Skip to main content Skip to navigation

A Toolbox for Modelling Engagement with Educational Videos

Project Overview

This document explores the integration of generative AI in education, specifically through the introduction of the PEEKC dataset and the TrueLearn Python library, both of which aim to enhance personalized learning. The PEEKC dataset comprises more than 20,000 examples of informal learners interacting with educational videos, particularly centered on AI and machine learning topics. TrueLearn employs Bayesian algorithms to effectively model learner engagement and provides visual tools that allow users to gain insights into their learning states. The library demonstrates superior predictive performance compared to existing baseline models, facilitating tailored educational experiences that cater to individual learner needs. Overall, the findings suggest that the application of generative AI tools like TrueLearn can significantly improve the personalization of education, making learning more effective and responsive to the unique styles and needs of learners.

Key Applications

TrueLearn Python library

Context: Educational video engagement modeling for informal learners

Implementation: Developed using Bayesian models to analyze learner interactions with educational video fragments

Outcomes: High predictive performance for learner engagement, offering personalized recommendations

Challenges: Limited interpretability of some models, user understanding of visualizations may vary

Implementation Barriers

Data Scarcity

The lack of publicly available datasets for predicting learner engagement constrains the growth of personalized AI education.

Proposed Solutions: Release the PEEKC dataset which includes engagement from informal learners with AI-related videos.

User Understanding

Not all visualizations may be equally understood by a wide variety of end-users, which can hinder the effectiveness of the learner models.

Proposed Solutions: Implement user studies to refine visualizations and ensure they are user-friendly.

Project Team

Yuxiang Qiu

Researcher

Karim Djemili

Researcher

Denis Elezi

Researcher

Aaneel Shalman

Researcher

María Pérez-Ortiz

Researcher

Emine Yilmaz

Researcher

John Shawe-Taylor

Researcher

Sahan Bulathwela

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Yuxiang Qiu, Karim Djemili, Denis Elezi, Aaneel Shalman, María Pérez-Ortiz, Emine Yilmaz, John Shawe-Taylor, Sahan Bulathwela

Source Publication: View Original PaperLink opens in a new window

Project Contact: Dr. Jianhua Yang

LLM Model Version: gpt-4o-mini-2024-07-18

Analysis Provider: Openai

Let us know you agree to cookies