Headline
Adaptive Student Learning
Inspiration
Adaptive testing has been an innovative approach in educational assessments, particularly seen in math placement tests. The idea behind creating a more tailored testing experience inspired the development of this adaptive program aimed at refining students' learning experiences.
What it does
This Google AI-powered adaptive testing program is designed to identify and target students' weak areas in their studies. By adapting to the individual's responses, the model ensures that students receive questions that match their proficiency levels. The goal is for students to focus on areas that need improvement without being held back by subjects they have already mastered.
How we built it
The adaptive AI model was developed using Google Cloud, leveraging past math placement tests as a foundational dataset. These historical tests helped train the AI to generate questions similar to those previously administered. The model functions by presenting questions one at a time, dynamically adjusting the difficulty based on the student's response. Correct answers prompt the AI to present more challenging questions, while incorrect responses result in easier questions. This system was structured using six different tests, each containing 20 questions—10 crafted by the development team and 10 generated by the AI. These tests were ranked by difficulty to create an adaptive learning path.
Challenges we ran into
One significant challenge was determining the relative difficulty of questions. Initially, we recognized that our judgments were subjective, introducing potential bias. To overcome this, we implemented testing phases where the model was deployed in real-world scenarios. We collected data on which questions were frequently answered correctly or incorrectly, enabling the AI to learn and refine its understanding of difficulty levels more objectively.
Accomplishments that we're proud of
We are proud of conceptualizing a tool that can genuinely assist students in enhancing their educational outcomes. By developing a model that targets weak areas, we created an efficient learning tool that addresses gaps in students' knowledge. We are also proud of overcoming the challenges of data collection and difficulty ranking, making the AI model more robust and effective.
What we learned
A major takeaway from this project is that even seemingly simple ideas can be complex to execute without meticulous planning. Developing this adaptive model required creative problem-solving, navigating the limitations of existing AI models, and gaining a deeper understanding of how human intelligence can be mirrored in a learning tool.
What's next for Adaptive Student Learning
The next step for adaptive student learning is to expand access and make the tool widely available. Integrating the program into classrooms so that teachers and professors can utilize it for post-lecture assessments would strengthen students' comprehension of core concepts. This expansion could enhance educational support and foster better learning outcomes on a broad scale.
Log in or sign up for Devpost to join the conversation.