AI Teachable Agents for Knowledge Gap Detection and Reasoning Insight
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing online learning systems fail to effectively identify and address knowledge gaps in students, lack sufficient insight into student reasoning, and struggle to scale personalized and engaging learning experiences, leading to ineffective study strategies and discouragement.
Innovation Solution
A computer-implemented platform that utilizes AI-based teachable agents with recursive feedback, enabling students to teach concepts to an agent, observe its performance, and receive indirect feedback on their understanding, while generating and managing questions to enhance metacognition and engagement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional multiple choice tests are used for assessment, then grading efficiency is improved, but insight into student reasoning is lost
Solution Approach 1:
The patent introduces an AI character as an intermediary between the student and the assessment system. The student teaches the AI character, and the system evaluates the teaching quality to assess student understanding. This intermediary approach maintains automated grading efficiency while capturing detailed reasoning processes through the teaching interaction, thereby resolving the contradiction between grading speed and insight acquisition.
2Measurement precision
If one-on-one tutoring is implemented, then student understanding is improved, but scalability is reduced
Solution Approach 1:
The patent creates virtual AI character copies that can be distributed to multiple students simultaneously. Each student has their own AI character to teach, providing personalized one-on-one tutoring experiences at scale. The AI character serves as a copyable virtual tutor that maintains the benefits of individualized attention while enabling system-wide scalability, resolving the contradiction between understanding quality and scalability.
3Productivity
If students passively learn from chat bots, then learning coverage is improved, but student agency is reduced
Solution Approach 1:
The patent inverts the traditional teaching relationship by having the student teach the AI character instead of the AI teaching the student. This role reversal maintains high learning coverage through automated AI interactions while significantly boosting student agency, as students take an active instructional role rather than passively receiving information. This inversion resolves the contradiction between learning coverage and student agency.
4Reliability
If students focus on studying known material, then confidence is improved, but knowledge gap identification is worsened
Solution Approach 1:
The patent implements a feedback mechanism where the AI character's questions and responses provide students with information about their own understanding gaps. When the AI character asks clarifying questions or demonstrates confusion, it gives students feedback on what they haven't fully mastered, allowing them to maintain confidence in known areas while identifying and addressing knowledge gaps systematically.
Data Source
AI summary
In an approach to a learn by teaching platform, one or more processors display a graphic user interface that includes a first portion configured to enable a user to create a knowledge graph for a learning module. The one or more processors create the knowledge graph based on inputs received from the first portion of the graphic user interface. The one or more processors can select one or more questions for a teachable agent to answer using the created knowledge graph. In response to receiving answers for the one or more questions, the one or more processors generate a real-time visualization of thought processes of the teachable agent.


