Automated Essay Scoring

Automated essay scoring studies how to assign reliable scores to essays and written responses, often by matching model judgments to human ratings and rubric criteria. Approaches range from linguistic-feature models and fine-tuned encoders to prompted large language models, with methods such as trait-based scoring, ordinal prediction, retrieval of scored examples, and rubric-guided feedback. Researchers test performance on benchmarks such as ASAP and TOEFL essays, while probing transfer across prompts, languages, scoring traits, and assessment settings. A central challenge is that strong average agreement does not ensure valid, stable, or fair scores: models can vary with wording, rubric, institution, and writer background, so careful validation and human oversight remain important.