AI-driven automated scoring systems offer scalable and efficient means of evaluating complex student-generated responses. Yet, despite increasing demand for transparency and interpretability, the field has yet to develop a widely accepted solution for interpretable automated scoring to be used in large-scale real-world assessments. This work takes a principled approach to address this challenge. We analyze the needs and potential benefits of interpretable automated scoring for various assessment stakeholder groups and develop four principles of interpretability – (F)aithfulness, (G)roundedness, (T)raceability, and (I)nterchangeability (FGTI) – targeted at those needs. To illustrate the feasibility of implementing these principles, we develop the AnalyticScore framework as a baseline reference framework. When applied to the domain of text-based constructed-response scoring, AnalyticScore outperforms many uninterpretable scoring methods in terms of scoring accuracy and is, on average, within 0.06 QWK of the uninterpretable SOTA across 10 items from the ASAP-SAS dataset. By comparing against human annotators conducting the same featurization task, we further demonstrate that the featurization behavior of AnalyticScore aligns well with that of humans.

LLM autograders are usually evaluated against human “ground truth,” a single reference label provided by one annotator or constructed by aggregating multiple annotations. However, human grading is often variable and subjective. We instead propose to evaluate LLM graders by treating humans and LLMs as two grading systems and characterize their agreement and disagreement. We operationalize this approach using distance correlation and generalizability theory. Across two programming-grading datasets, we find higher Human–LLM agreement on more objectively specified rubric items. On items involving greater subjective judgment, LLMs exhibit strong internal consensus, diverging more from human graders.