← العودة إلى أخبار البحث
Chinese secondary students complete homework with digital assistance before taking a separate closed-book assessment observed by a teacher
ورقة مجلةPeer-reviewed study20266 أغسطس 2026· 2 min

Generative AI adoption was linked to higher homework scores but lower unaided exams in a 26,811-student panel

David Strömberg, Victor Lei, Yanhui Wu

CEPR Discussion Paper No. 21577

ملخص 500 كلمة

Chinese secondary students complete homework with digital assistance before taking a separate closed-book assessment observed by a teacher

Strömberg, Lei and Wu examine a pattern that homework dashboards alone can conceal: assisted work may improve while independent performance declines. Their CEPR discussion paper combines 30 months of records for 26,811 Chinese students in Grades 7 through 12 across nine subjects. The data include homework scores and completion time, monthly closed-book examinations and two types of high-stakes entrance examination. Because students and schools adopted generative AI at different times, the authors use a staggered-adoption difference-in-differences design to estimate changes after adoption.

The reported contrast is large. After adoption, homework scores rose by an estimated 18% while homework completion time fell by 30%. Within six months, monthly unaided examination scores were estimated to be 20% lower. Scores on the two high-stakes entrance examinations were estimated to fall by 18% and 24%, with the full entrance-exam penalty appearing after about two years. The paper reports the largest losses in social-science subjects, followed by STEM and languages, and larger estimated effects for younger students, high achievers and boys.

The authors investigate whether use resembles learning support or task outsourcing. About 80% of AI users displayed a combination of unusually high homework scores and unusually short completion time, which the paper interprets as outsourcing-like behaviour. The learning losses concentrate in that group. Users who maintained homework time closer to non-users showed much smaller losses. This pattern is consistent with reduced cognitive engagement, but it is not a direct observation of prompts, copied answers or student intention.

The evidence is consequential but not experimental. AI adoption was not randomly assigned, so the causal interpretation depends on parallel-trends assumptions and on whether other time-varying changes are adequately handled. The outsourcing measure is inferred from administrative traces rather than chat logs. The setting combines one national examination culture and particular homework platforms, and the paper is a discussion paper that may be revised. Its estimates should not be generalized to every learner, subject or guarded tutoring system.

For schools, the study argues for separating product completion from learning evidence. An AI-supported homework task can be paired with a brief unaided explanation, oral defence, retrieval quiz or transfer problem. Teachers can inspect both time-on-task patterns and the reasoning shown, without assuming that fast work is automatically misconduct. Assignments can permit AI for hints, examples and feedback while requiring students to identify what they changed and why.

For Hong Kong education, the strongest lesson is methodological: evaluate the outcome after assistance is removed. Higher homework scores and faster completion may be valuable, but they cannot establish durable learning by themselves. Pilots should compare assisted products, independent assessments and process evidence, while treating the paper's striking estimates as provisional findings from an observational design rather than a universal causal law.

A robust replication would preregister adoption measures and independent outcomes, test the parallel-trends assumption openly, and examine students who use AI for explanation rather than completion. That would help separate harmful outsourcing from forms of support that preserve effort.

أوراق ذات صلة

Two educators and a learner compare circuit experiment sketches and written observations in a science classroom.
ورقة مجلة2026
ورقة مجلة 198

Evidence of impact and interpretational limits of generative AI in STEM education: A systematic review and meta-analysis on cognitive learning outcomes

Chiara Boolzen, Jochen Kuhn, Salome Flegr, Eva-Maria Rott, Niklas Stausberg, Stefan Küchemann

Artificial Intelligence Review

A STEM-focused review shows why a positive average effect can be a fragile guide to classroom decisions. Across 49 studies, the conventional pooled estimate favored generative AI, but a publication-bias adjustment moved the average close to zero with substantial uncertainty. Effects still varied widely across contexts. The result calls for better comparisons and clearer descriptions of what students actually do, rather than treating every AI-supported activity as the same intervention.

STEM EducationResearch MethodsGenerative AI
اقرأ ملخص 500 كلمة →
Three teachers revise an environmental science lesson using a water-cycle diagram, a river catchment map and annotated planning sheets in a sunlit classroom
ورقة مجلة2026
ورقة مجلة 170

Teachers used AI to develop science lessons, while keeping responsibility for evidence and local relevance

Won Jung Kim

Research in Science Education

A qualitative study of 23 teachers followed AI-assisted science lesson design through surveys, lesson documents, chat logs and instructor notes. Teachers used AI to develop concepts, refine activities and connect environmental learning with student action. They also had to verify claims and repeatedly adapt generic suggestions. The study documents a supervised design process; it does not establish that AI saved preparation time or improved pupils' learning.

teacher educationlesson planningenvironmental science agency
اقرأ ملخص 500 كلمة →
A programming lecturer and two diverse university students inspect compiled code, an inheritance diagram, and a grading rubric in a computer laboratory
ورقة مجلة2026
ورقة مجلة 102

Five AI systems outscored the average OOP cohort but still failed compilation and advanced concepts

Marina Lepp, Joosep Kaimre

arXiv preprint

Lepp and Kaimre evaluated ChatGPT-5.2, DeepSeek-V3, Gemini 2.5 Flash, Claude Sonnet 4.5, and Microsoft 365 Copilot on authentic introductory OOP tests and examinations using student grading criteria. Systems exceeded the historical average and often solved long tasks, yet some code did not compile and interfaces, abstract classes, inheritance, and image-based questions remained difficult. The results challenge take-home assessment validity without proving student learning.

programming assessmentobject-oriented programminggenerative AI
اقرأ ملخص 500 كلمة →