning needs. We're also extremely aware of the vast overlap in these three groups. Modal Education was created to serve students from diverse backgrounds and needs, especially those facing the biggest barriers to opportunity. Our goal is to prove that thoughtfully-designed EdTech solutions have the power to bridge the equity gap in education. Then, we'd like to use that data to encourage policymakers, investors, and educational leaders across the country to help us lead the charge on investing in accessible, adaptive EdTech for all.

Current Core AI Technologies

Machine Learning and Deep Learning: The core of Modal Education's solution will be ML and DL models designed to adaptively curate and personalize learning experiences. This will enable our platform to analyze learner data over time, identify patterns, and create adaptive learning paths. 

Future AI Technologies

Note that we haven't begun to implement the following, but we believe they'll eventually play crucial roles in improving the Modal Education platform:

Computer Vision: STEM students are frequently asked to draw graphs, charts, and create other visual or interactive models. Using computer vision will allow the platform to interpret and respond to some of these visual inputs from users.

Speech Recognition: Some students are unable to use keyboards or trackpads, due to disability, injury, or other various reasons. Implementing speech recognition will enable us to build in voice-activated controls and auditory learning modules, which will only further improve accessibility and inclusivity. 

Natural Language Processing: Our eventual goal is to create a platform so robust that it serves functionally as a tutor-in-a-box. NLP will help us ensure that Modal Education is able to understand student text-based input and provide meaningful, constructive feedback. This will be especially important as we eventually add Language Arts learning paths.

Technology Dependence and Integration 

As Modal Education is still in the Concept phase, we're currently relying heavily on third-party libraries and tools to train our machine learning model. This includes, but isn't limited to:

• TensorFlow
• NumPy
• pandas
• scikit-learn
• Matplotlib

Other tools we're currently using include:

• Colab
• Mockaroo, to create the sample data we'll use to train our models
• Airtable, to experiment with data architecture. Note that Airtable is not used to host genuine student data

In order to build and test quickly, we're currently building a sample front-end using FlutterFlow, a no-code drag-and-drop app builder that allows for rapid building and iteration. Sample data is hosted in Firebase, which will allow us to efficiently manage fluctuating usage without compromising on user experience or data processing. Our eventual platform will likely be built using React and React Native, and will allow for interactive elements such as drag-and-drop activities and virtual labs.

Modal Education is still in the early stages of development, and we're still currently refining our machine learning models. However, while we have not yet implemented a pilot program, we are building our platform based on proven methodologies and the latest research in both pedagogy and educational technology.

1. Development Based on Educational Best Practices: We are developing and designing Modal Education in close alignment with established educational theories related to the benefits of multimodal education. Additionally, our founder is a teacher of the Deaf and a doctoral student in educational technology and psychology, and we are leveraging this background to ensure that the platform we build is suitable and accessible for young learners.

2. Planned Prototype Testing: We are in the process of planning controlled pilot tests within diverse educational settings. These tests will compare our AI-driven adaptive learning system against traditional methods, focusing on engagement levels and comprehension among students. We aim to document these tests thoroughly in order to better refine our models and provide clear metrics and feedback.

3. Proposed Collaborations for Third-Party Evaluation: We intend to partner with independent educational research organizations and institutional review boards to validate our approaches and ensure our platform meets academic, educational, and privacy standards.

4. Engagement with the Educational Community: Our ongoing dialogue with educators and educational technologists—particularly those with experience with disadvantaged groups—has and will continue to profoundly shape our development process, and are vital as we design intuitive, effective learning modules.

Grounding Modal Education's development in rigorous academic and practical foundations will enable us to build a robust, validated platform ready for future testing and implementation. Although we are still in the Concept phase, our commitment to evidence-based development and continuous improvement positions us to create a significant impact for students with special learning needs across the world.

Modal Education's primary goal is to bridge the educational opportunity gap for disadvantaged learners. The only way to do so is to ensure that equity is built into the very DNA of our platform. In order to do so:

Diverse Data: Our pilot programs will largely be conducted in schools with widely diverse populations, to include a range of ethnic backgrounds, socio-economic statuses, and learning abilities. This will reduce the risk of bias that could disadvantage any particular group. 

Bias Monitoring & Iterative Feedback Loops: Our development process will include the implementation of bias monitoring tools designed to detect and alert if the educational outcomes disproportionately favor or disfavor any specific group. This will allow for immediate corrective adjustments in our models. Post-deployment, we will continue to quantitatively and qualitatively evaluate the impact of Modal Education on all student populations. Any indication of bias will lead to immediate reassessment and  updates.

Transparent Development: We plan on maintaining extreme transparency during development by carefully documenting and publishing each of our methodologies, including how data is collected, used, and processed. We will not only welcome but actively solicit independent reviews from third-party experts specializing in ethical AI. 

Ethical AI: In 2021, UNESCO published its Recommendations on the Ethics of Artificial Intelligence. Our team studies and adheres to these recommendations strictly, and will continue to do so as the guidelines are updated to reflect advancements in machine learning.

Stakeholder Engagement: We will maintain an ongoing and continuous dialogue with community leaders, educators, and advocacy groups to ensure our solution aligns with the real-world needs and values of the many communities we serve. This engagement also helps in identifying any unforeseen impacts or areas for improvement. 

We fully believe that this is not only a commitment to equity but also an essential element of our mission. Modal Education will not work if equity and diversity are not built into its core.

Full-time: 1

Part-time: 1 (to become full-time if grant funding is awarded)

Depending on grant funding, we also plan to hire 1-2 additional full-time machine learning experts to help with further development beyond what currently exists, and 1-2 additional part-time educators to assist with content creation.

Our development plan is aggressive, but we fully believe that we will be pilot ready within the next 12-18 months. Our progress thus far includes:

• We have created a content library of dozens of lessons and thousands of curriculum-aligned, multimodal math practice problems for PreK-5th grade students. This content library is not yet adaptive, but is currently being used by 100+ students.
• We've identified 42 data points that our machine learning models will use in order to construct adaptive learning paths for students.
• We've created a mock dataset of 1000+ students, each of which has been provided with randomized data across these 42 data points.
• We've begun testing several unsupervised learning algorithms against this data. While we are still in the early phases of development, our results so far have been exciting.
• We have begun creating mock interfaces using no-code tools.

If selected for the MIT grant, we will expand our team, particularly in areas of AI development and educational content creation, to support the scaling of our platform. The next 9-12 months will be devoted almost entirely to building, testing, and refining our machine learning models, with the goal of creating a platform that will allow us to deliver fully adaptive content to each student. Our development timeline also includes several rounds of testing with simulated user environments, which will transition into small-scale real-user testing in educational settings. This will allow us to gather valuable feedback and make necessary adjustments.

Once we've refined our models and are pilot ready, we will begin initiating conversations with local educational institutions to pilot Modal Education and gather preliminary feedback. We will also explore partnerships with technology providers and educational organizations to ensure support for deployment and growth.

Global access is of paramount importance to us, and there are a number of ways we'll work to ensure it.

Tiered Pricing Model: We plan to implement a tiered pricing structure that varies not only based on number of students but also on student demographics. We want public schools serving large numbers of priority learners to access the platform at reduced rates or for free.

Grant Funding: We will actively seek additional grants and funding opportunities dedicated to educational technology and innovation, to help us offer our platform at minimal or no cost to the most underserved communities.

Multi-Platform Availability: Our solution will be designed to work across various devices, including smartphones, tablets, and computers. We also plan to eventually introduce free, offline materials as well. This will help us ensure that Modal Education is accessible regardless of the technology available at home or in school.

Accessible Design: Our platform was designed for students with special learning needs. We'll ensure that Modal Education's interface is easy to navigate for all age groups and adaptable to the needs of students with disabilities, including those who require screen readers or other assistive technologies.

Continuous Feedback Loop: We plan to establish a continuous feedback loop with all user types—teachers, administration, parents, and students—to allow us to gather constant input on the platform's usability, accessibility, and effectiveness. This will allow us to make necessary adjustments and ensure the platform remains aligned with learner needs and educational standards.

There are two main reasons we're applying to the Learner//Meets//Future Challenge. 

First, there is a breadth of knowledge within this community that defies categorization. MIT Solve and the Bill & Melinda Gates Foundation have created a platform overflowing with driven, ambitious leaders who are all equally passionate about improving educational outcomes for our most disenfranchised students. We want to meet and learn from them, and we hope that we'll be able to share our own knowledge along the way.

Secondly, the grant funding from this challenge will give us the ability to hire additional talent. When we began building Modal Education, we weren't—and still aren't—machine learning experts. We began learning ML and AI because we knew our students needed help, and we've come a remarkably long way. This grant funding would allow us to hire a machine learning engineer, however, which in turn would allow us to accelerate our progress even more.