# An Evaluation of Explanation Methods for Black-Box Detectors of Machine-Generated Text

arXiv: https://arxiv.org/abs/2408.14252 · HTML: https://mlsanity.com/paper/2408.14252/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 21 of its 42 references that are in the map.
- First posted on arXiv 26 Aug 2024.

## Builds on (most cited first, 21 of 21)

- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [A Unified Approach to Interpreting Model Predictions](https://mlsanity.com/paper/1705.07874/) (2017)
- [DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter](https://mlsanity.com/paper/1910.01108/) (2019)
- [OPT: Open Pre-trained Transformer Language Models](https://mlsanity.com/paper/2205.01068/) (2022)
- [Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling](https://mlsanity.com/paper/2304.01373/) (2023)
- [Towards A Rigorous Science of Interpretable Machine Learning](https://mlsanity.com/paper/1702.08608/) (2017)
- [A Survey Of Methods For Explaining Black Box Models](https://mlsanity.com/paper/1802.01933/) (2018)
- [DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvature](https://mlsanity.com/paper/2301.11305/) (2023)
- [Towards Faithfully Interpretable NLP Systems: How should we define and evaluate faithfulness?](https://mlsanity.com/paper/2004.03685/) (2020)
- [ELI5: Long Form Question Answering](https://mlsanity.com/paper/1907.09190/) (2019)
- [Release Strategies and the Social Impacts of Language Models](https://mlsanity.com/paper/1908.09203/) (2019)
- [Towards Robust Interpretability with Self-Explaining Neural Networks](https://mlsanity.com/paper/1806.07538/) (2018)
- [How Close is ChatGPT to Human Experts? Comparison Corpus, Evaluation, and Detection](https://mlsanity.com/paper/2301.07597/) (2023)
- [From Anecdotal Evidence to Quantitative Evaluation Methods: A Systematic Review on Evaluating Explainable AI](https://mlsanity.com/paper/2201.08164/) (2022)
- [Metrics for Explainable AI: Challenges and Prospects](https://mlsanity.com/paper/1812.04608/) (2018)
- [Evaluating Explainable AI: Which Algorithmic Explanations Help Users Predict Model Behavior?](https://mlsanity.com/paper/2005.01831/) (2020)
- [Robust and Stable Black Box Explanations](https://mlsanity.com/paper/2011.06169/) (2020)
- [CHEAT: A Large-scale Dataset for Detecting ChatGPT-writtEn AbsTracts](https://mlsanity.com/paper/2304.12008/) (2023)
- [Explaining Predictions of Non-Linear Classifiers in NLP](https://mlsanity.com/paper/1606.07298/) (2016)
- [Smaller Language Models are Better Black-box Machine-Generated Text Detectors](https://mlsanity.com/paper/2305.09859/) (2023)

## Built on by (most cited first, 2 of 2)

- [Do LLM Self-Explanations Help Users Predict Model Behavior? Evaluating Counterfactual Simulatability with Pragmatic Perturbations](https://mlsanity.com/paper/2601.03775/) (2026)
- [EMMM, Explain Me My Model! Explainable Machine Generated Text Detection in Dialogues](https://mlsanity.com/paper/2508.18715/) (2025)

## Abstract (arXiv)

The increasing difficulty to distinguish language-model-generated from human-written text has led to the development of detectors of machine-generated text (MGT). However, in many contexts, a black-box prediction is not sufficient, it is equally important to know on what grounds a detector made that prediction. Explanation methods that estimate feature importance promise to provide indications of which parts of an input are used by classifiers for prediction. However, these are typically evaluated with simple classifiers and tasks that are intuitive to humans. To assess their suitability beyond these contexts, this study conducts the first systematic evaluation of explanation quality for detectors of MGT. The dimensions of faithfulness and stability are evaluated with five automated experiments, and usefulness is assessed in a user study. We use a dataset of ChatGPT-generated and human-written documents, and pair predictions of three existing language-model-based detectors with the corresponding SHAP, LIME, and Anchor explanations. We find that SHAP performs best in terms of faithfulness, stability, and in helping users to predict the detector's behavior. In contrast, LIME, perceived as most useful by users, scores the worst in terms of user performance at predicting detector behavior.