StudyPal

Inspiration

We understand that traditional educational approaches don't always cater to the diverse needs of all learners, and this is especially true for those who are either fast or slow learners. Our platform aims to bridge this gap by offering adaptive pathways and differentiated content, ensuring that every child, regardless of their learning pace, feels included, supported, and challenged appropriately.

Furthermore, from personal experience, we understand that class sizes in primary school can go up to 40, making it incredibly challenging for teachers to provide individualised attention to every student. However, even with grouping strategies, it can still be difficult for teachers to consistently monitor each student's progress and adapt their teaching in real life. Our platform is designed to support this process and complement the existing curriculum by helping teachers identify students' weak areas more efficiently, so they can provide targeted support where it's needed most.

What it does

Study Pal is an adaptive learning platform that personalises education for students of all ages and regions. While our current demo focuses on Primary 1–3 learners, the platform's scalable architecture and pedagogy support a much wider educational scope.

Our approach is built on three pillars:

Skill-Based Topic Mapping Study Pal breaks content into fine-grained skills and maps topic dependencies. This ensures students build knowledge progressively—for instance, mastering arithmetic before advancing to algebra.
Differentiated Quizzes for Mastery Quizzes are tailored by difficulty and dynamically assigned based on student progress. For young learners, these take on playful formats. Struggling students are flagged for teacher support.
AI-Powered Learning Paths Our AI analyses performance trends to predict mastery, adapting content in real-time. This keeps students challenged without overwhelming them, optimising learning pace and retention. Together, these features enable Study Pal to deliver a personalised, engaging, and effective learning experience—anytime, anywhere.

How we built it

We've decided to initially concentrate our efforts on Primary 1-3 students in Singapore. This age group currently lacks adequate adaptive learning solutions (ALS), presenting a clear opportunity for impact. Our pilot program will focus on selected topics to test the solution's feasibility. If successful, we envision scaling the platform to different regions, with teachers playing a key role in curriculum planning and adaptation.

To personalize the learning experience, we're leveraging the Bayesian Knowledge Tracing (BKT) model. This well-known statistical model allows us to predict a student's mastery of specific skills, enabling the platform to dynamically adjust the pacing and difficulty of the online course. This ensures each child receives content tailored to their individual needs.

Recognizing the critical importance of foundational knowledge in early education, our platform's design prioritizes engagement for young children. It will be highly gamified and incorporate interactive kinesthetic elements to keep children actively involved and motivated in their learning journey. We leverage React to build a dynamic and responsive user interface, allowing for seamless integration of these interactive and kinesthetic components, ensuring a fluid and engaging experience for our young learners.

Challenges we ran into

Developing a Minimally Viable Product (MVP) within a short timeframe, especially one that incorporates Machine Learning (ML) algorithms for student pacing and robust data analytics, presents several significant challenges. We decided to come up with a framework that focuses on selected topics as a proof of concept.

Accomplishments that we're proud of

We've successfully developed a working BDK mastery predictor that's a cornerstone of our platform. This system effectively adjusts the difficulty and pacing of the course material, ensuring each student receives a learning experience tailored to their individual progress and comprehension.

What we learned

Developing an adaptive learning platform taught us the complexities of personalizing education. We realized the extensive data and rigorous planning required for effective syllabus design, especially when integrating adaptive algorithms.

What's next for StudyPal

With more contributions from different organizations, we can have a greater pool of questions grouped according to topics and difficulty level that is applicable to a diverse range of educational standards. This greater pool of questions also ensures that students get more practise and are exposed to more question types and ways of thinking.

Organizations can import existing dataset e.g their school’s markers report to be used as training data for our BKT model. Over time, the model can be enhanced with additional features like response time, allowing for even more personalised learning paths.

Features such as adjusting the level of difficulty of questions based on the proportion of students who get the question correct can also be implemented. This ensures that questions that students find hard will not be mistakenly labelled as easy or medium, allowing for better representation of how they are doing among their peers.

To prevent the accumulation of learning deficits, the AI can proactively reinforce previous topics whenever any gaps or weaknesses in students’ understanding is detected. Additionally, online lessons can also be added to supplement students' learning. They can be precursors to attempting the quizzes or optional resources that students can take if they feel they need help in that area. Teacherscan also recommend or prompt students to engage with specific lessons through their dashboard, ensuring targeted guidance.