Date
Publisher
arXiv
The rapid advancement of large language models (LLMs) has enabled the
generation of coherent essays, making AI-assisted writing increasingly common
in educational and professional settings. Using large-scale empirical data, we
examine and benchmark the characteristics and quality of essays generated by
popular LLMs and discuss their implications for two key components of writing
assessments: automated scoring and academic integrity. Our findings highlight
limitations in existing automated scoring systems, such as e-rater, when
applied to essays generated or heavily influenced by AI, and identify areas for
improvement, including the development of new features to capture deeper
thinking and recalibrating feature weights. Despite growing concerns that the
increasing variety of LLMs may undermine the feasibility of detecting
AI-generated essays, our results show that detectors trained on essays
generated from one model can often identify texts from others with high
accuracy, suggesting that effective detection could remain manageable in
practice.
Study at a glance
- Question: How do AI-generated essays compare to human essays in quality, automated scoring, and detectability?
- Setting and data: 200 human essays from 2021 GRE tests and 2,000 AI-generated essays across two GRE prompts from 10 language models.
- Method: Model and benchmark evaluation comparing human ratings, e-rater scores, linguistic features, and machine learning classifiers (XGBoost)
Why it matters
Widespread use of AI writing tools raises concerns for academic integrity and challenges the ability of testing systems to evaluate writing accurately. Over-reliance on AI also risks diminishing individuals' writing proficiency and critical thinking skills.
Key findings
- The e-rater automated scoring engine systematically assigned higher scores to AI-generated essays than human raters did across all evaluated models.
- Human essays demonstrated stronger development but made more grammar, mechanics, and usage errors than AI-generated essays, while showing substantially higher perplexity.
- Classifiers accurately detected AI essays within the same model and generally maintained strong accuracy across models, though some cross-model combinations performed less effectively.
Summary drafted with AI from the full paper and may contain errors. See the full paper for details.
What is the application?
Who is the user?
Who age?
Why use AI?
Study design
