The Empirical Effect Analysis of GAI-Assisted Multidimensional Evaluation in Primary School Composition

Confronting critical challenges in primary school composition pedagogy—including excessive teacher workload, imbalanced feedback mechanisms, and limitations in applying advanced technologies—this empirical study investigates the viability of generative artificial intelligence (GAI) as an evaluative aid. Through quantitative analysis of scoring consistency and qualitative examination of comment quality across 50 student essays assessed by both educators and GAI systems, three core findings emerge. First, GAI exhibits systematic scoring deviations from human evaluators, prioritizing surface-level linguistic accuracy over curriculum-aligned dimensions such as conceptual depth. Second, despite scoring limitations, GAI demonstrates robust auxiliary capacity in generating pedagogically structured comments, significantly reducing mechanical correction burdens while maintaining strategic alignment with teacher feedback. Third, grounded in cognitive development theory, a Teacher–AI Co-evolution Model is proposed to formalize collaborative roles: GAI handles normative diagnostics and initial feedback drafting, while teachers focus on higher-order guidance and motivational scaffolding. Results confirm that GAI cannot supplant teacher judgment but effectively enables a multidimensional evaluation paradigm—precision, strategy, and encouragement—thereby addressing structural inefficiencies in writing assessment. This synergy offers a pragmatic pathway to enhance pedagogical quality within resource-constrained educational contexts.