Skip to main content Skip to navigation

YouLeQD: Decoding the Cognitive Complexity of Questions and Engagement in Online Educational Videos from Learners' Perspectives

Project Overview

The document explores the role of generative AI in education, specifically highlighting the creation of the YouTube Learners' Questions on Bloom's Taxonomy Dataset (YouLeQD). It emphasizes the use of Large Language Models (LLMs) to analyze and generate educational questions that align with the cognitive complexity outlined in Bloom's Taxonomy. The primary objectives of the study are to deepen the understanding of questioning strategies in the educational context, boost student engagement, and support the advancement of effective AI models tailored for educational applications. The findings suggest that utilizing generative AI can significantly improve the quality of questions posed in educational settings, fostering a more interactive and engaging learning environment. Overall, the document underscores the potential of generative AI to enhance educational practices and outcomes through sophisticated question generation that aligns with cognitive learning frameworks.

Key Applications

YouLeQD: YouTube Learners' Questions Dataset

Context: Online educational videos, targeting learners engaging with STEM content on YouTube.

Implementation: Data collected from comments on 1,762 educational YouTube videos across five STEM subjects. Developed RoBERTa-based classification models to detect and analyze question complexity using Bloom's Taxonomy.

Outcomes: Generated a dataset of 57,242 questions, providing insights into cognitive complexity and engagement metrics, aiding in better AI model development for education.

Challenges: Challenges include the quality of extracted questions from informal comments, alignment with educational standards, and ensuring the effectiveness of generated questions.

Implementation Barriers

Quality of input data

The extracted questions from YouTube comments are often low-quality and irrelevant, making it difficult to classify them accurately according to Bloom's Taxonomy.

Proposed Solutions: Implementing a two-stage training process and using LLMs for data augmentation to improve the quality and relevance of the training dataset.

Alignment with instructional design

AI models sometimes lack a comprehensive grasp of instructional design principles, leading to ineffective questioning strategies.

Proposed Solutions: Bridging the gap between AI and instructional design to ensure AI-generated questions are thoughtfully crafted.

Project Team

Nong Ming

Researcher

Sachin Sharma

Researcher

Jiho Noh

Researcher

Contact Information

For information about the paper, please contact the authors.

Authors: Nong Ming, Sachin Sharma, Jiho Noh

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