evaluation quality

A primary mathematics AI pilot reported before-and-after gains, with comparative effectiveness still unclear

Jingxi Liu, Daner Sun, Jin Sun, Jingyun Wang, Philip Leung Ho Yu

Computers and Education: Artificial Intelligence

A ChatGPT-supported word-problem system was studied in two fifth-grade classes totaling 104 pupils. The detailed performance analysis reported gains among 49 AI-group pupils after three post-task outliers were removed. Although the abstract describes a comparative advantage, the main results present a within-group paired analysis rather than a clear between-group effect. The paper offers a learning-system design and preliminary task evidence, while leaving stronger claims about comparative and lasting effectiveness unresolved.