WriteWiseAI by CommonLit

WriteWiseAI by CommonLit employs advanced machine learning (ML) and natural language processing (NLP) technologies to provide immediate, precise feedback on student writing, distinguishing itself from typical chatbots by integrating direct instruction with feedback. Unlike other tools that offer feedback in isolation, WriteWiseAI ensures that each piece of feedback is directly tied to instruction, enhancing relevance and impact. This integration supports a holistic learning approach, helping students not only understand their errors but also learn the underlying principles to improve their writing skills. It also reduces the administrative burden for teachers looking to assess students’ skills, plan and teach an impactful writing lesson, and provide immediate feedback to a classroom of students. The system is designed to train against robust writing rubrics, integrating teacher oversight as the final check to ensure personalized and contextually relevant feedback. This dual focus on feedback efficiency and instructional depth makes WriteWiseAI a helpful tool for educators and students alike, making writing instruction more connected, immediate, and effective.

Technologies used include:

• Adaptive Learning Algorithms: Utilizes adaptive learning algorithms to dynamically adjust the difficulty of writing tasks based on real-time student performance data, ensuring personalized learning experiences.
• Natural Language Processing (NLP): At the heart of WriteWiseAI is a sophisticated NLP engine that analyzes text, providing feedback on grammar, style, and coherence. This technology enables the solution to understand and interact with human language in a nuanced and educative manner.
• Predictive Analytics: By incorporating predictive analytics, WriteWiseAI forecasts students’ writing proficiency developments, allowing educators to proactively tailor instruction to meet anticipated needs and skill gaps.
• Automated Rubric Scoring: Employing ML models trained on diverse datasets, WriteWiseAI automates scoring against standardized rubrics, ensuring consistent and unbiased evaluations of student writing.
• Feedback Loop Integration: WriteWiseAI features a robust feedback loop system that adjusts to student responses and evolves from teacher inputs, enhancing its decision-making algorithms to support educational outcomes better.
• AI-Driven Personalization: The platform’s AI-driven personalization engine analyzes individual learning patterns and tailors writing prompts and feedback accordingly, promoting optimal learning efficiency and engagement.
• Sentiment Analysis: Utilizing sentiment analysis, WriteWiseAI evaluates the tone and style of student writing, providing insights that help teachers understand student engagement and adapt their instructional strategies.

Primarily targeting grades 6-12, WriteWiseAI alleviates significant burdens from educators by automating the labor-intensive process of writing feedback. This technology is crucial in resource-strained educational environments, promising to elevate the quality of education by allowing teachers to focus more on personalized instruction rather than time-consuming assessments. Additionally, middle school is a critical period for writing development, making WriteWiseAI particularly beneficial in fostering essential writing skills during these formative years.

Built by former educators, CommonLit is on a mission to provide educators without access to high-quality learning materials with the tools to teach the next generation the reading, writing, critical thinking, and communication skills needed to succeed in school, regardless of socioeconomic background. The foundation of our work is our proximity to and collaboration with the communities we serve.

The team leading this initiative uniquely represents and is connected to our target communities. Many team members, including our CEO, Michelle Brown, have personal or professional backgrounds in these communities. This connection fosters a genuine understanding of the specific challenges and needs faced by the educators and students we aim to support. They bring firsthand insights into the educational gaps and the technological solutions that can address them effectively.

Our dedicated engineering team, which has been pivotal in developing comprehensive educational tools that integrate curriculum, assessments, and professional development, is now expanding its capabilities into data science and AI. While AI is a new avenue for us, our team's strong foundation in handling large datasets, training models, and collaborating with researchers positions us well for this venture. We have a proven track record of success, evidenced by our creation and widespread publication of the CLEAR corpus—an open-source dataset developed by 2,000 teachers to assess text complexity, specifically designed to train AI in recognizing and generating accessible educational content.

The design and implementation of our AI tool are deeply guided by the communities' input. The insights gained from our extensive network of educators have been instrumental in shaping our approach. These educators directly interact with our target student population and provide continuous feedback that drives our innovation. Their contributions ensure that our solutions are theoretically sound, practically applicable, and tailored to real-world educational needs.

WriteWiseAI is currently in its development phase with the concept having been scoped and identified as a solution.

At large, CommonLit’s digital platform attracts ~7M student and teacher users each school year, with over 35M students and 1M teachers registered. In terms of impact, CommonLit has helped more than 5M students move towards proficiency since our program began in 2013 (1.5M children from the last school year alone), demonstrating a significant stride towards our objective and underscoring our commitment to moving the needle for children’s literacy on a substantial scale. 

WriteWiseAI revolutionizes writing education by seamlessly integrating real-time, actionable feedback with direct instructional content. This approach not only enhances student learning by providing specific, contextual feedback but also actively teaches writing principles during the feedback process, a significant advancement over existing technologies that offer feedback without instructional context.

• Innovative Approach:

  • WriteWiseAI leverages sophisticated natural language processing (NLP) and machine learning (ML) technologies. Unlike traditional tools that merely correct errors, WriteWiseAI’s NLP capabilities allow it to understand the context and intent behind students' text, making its feedback more relevant and instructive. This helps students recognize their mistakes and understand why they are mistakes and how to correct them in the future. Additionally, the ML algorithms adapt to each student's learning progress, tailoring feedback and instruction to suit individual needs, thus promoting a personalized learning experience.

• Broader Positive Impacts:

  • WriteWiseAI serves as a catalyst for broader positive changes within the educational technology space:

    • Integrative Educational Tools: By demonstrating the successful integration of AI in enhancing the educational process, WriteWiseAI encourages the development of new tools that can bridge gaps between learning management systems and actual teaching. This integrative approach can lead to more holistic educational solutions that consider feedback and instruction as interconnected elements of the learning process.

    • Promotion of AI in Education: The effectiveness of WriteWiseAI in improving educational outcomes showcases the potential of AI not just for administrative tasks but as a core component of educational methodologies. This can inspire educational institutions to adopt AI technologies more widely, raising the overall quality of education.

• Market/Landscape Changes:

  • WriteWiseAI is poised to transform the educational landscape by changing the way writing skills are taught:

    • Emphasis on Instructional Feedback: The market is saturated with quantitative feedback tools. WriteWiseAI shifts the focus to qualitative instructional feedback, setting a new standard for educational tools that value teaching as much as assessing.

    • Enhanced Teacher Efficiency: By automating the feedback process, WriteWiseAI allows educators to devote more time to interactive teaching and personalized instruction, potentially redefining the educator’s role from evaluator to facilitator.

In summary, WriteWiseAI introduces a novel, impactful solution that addresses the immediate needs within its application area and sets the stage for a fundamental transformation in educational practices and AI’s role in education. This innovative approach promises a more integrated, effective, and student-centered learning environment, setting a benchmark for future educational technologies.

• Natural Language Processing (NLP):

  • NLP is at the heart of WriteWiseAI. We use several NLP techniques, including syntax analysis, sentiment analysis, and entity recognition, to thoroughly analyze text submissions. This allows our system to provide feedback that goes beyond surface errors, addressing complex issues such as coherence, argument strength, and use of language, which are critical for developing writing skills at the middle school level.

• Machine Learning (ML):

  • Our ML models are trained on a large corpus of educational data, which includes anonymized student writings and feedback histories. These models are designed to learn from each interaction, thereby improving their accuracy and the relevance of feedback over time. We employ supervised learning to train our models against established educational rubrics, ensuring that the feedback provided aligns with educational standards and learning goals.

• Technology Integration and Development:

  • Model Training and Implementation:

    • WriteWiseAI’s AI models are primarily developed in-house. We rely on open-source libraries such as TensorFlow and PyTorch to build and train our models. However, the intellectual property—the algorithms and the data models—are proprietary and tailored specifically for educational purposes. This self-sufficiency in model development allows for greater control over features and customization based on user feedback and educational trends.

  • Third-Party Models and APIs:

    • For certain specific functions, such as syntax error detection, we integrate third-party APIs that specialize in language mechanics. These APIs complement our in-house technologies by enhancing the robustness of the grammar checks without burdening our core ML models with tasks already well-served by existing technologies.

  • Frontend User Interface:

  • The front end of WriteWiseAI is built using modern web development technologies such as React.js and Redux for state management. This setup ensures a responsive, intuitive user experience across various devices and platforms, which is crucial for accessibility in educational environments. The interface is designed to be user-friendly, allowing both students and educators to interact easily with the AI feedback and instructional materials.

• Ethical AI Use and Adaptation:

  • We are committed to ethical AI practices, ensuring our models are transparent and fair. Regular audits and updates are part of our operational model to mitigate any potential biases and ensure that our technology remains inclusive and effective for all users.

CommonLit has been shown to deliver 2x learning acceleration with gains replicated in multiple studies, each showing significant improvement for all students, especially for low-income students, English language learners, and students with disabilities. It is one of a handful of edtech tools in the market that have shown the rigor and efficacy to recently receive a Tier II rating under the Every Student Succeeds Act (ESSA). This rating is indicative of our commitment to evidence-based solutions and provides a strong foundation for the technology we are developing.

CommonLit’s premiere program, CommonLit360 (“CL360”), is a comprehensive program that recently demonstrated success in a large-scale study of over 100k students, conducted through a Gates-funded study by third-party evaluators from Mathematica. The study showed reading growth for students of teachers with high utilization of CL360 was statistically significantly higher than that of students in the comparison group. Students of high-utilizing teachers grew +0.22 standard deviations or 2-6 months more reading gains than we typically expect in 1 year compared to low-utilizing teachers. Significant positive effects were also shown in the high utilization group in the areas of affirming cultural identity, behavior management, and student engagement. Roughly 9 in 10 teachers surveyed indicated that the CL360 curriculum helps students improve their reading (98%), writing (89%), and discussion skills (91%). Both high and low-utilizing teachers agreed that CL360 improved their teaching practice. 

These strong outcomes led CommonLit to be selected for a high-profile Education Innovation and Research grant from the Department of Education. Through this research grant, CL360 will be studied in a large-scale longitudinal 5-year RCT evaluation in 100 schools in partnership with the Teacher College at Columbia University. Our underlying pedagogical approach also recently received the green lighting from EdReports, indicating the ready-for-scale effectiveness of our program.

Helpful Links:

• Summary of Efficacy Studies (Jan 2024) - a compilation of major studies conducted by CommonLit alongside research partners such as LearnPlatform, Consortium for Policy Research in Education (CPRE), and Mathematica  

• EdTech Evidence: 2023 Mid-Year Report from LearnPlatform - this report looks at the evidence available for the 100 learner-focused and educator-focused K-12 solutions accessed by most students and educators during the first half of the 2022-23 school year, highlighting the lack of efficacy of most edtech tools in the market

• ESSA Tier II Study Blog Post - published blog post with details on the study conducted by LearnPlatform to receive an ESSA Tier II rating

• Impact Page - includes highlights of our research studies

• Insights Blog Page - includes blog posts that showcase our ongoing research and subsequent results and findings

We are committed to ensuring equity by designing AI systems that enhance educational outcomes and actively combat biases that could impact marginalized groups, particularly Black and Latino learners and those experiencing poverty. Our approach to equity involves both the careful development of AI technologies and ongoing efforts to monitor and refine our solutions in response to diverse learner needs.

• Equity and Bias Mitigation Goals:

  • Diverse Data Sets:

    • To combat algorithmic bias, WriteWiseAI uses training data that is representative of the diverse student populations we serve, including substantial inputs from Transitional Kindergarten, Pre-K, and K-8 environments. We prioritize including data from Black and Latino communities and learners experiencing poverty to ensure our AI models perform equitably across all demographic groups.

  • Bias Detection and Correction:

    • We employ advanced statistical techniques to detect and correct biases in our AI models. This includes regular audits of the model outputs to identify any disparities in feedback that could reflect underlying biases. Adjustments are made through retraining sessions with curated, balanced datasets to mitigate detected biases.

  • Ethical AI Development:

    • Our development team adheres to ethical AI guidelines which emphasize fairness and transparency. We involve stakeholders from diverse backgrounds in the design and testing phases to gather a broad range of insights and experiences which guide our development process.

  • Equitable AI Design:

    • WriteWiseAI is designed with a focus on equitable user experiences. We ensure that our user interface and content are accessible to all students, including those with disabilities, and are sensitive to cultural and socioeconomic contexts. This is crucial for providing equitable access to learning opportunities, as defined by MIT’s equity standards.

  • Monitoring and Deployment Strategy:

    • Continuous Feedback Loop:

      • We implement a continuous feedback loop with educators and students from diverse backgrounds to understand the impact of our AI tool in real-world educational settings. This feedback is crucial for the ongoing refinement of our algorithms and helps ensure that our tool evolves in response to the actual needs of all users.

    • Collaboration with Educational Experts:

      • We collaborate with experts in educational equity to ensure that our AI tools do not perpetuate or exacerbate existing educational disparities. These experts help us identify potential barriers to full participation by marginalized groups and develop strategies to eliminate these barriers.

    • Transparency and Accountability:

      • WriteWiseAI maintains a high level of transparency regarding how our AI models are built, trained, and deployed. We provide documentation and reports on the performance of our models across different demographics, ensuring accountability and facilitating trust among users and stakeholders.

    • Regular Impact Assessments:

      • We will conduct regular impact assessments to measure how well WriteWiseAI is meeting its goals of equitable access to educational opportunities. These assessments help us to understand whether we are successfully addressing disparities and achieving meaningful advancements for all learners.

The WriteWiseAI team comprises of 9 FTEs: a product manager, a product designer, two data scientists, three curriculum writers, and two engineers. This dedicated group is focused on the rapid development and iteration of a market-ready MVP, ensuring a robust and user-friendly product.

WriteWiseAI is on track to be pilot-ready within the next year, with our comprehensive development timeline and preparation strategies aligned to meet the specific needs of priority learners, including public Transitional Kindergarten, Pre-K, and K-8 students, with a focus on Black and Latino learners and those experiencing poverty.

• Pilot Readiness Plan:
  • Development Timeline:
    • We are currently finalizing the alpha version of WriteWiseAI and will begin beta testing in three months. This testing phase involves a select group of educators from diverse backgrounds who will use the tool in classroom settings and provide feedback. The insights gained will guide final adjustments, ensuring the tool meets the diverse needs of our target student groups. Our schedule projects that WriteWiseAI will be ready for a larger pilot launch by Spring 2025.
• Collaborative Testing Framework:
  
  • Collaborative Testing Framework:
    • Our pilot preparation involves collaboration with schools in urban and rural areas to ensure our AI tool is adaptable and effective across various educational settings. These partnerships are crucial for understanding the specific challenges and needs of priority learners and for tailoring our solution to better fit these contexts.
• Evidence of Progress:
  • Evidence that we are on track includes:
    • Development Milestones: Meeting key software development milestones, with positive initial feedback from early tester educators who highlight the tool’s usability and impact on student engagement.
    • Engagement Metrics: Preliminary tests show that our tool significantly reduces the time teachers spend on grading and feedback, allowing more focus on direct instruction.
    • Advisory Board Feedback: Ongoing guidance from our educational technology advisory board, which includes experts in equity and pedagogy

As a 501(c)(3) nonprofit, CommonLit is on a mission to close the literacy opportunity gap and unlock the potential of children through a highly effective, super scalable digital English Language Arts program. Our platform, recognized for its high accessibility to underserved communities, is designed to be both user-friendly and effective in diverse educational settings. We leverage our nonprofit status to offer our services at approximately one-third the cost of our competitors, making our solutions more affordable for schools and districts that serve priority learners.

While there is uncertainty in scaling AI solutions, we are actively developing a robust model that allows us to incrementally increase our reach without significant cost increases, which includes sustaining WriteWiseAI through our paid services. This model will ensure that as our AI tool is scaled, it remains economically accessible to the schools that need it most.

CommonLit is applying to the Learner//Meets//Future Challenge because, despite our high-growth trajectory and mature infrastructure, integrating AI into our educational solutions presents significant financial and technical risks that are challenging to undertake alone. As a nonprofit dedicated to delivering high-quality educational materials at accessible prices, venturing into AI necessitates substantial upfront investment and expertise, which are currently beyond our conventional funding and technical capabilities.

Through this challenge, we believe that support from Solve and the Gates Foundation can help alleviate this financial barrier as well as provide needed guidance through both organizations’ technical expertise to develop WriteWiseAI.