Peer-reviewed article
Cheating among elementary school children: A machine learning approach
Li Zhao, Yi Zheng, Junbang Zhao, Guoqiang Li, Brian J. Compton et al. ·
Summary
Written by AI, not yet AICS's view · claude-haiku-4-5-20251001This study examined academic cheating among Chinese elementary school children in grades 2-6, finding that about 26% reported cheating. Using machine learning analysis, researchers identified that children's beliefs about whether cheating is acceptable and their observations of peer cheating were the strongest predictors of their own cheating behaviour. The study demonstrates how machine learning can help understand patterns in child development and educational behaviour.
Key points
- 25.74% of Chinese elementary school children (grades 2-6) reported having cheated academically
- Children's beliefs about the acceptability of cheating was the strongest predictor of cheating behaviour
- Observed prevalence and frequency of peer cheating at school was the second strongest predictor
- A Random Forest machine learning algorithm achieved 81.43% accuracy in predicting cheating behaviour
- Academic cheating emerges early in elementary school and is influenced by social and normative factors
Limitations
The study is based on self-reported cheating behaviour from Chinese children only, so findings may not generalise to other cultural contexts or age groups. The cross-sectional design cannot establish causal relationships between predictors and cheating behaviour.
Abstract
Academic cheating is common, but little is known about its early emergence. It was examined among Chinese second to sixth graders (N = 2094; 53% boys, collected between 2018 and 2019) using a machine learning approach. Overall, 25.74% reported having cheated, which was predicted by the best machine learning algorithm (Random Forest) at a mean accuracy of 81.43%. Cheating was most strongly predicted by children's beliefs about the acceptability of cheating and the observed prevalence and frequency of peer cheating at school. These findings provide important insights about the early development of academic cheating, and how to promote academic integrity and limit cheating before it becomes entrenched. The present research demonstrates that machine learning can be effectively used to analyze developmental data.