AI for Education: Designing Conversational Teaching Agents
Abstract
From online instruction to personalized testing, there has been a push in recent years to improve education through technology. Continuing in this direction, the goal of this project is to build a natural-language conversational teachingagent (i.e., an educational chatbot). The proposed chatbot would augment in-class instruction by reviewing course material and helping students work through problems designed to enhance their understanding of key concepts. Theapproach would complement popular video-based instruction tools by providing increased interactivity. The chatbot woud also be designed with personalization in mind. In particular, drawing on modern machine learning techniques,including contextual bandit learning, the chatbot would adapt to individual learning styles. The performance of the chatbot would be evaluated through a series of online experiments, benchmarking the approach against existing state-of-the-art online teaching resources. I plan both to carry out basic research and to deploy a fully functioning educational platform that can be used by students and teachers. As the primary apFrom online instruction to personalized testing, there has been a push in recent years to improve education through technology. Continuing in this direction, the goal of this project is to build a naturallanguageconversational teaching agent (i.e., an educational chatbot). The proposed chatbot would augment in-class instruction by reviewing course material and helping students work through problems designed to enhance theirunderstanding of key concepts. The approach would complement popular video-based instruction tools by providing increased interactivity. The chatbot woud also be designed with personalization in mind. In particular, drawing onmodern machine learning techniques, including contextual bandit learning, the chatbot would adapt to individual learning styles. The performance of the chatbot would be evaluated through a series of online experiments, benchmarking the approach against existing state-of-the-art online teaching resources. I plan both to carry out basic research and to deploy a fully functioning educational platform that can be used by students and teachers. As the primary apFrom online instruction to personalized testing, there has been a push in recent years to improve education through technology. Continuing in this direction, the goal of this project is to build a natural-language conversational teaching agent (i.e., an educational chatbot). The proposed chatbot would augment in-class instruction by reviewing course material and helping students work through problems designed to enhance their understanding of key concepts. The approachwould complement popular video-based instruction tools by providing increased interactivity. The chatbot woud also be designed with personalization in mind. In particular, drawing on modern machine learning techniques, includingcontextual bandit learning, the chatbot would adapt to individual learning styles. The performance of the chatbot would be evaluated through a series of online experiments, benchmarking the approach against existing state-of-the-artonline teaching resources. I plan both to carry out basic research and to deploy a fully functioning educational platform that can be used by students and teachers. As the primary application, I propose focusing on teaching high-schoollevel. corpus of questions, dissecting each question into discrete steps, enumerating possible correct and incorrect answers to each step, and ?nally, tying the possible responses together in a graph structure that describes the conversational o w. I plan to take a computationaapproach to re?ne the chatbot, to identify conversational and problem structures that are optimized both for engagement and for learning. I propose to accomplish this goal in twostages. First, I plan to use crowdsourcing to generate multiple pieces of content for each concept to be taught, including variations on problems,
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 27, 2018
- Source ID
- N000141812763
Entities
People
- Sharad Goel
Organizations
- Office of Naval Research
- Stanford University
- United States Navy