WeAreAbility

Inspiration WeAreAbility started with people we care about most. Some of our closest friends have dyslexia. Over the years, we watched them often become frustrated with materials that simply were not designed for the way they learn. What always stood out to us was that these were some of the smartest, most creative, and most hardworking people we knew, yet they often had to overcome barriers that others never had to think about. As we spoke with more students, educators, researchers, and advocates, including individuals we met through the Harvard Health Systems Innovation Lab (HSIL), we realized how many others shared similar experiences. We heard stories about feeling overlooked, misunderstood, or forced to fit into educational systems that were never built with them in mind. Those conversations became more than research. They became personal. We began asking ourselves a simple question: what if educational content could adapt to the learner instead of forcing the learner to adapt to the content? That question became the foundation of WeAreAbility.

What it does WeAreAbility is developing Promea, an adaptive AI-powered learning companion specifically designed for learners with dyslexia. The system consists of two major components: a wearable lecture-capture device and an intelligent adaptive learning platform. The wearable component captures auditory and visual information during classroom lectures through integrated microphone and camera systems. Rather than acting as a passive recording device, the wearable serves as the input layer of a larger adaptive learning architecture. To prioritize privacy and responsible AI use, the system follows a data-minimization approach, collecting only the information necessary to support learning and transforming raw lecture inputs into structured educational representations whenever possible. Lecture data is processed through a multimodal AI pipeline that extracts key concepts, vocabulary terms, diagrams, instructional explanations, and contextual relationships between concepts. This information is then transformed into personalized educational content. Students receive dynamically generated learning materials including simplified summaries, structured note outlines, vocabulary support and definitions, AI-generated visual representations, active recall activities, guided comprehension questions, phonological learning exercises, memory reinforcement activities, text-to-speech support, and personalized review pathways. What differentiates Promea from traditional assistive technologies is KAPN, our Knowledge-Augmented Policy Network. Rather than providing static accommodations, KAPN continuously learns from each student's interactions to build an evolving understanding of their strengths, challenges, and learning progress. By combining Bayesian Knowledge Tracing (BKT), Dynamic Bayesian Networks (DBNs), Deep Knowledge Tracing (DKT), Retrieval-Augmented Generation (RAG), and Reinforcement Learning, the system can estimate a learner's current knowledge state, identify emerging knowledge gaps, and predict future learning difficulties before they become significant barriers. It then uses lecture-specific content and learner data to determine the instructional strategy most likely to improve comprehension, retention, and confidence. To promote accuracy and reduce the risk of hallucinations, generated explanations and learning materials are grounded in lecture-specific content through Retrieval-Augmented Generation rather than relying solely on model-generated responses. Human oversight remains central to the platform's design. Promea is intended to augment students and educators, not replace them. Learners retain control over their learning experience, while educators can review, modify, or override recommendations and generated materials when appropriate. As students interact with the platform, KAPN continuously adapts instructional pacing, content complexity, presentation methods, and reinforcement strategies to meet their evolving needs. By combining adaptive AI with transparency, privacy-conscious design, and human-in-the-loop decision-making, Promea aims to deliver personalized support while maintaining trust, accountability, and learner autonomy. In essence, Promea does not simply help students access information—it learns how they learn and continuously personalizes instruction to help them learn more effectively.

How we built it Although the full production system is still under development, every aspect of Promea was shaped by extensive research and conversations with the dyslexic community. We began by studying the challenges dyslexic learners face and examining why many existing educational technologies fail to provide meaningful long-term support. Our research revealed that many dyslexic learners spend a significant amount of cognitive effort decoding and processing information, leaving fewer mental resources available for comprehension, retention, and critical thinking. We also found that dyslexia affects learners in highly individualized ways. While one student may struggle with phonological processing, another may face challenges related to visual attention, working memory, or reading comprehension. These findings reinforced a key insight: effective support must be personalized rather than standardized. We also explored evidence-based learning approaches, including multisensory instruction and adaptive learning systems. Research consistently showed that combining visual, auditory, and interactive learning experiences can significantly improve engagement and understanding. This inspired us to design a platform capable of generating content in multiple formats and adapting instruction based on each learner's needs. To bring these ideas together, we developed a multi-stage adaptive learning architecture. The system begins with learner profiling, where onboarding surveys, accessibility preferences, confidence baselines, and dyslexia-specific learning characteristics are used to create a personalized learner model. This profile helps determine how content should be presented, whether through visual explanations, audio support, structured notes, or other adaptive interventions. Educational content such as lecture transcripts and slides is then processed through a knowledge extraction pipeline. Using natural language processing, the platform identifies key concepts, definitions, prerequisite relationships, and contextual connections between ideas. These concepts are organized into a knowledge graph and transformed into semantic embeddings, creating a structured knowledge base that supports retrieval, reasoning, and personalized content generation. From there, Promea continuously models learner understanding at the concept level rather than simply tracking overall course performance. For every concept, the system monitors mastery, confidence, review history, hint usage, engagement, and time spent learning. This allows Promea to identify both knowledge gaps and mismatches between perceived and actual understanding, enabling more targeted support. At the center of the system is KAPN, our Knowledge-Augmented Policy Network. By combining Bayesian Knowledge Tracing, Dynamic Bayesian Networks, Deep Knowledge Tracing, Retrieval-Augmented Generation, and Reinforcement Learning, KAPN estimates a learner's knowledge state, predicts future learning difficulties, and selects the instructional intervention most likely to improve learning outcomes. Depending on the learner's needs, the system may generate simplified notes, visual explanations, adaptive quizzes, targeted review recommendations, or increased challenge when mastery has been achieved. Every intervention is logged and evaluated through a continuous feedback loop. The platform measures changes in mastery and confidence before and after each intervention, allowing it to learn which strategies are most effective for different learners. Promea also maintains a forgetting-risk model inspired by memory decay research, proactively recommending review when concepts are at risk of being forgotten. Over time, this creates a closed-loop adaptive system that continuously learns from each student and becomes increasingly personalized as more learning interactions occur. Rather than providing static accommodations, Promea is designed to evolve alongside each learner, delivering increasingly personalized support as it develops a deeper understanding of how they learn best.

Challenges we ran into One of the biggest challenges was understanding the true complexity of dyslexia. Initially, we approached dyslexia primarily as a reading challenge. However, our literature review quickly revealed that dyslexia encompasses a wide range of cognitive difficulties that vary dramatically between individuals. This raised an important design challenge: how can an educational system personalize support when every learner's needs are different? Another challenge involved translating educational theory into computational systems. Concepts such as cognitive load, working memory, learner autonomy, motivation, and instructional pacing are easy to discuss theoretically but significantly more difficult to model mathematically. We spent considerable time researching Bayesian learning models, reinforcement learning systems, and student modeling frameworks to identify methods capable of representing evolving learner states. Balancing personalization and transparency also proved challenging. Many AI systems function as black boxes, yet educational environments require accountability and explainability. We therefore designed KAPN to maintain interpretable learner profiles and ensure that recommendations remain understandable to educators and students. Finally, designing a wearable form factor introduced practical challenges related to comfort, privacy, classroom integration, and social acceptance. Research consistently showed that assistive technologies are often abandoned when they create stigma or unnecessary complexity, so user-centered design became a major focus throughout development. Accomplishments that we're proud of Our greatest accomplishment is turning something deeply personal into something meaningful. What started as conversations with friends who have dyslexia grew into a research-driven effort to better understand the challenges they face and how technology could provide more effective support. Throughout this project, we constantly challenged ourselves to avoid making assumptions. Instead, we grounded our decisions in both research and real experiences. Those insights led us to design KAPN, an adaptive learning system that recognizes that no two dyslexic learners are the same and that meaningful support must evolve alongside the learner. We are also proud that Promea addresses a community that is often overlooked. While many existing tools focus on basic reading assistance, we wanted to explore how AI could support students navigating complex, lecture-heavy learning environments where information moves quickly and personalized support is often limited. More than any technical achievement, we are proud that WeAreAbility became something bigger than a competition project. Throughout this journey, we had the opportunity to connect with friends, mentors, researchers, educators, and advocates who shared their experiences, challenged our assumptions, and helped shape our understanding of what meaningful accessibility looks like. Their voices remained at the center of every decision we made, and their support ultimately helped transform WeAreAbility from an idea into a vision we genuinely believe can make a difference.

What we learned Throughout this project, we learned that building meaningful solutions requires far more than a good idea. It requires patience, research, collaboration, and a willingness to constantly challenge your own assumptions. As we spoke with students, educators, researchers, healthcare professionals, and accessibility advocates, we saw how complex the problem really is. Dyslexia is not a single experience, and there is no universal solution. Every conversation reinforced the importance of designing systems that adapt to the learner rather than expecting the learner to adapt to the system. We also learned the value of bringing together ideas from different fields. The challenges we were trying to address could not be solved through technology alone. Building Promea required us to connect insights from cognitive science, educational psychology, accessibility research, machine learning, and human-centered design into a single framework. Most importantly, this project showed us how powerful it can be to simply ask questions and listen. Some of our most valuable insights did not come from research papers or technical discussions. They came from conversations with friends, mentors, and members of the dyslexic community who were willing to share their experiences with us. Those conversations shaped not only what we built, but how we think about innovation itself. WeAreAbility taught us that the strongest solutions are not created in isolation. They are built through curiosity, collaboration, and the collective wisdom of the people who inspire them.

What's next for WeAreAbility Our next step is turning Promea from an idea into something students can actually use. We hope to fully implement the KAPN architecture, develop the wearable smart-glasses system, and create a seamless platform capable of capturing lectures, understanding what is being taught, and transforming that information into personalized learning support. What is currently a vision on paper would become a real tool that can sit in a classroom, listen alongside a student, and help them engage with material in a way that works best for them. As development continues, we hope to work closely with students, educators, researchers, and accessibility advocates to ensure Promea remains grounded in the needs of the people it is designed to serve. Their experiences and feedback will help shape every stage of the platform's evolution. Long term, we envision Promea becoming more than just a device. We hope it becomes a companion that helps students with dyslexia feel more confident raising their hand in class, less overwhelmed by fast-paced lectures, and more empowered to pursue the subjects they love. We want students to spend less time struggling to access information and more time discovering what they are capable of achieving. Ultimately, our goal is not simply to build new technology. It is to help create a future where students are not limited by the way information is presented to them. If Promea can help even one student feel more understood, more confident, and more excited to learn, then every hour spent imagining and building it will have been worth it.