AI Authorship Detection via Text Redundancy Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to accurately distinguish between human-authored texts and those generated by generative artificial intelligence (GenAI) conversational agents, particularly in cases where AI assistance is used in human writing.
Innovation Solution
A computer-implemented method that combines stylometric features and redundancy analysis to differentiate between human and GenAI-generated texts, utilizing Bayes factors to evaluate the evidence and determine the likelihood of human or AI authorship.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If stylometric features are used to detect AI-generated texts, then detection capability is improved, but accuracy deteriorates because GenAI texts can be instructed to write by applying specially defined types of style
Solution Approach 1:
The patent segments the text analysis into multiple independent stylometric features (word length, sentence length, paragraph length, N-grams, etc.) and evaluates each feature's contribution to detection accuracy. This allows the system to identify patterns that remain consistent across different stylistic instructions while filtering out style-specific variations.
Solution Approach 2:
The patent changes the parameters used for detection from simple stylistic features to a comprehensive set including low-level stylometric features (N-grams), semantic features, and syntactic features. By transforming the detection approach to analyze multiple parameter dimensions simultaneously, the system maintains accuracy even when AI texts are instructed to adopt specific human-like styles.
2Ease of operation
If GenAI agents are instructed to write by applying specially defined types of style, then the text becomes more human-like, but the specificity of the style becomes non-existent or difficult to discriminate
Solution Approach 1:
The patent introduces an intermediary analysis layer that examines the underlying structural patterns of text generation. By analyzing the probabilistic relationships between words, sentences, and paragraphs at multiple levels, the system can detect AI-generated content even when surface-level stylistic features have been customized to resemble human writing patterns.
3Stability of the object's composition
If a human writer maintains personal style along the text, then style consistency is improved, but the ability to detect AI assistance deteriorates
Solution Approach 1:
The patent dynamically analyzes style consistency across different segments of the text, comparing local stylometric patterns against global patterns and against known author profiles. This dynamic approach can detect AI assistance even when a human writer maintains personal style, by identifying inconsistencies in the probabilistic structure of the text that are characteristic of AI generation.
Data Source
AI summary
A computer implemented method for determining if a questioned text has been produced by a human or by a generative artificial intelligence based conversational agent. The method includes the steps of: retrieving from the test text a feature (y) representing the redundancy of the test text; using this feature for determining if the test text has been produced by a human or by a generative artificial intelligence based conversational agent.