The high risk of construction operations makes workers' safety attitudes an important factor in accident prevention, but the current understanding of the aggregation characteristics of safety attitudes and their propagation laws is still insufficient. In order to improve the pertinence and effectiveness of construction safety management, based on the survey data of construction workers' safety attitudes and social networks, the network-driven mechanism of safety attitudes is systematically revealed by constructing work networks and social networks, combined with the Moran index, QAP regression and centrality path effect analysis. The results show that the safety attitude of construction workers shows significant aggregation in both work networks and social networks, especially in the work network, the closer the distance between nodes, the more similar their safety attitudes are. Nodes with high eigenvector centrality can significantly promote the convergence of safety attitudes of surrounding nodes. It is worth noting that in the process of aggregation and disseminating safety attitudes, the structural effect of work networks is significantly stronger than that of social networks. The research results can provide a theoretical basis and practical reference for networked intervention strategies in building safety management.