PsyScore: A Psychometrically-Aware Framework for Trait-Adaptive Essay Scoring and ZPD-Scaffolded Feedback

PsyScore is a new psychometrically-aware framework for Automated Essay Scoring (AES) that integrates scoring and instructional feedback through a shared latent ability representation. It features a Trait-Adaptive Neural IRT Scorer utilizing the Graded Partial Credit Model (GPCM) for precise student ability estimation, a ZPD-Scaffolded Feedback Generator that adapts feedback based on diagnosed proficiency levels, and a Multi-Perspective Feedback Evaluation Strategy for assessing feedback quality. Experiments on the ASAP++ dataset show that PsyScore not only achieves competitive scoring performance but also offers feedback that is more aligned with pedagogical needs, making it significant for practitioners seeking to enhance the effectiveness of LLM-based educational tools.