LLM Authorship Verification Through Assignment Comprehension Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI detection systems fail to effectively verify student comprehension of their assignments, allowing students to plagiarize by using AI to complete their work without demonstrating understanding.
Innovation Solution
A comprehension-based authorship verification system that uses Large Language Models to generate knowledge-testing questions and evaluate student responses to determine their understanding of the assignment content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional AI detection systems are used to analyze assignment text, then detection capability is provided, but they fail to verify student comprehension of the assignment content
Solution Approach 1:
The patent introduces an intermediary comprehension verification system that bridges the gap between AI detection and student understanding. The system generates questions based on the assignment content and evaluates student responses to verify comprehension, acting as a mediator between the detection system and the student's knowledge state.
Solution Approach 2:
The system implements feedback by analyzing student responses to comprehension questions and providing evaluation results. This feedback loop allows the system to assess whether students truly understand the assignment content they submitted, complementing the initial AI detection of potential plagiarism.
2Productivity
If students use AI to generate assignments, then assignment completion efficiency is improved, but academic integrity deteriorates due to plagiarism
Solution Approach 1:
The system performs preliminary action by detecting AI-generated content in assignments before final submission is confirmed. By identifying potential plagiarism early, the system can then apply additional comprehension verification to determine whether the student actually understands the material, preventing undetected academic misconduct.
Solution Approach 2:
The patent applies preliminary anti-action by implementing comprehension verification as a countermeasure against AI plagiarism. The system proactively generates questions and evaluates student understanding to counterbalance the harmful effect of students using AI to complete assignments without learning.
3Measurement precision
If comprehension questions are generated and student responses are evaluated, then authorship verification accuracy is improved, but system complexity increases
Solution Approach 1:
The system applies universality by using a multi-functional AI processing system that can both generate comprehension questions and evaluate student responses. This single system performs multiple functions (question generation, response analysis, comprehension assessment) that would otherwise require separate components, reducing overall system complexity while maintaining verification accuracy.
Solution Approach 2:
The patent merges the question generation and response evaluation functions into a unified comprehension verification process. By combining these functions that work with the same assignment content and student interaction, the system reduces complexity compared to having entirely separate systems for each function.
Data Source
AI summary
The method and system for comprehension-based authorship verification evaluates a user's response to determine how well the user comprehends their own submitted assignment. The user submits their assignment into a validation system where it is analyzed by a Large Language Model (LLM) to generate a question based on the assignment. The generated question is then presented to the user wherein the user provides a user response with a microphone, camera or keyboard. The user response is then analyzed by the LLM again to generate a score that ranks the level of comprehension the user has based on their submitted assignment. Each user response, question, score, and assignment are stored within a database and visually presented to an administrator, wherein the database highlights users that require more attention due to low scores.