Identifying AI-Written Text in Academia: A Machine Learning-Based Framework

The rise of AI-generated text poses new challenges to academic integrity, requiring robust mechanisms to identify and differentiate human-written content from machine-generated material in scholarly publications. This paper presents a machine learning-based framework for detecting AI-written text in academia, combining multiple techniques including stylometry, semantic analysis, and citation pattern recognition. The framework utilizes supervised classification models trained on features such as lexical diversity, syntactic complexity, and redundancy patterns to distinguish AI-generated content from genuine research writing. Additionally, unsupervised anomaly detection techniques are employed to flag unusual stylistic deviations. The framework also integrates traditional plagiarism detection tools and enlists human expert review for validating suspicious sections. By addressing both technical and ethical considerations, this approach aims to preserve the authenticity of academic work while adapting to the evolving landscape of AI-driven content generation.

• AI-generated text detection
• academic integrity
• stylometry
• semantic analysis
• lexical diversity
• syntactic complexity
• anomaly detection
• plagiarism detection

• Authorship Attribution and Profiling
• Academic integrity and plagiarism
• Spam and Phishing Detection
• AI-generated text detection
• academic integrity
• stylometry
• semantic analysis
• lexical diversity
• syntactic complexity
• anomaly detection
• plagiarism detection
• Artificial Intelligence
• Safety Research
• Information Systems
• Computer Science
• Social Sciences