HULLMI: Human vs LLM identification with explainability

TL;DR

This paper compares traditional ML models and modern NLP detectors for human vs AI text detection, emphasizing interpretability through LIME, and highlights the potential for reliable, explainable detection tools across various domains.

Contribution

It demonstrates that traditional ML models perform on par with modern NLP detectors in AI detection and introduces explainability techniques to improve trustworthiness.

Findings

Traditional ML models match modern NLP detectors in detection accuracy.

LIME provides interpretable insights into model predictions.

The approach enhances reliability of AI detection tools in critical domains.

Abstract

As LLMs become increasingly proficient at producing human-like responses, there has been a rise of academic and industrial pursuits dedicated to flagging a given piece of text as "human" or "AI". Most of these pursuits involve modern NLP detectors like T5-Sentinel and RoBERTa-Sentinel, without paying too much attention to issues of interpretability and explainability of these models. In our study, we provide a comprehensive analysis that shows that traditional ML models (Naive-Bayes,MLP, Random Forests, XGBoost) perform as well as modern NLP detectors, in human vs AI text detection. We achieve this by implementing a robust testing procedure on diverse datasets, including curated corpora and real-world samples. Subsequently, by employing the explainable AI technique LIME, we uncover parts of the input that contribute most to the prediction of each model, providing insights into the detection process. Our study contributes to the growing need for developing production-level LLM detection tools, which can leverage a wide range of traditional as well as modern NLP detectors we propose. Finally, the LIME techniques we demonstrate also have the potential to equip these detection tools with interpretability analysis features, making them more reliable and trustworthy in various domains like education, healthcare, and media.