Modelling and analysis of risk-taking behaviours propagation among construction workers based on SIRS

Purpose Risk-taking behaviour is a prevalent unsafe act in construction crews and may spread through onsite interactions, undermining safety performance. This study develops a modified susceptible–infectious–recovered–susceptible (SIRS) model to simulate the occurrence, propagation, mitigation, and recurrence of risk-taking behaviour at the crew level. Design/methodology/approach Behaviour-specific influencing factors are linked to analyse risk-taking transmission. Influencing factors were elicited using grounded theory and operationalized as indicators associated with five transition rates (transmission, recovery, awakening, forgetting, and control). A survey was used to calibrate model parameters via entropy-weighted aggregation. The model was implemented and examined in a real project case with a relatively stable crew, and controlled-variable simulations were conducted to assess parameter effects and structural sensitivity. Findings Risk-taking propagation exhibits rapid early diffusion followed by mitigation and convergence, with recurrence possible under relapse. Transmission (β), awakening (γ), and forgetting (δ) dominate the propagation tendency: higher β accelerates diffusion and increases peak prevalence, higher γ provides the strongest suppression and speeds convergence to a low-risk regime, and higher δ sustains long-run risk through relapse. Recovery (α) and control (κ) show incremental effects. Interaction intensity also amplifies propagation risk, whereas sparse connectivity bounds the peak by the initial infected proportion. Originality/value This study extends construction safety research by modelling risk-taking behaviour as a dynamic and recurrent propagation process. The framework links field-derived behavioural indicators with SIRS transition rates, providing quantitative evidence for prioritizing prevention to reduce recurrence on construction sites.