Background: The construction industry faces significant safety hazards, frequent accidents, and inadequate management. Studies identify unsafe worker behaviors as the primary cause of construction accidents. However, most research overlooks the psychological state, particularly emotions, of construction workers. Methods: This study designed a behavioral experiment integrating social cognitive neuroscience, collecting real-time EEG data to classify and recognize fear, anger, and neutral emotions. Variance analysis explored differences in safety hazard identification and risk assessment under these emotional states. A total of 22 male participants were involved, with data collection lasting three days. The role of psychological capital in mediating the effects of emotions on unsafe behaviors was also examined. Results: Emotional classification using EEG signals achieved 79% accuracy by combining frequency domain and nonlinear feature extraction. Fear significantly enhanced safety hazard identification accuracy compared to neutral and anger emotions (F = 0.027, p = 0.03). Risk assessment values under fear and anger were higher than under neutral emotion (F = 0.121, p = 0.023). Psychological capital interacted significantly with emotions in hazard identification accuracy (F = 0.68, p = 0.034), response time (F = 2.562, p = 0.003), and risk assessment response time (F = 1.415, p = 0.026). Safety hazard identification correlated with the number of safety trainings (p = 0.002) and safety knowledge lectures attended (p = 0.025). Risk assessment was significantly associated with smoking (p = 0.023), alcohol consumption (p = 0.004), sleep duration (p = 0.017), and safety training (p = 0.024). Conclusions: The findings provide insights into how emotions affect safety hazard identification and risk assessment, offering a foundation for improving emotional regulation, reducing accidents, and enhancing safety management in construction.