AI Text Detection via Human Verification of Key Terms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing difficulty in distinguishing between AI-generated and human-written content, particularly in educational contexts, leads to inaccurate assessments and undermines the educational process, as existing methods often rely solely on machine-based detection that can become obsolete and exclude human interaction.
Innovation Solution
A method that involves scoring word rarity and vocabulary levels to select key words and phrases from a submitted text, followed by a test to measure the author's understanding of these terms, ensuring human involvement in the detection process and reducing the reliance on solely machine-based methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If purely machine-based methods are used to detect AI-written material, then detection efficiency and speed are improved, but reliability and accuracy deteriorate due to rapid AI evolution and false positives/negatives
Solution Approach 1:
The patent introduces an intermediary human verification step between the text analysis and final detection conclusion. The system identifies suspicious text through machine analysis, then selects key words/phrases for human author understanding tests, combining automated efficiency with human judgment reliability to resolve the contradiction between detection speed and accuracy.
2Extent of automation
If purely machine-based detection methods are used, then automation extent is improved, but adaptability deteriorates as methods become obsolete rapidly against evolving AI composition capabilities
Solution Approach 1:
The patent implements a dynamic detection system where the human involvement level adjusts based on the suspiciousness score from machine analysis. When AI composition methods evolve, the system adapts by updating its analysis parameters and selecting different key words for human verification, maintaining effectiveness without complete method overhaul.
3Reliability
If human involvement is increased in the detection process, then reliability and educational value are improved, but device complexity and time consumption increase
Solution Approach 1:
The patent segments the detection process into distinct phases: initial machine analysis to identify suspicious text, automated selection of key words/phrases based on analysis results, and targeted human verification only for selected elements. This segmentation reduces overall system complexity while maintaining high reliability through focused human involvement.
4Reliability
If human involvement is increased in the detection process, then educational interaction value is improved, but time consumption increases
Solution Approach 1:
The patent applies partial human action by having humans verify only the key words and phrases identified as most suspicious by the machine analysis, rather than reviewing the entire text. This partial verification approach maintains educational interaction value and assessment integrity while significantly reducing the time investment required from human reviewers.
Data Source
AI summary
This invention is a method for detecting AI-written material which evaluates authorship more directly than purely machine-based methods of AI detection. This method automatically selects relatively-rare key words and/or phrases used in a submitted body of text which is asserted to have been written by a specific person and creates a test to measure understanding of those key words and/or phrases which is administered to that person.