PurposeThis study aimed to identify driving factors of safety attitudinal ambivalence (AA) and explore their influence. Construction workers' intention to act safely can be instable under conflicting information from safety management, co-workers and habitual unsafe behaviour. Existing research explained the mechanism of unsafe behaviours as individual decisions but failed to include AA, as the co-existence of both positive and negative attitude.Design/methodology/approachThis study applied system dynamics to explore factors of construction workers' AA and simulate the process of mitigating the ambivalence for less safety behaviour. Specifically, the group model building approach with eight experts was used to map the causal loop diagram and field questionnaire of 209 construction workers were used to collect empirical data for initiating parameters.FindingsThe group model building identified five direct factors of AA, namely the organisational safety support, important others' safety attitude, emotional arousal, safety production experience and work pressure, with seven feedback paths. The questionnaire survey obtained the initial values of the factors in the SD model, with the average ambivalence at 0.389. The ambivalence between cognitive and affective safety attitude was the highest. Model simulation results indicated that safety experience and work pressure had the most significant effects, and safety experience and positive attitude of co-workers could compensate the pressure from tight schedule and budget.Originality/valueThis study provided a new perspective of the dynamic safety attitude under the co-existence of positive and negative attitude, identified its driving factors and their influencing paths. The group model building approach and field questionnaire surveys were used to provide convincible suggestions for empirical safety management with least and most effective approaches and possible interventions to prevent unsafe behaviour with tight schedule and budget.
Developing social-ecological system indicators using group model building
