Artificial Intelligence for Educational Assessment: Prediction, Validity, and Responsible Use

Abstract

Ai Educational Assessment poses a recurring problem of coordinating performance with evidence quality, resource limits, and transfer across settings. The discussion is organized around placing performance prediction inside a broader validity, fairness, and intervention framework. The comparison integrates 1 focal paper with 13 independently retrieved publications whose authorship and venue fields were independently checked. The analysis is organized around construct validity, data drift, fairness, explainability, and teacher oversight. Rather than treating headline results as if they shared one denominator, the review compares study questions, technical premises, and validation scope. Across the literature, the recurring conclusion is that advances in AI educational assessment become credible when behavior, measurement, and decision context are evaluated together and when uncertainty about selection effects is reported explicitly. This organization relates method selection to decision risk and exposes recurring transfer threats, and proposes a research agenda centered on transparent baselines, stress testing, and reproducible evidence.