Public health measures at crowded events only work if people follow them, and whether people follow them depends on what everyone else is doing. We model this feedback in a local-scale exposure model, where agents keep observing the crowd and revise their behaviour accordingly. In a ticket-checkpoint queue of 101 agents, one of them infectious, agents are told to wear masks. Each agent sees how many others around them already comply. They make their own initial decision according to what respondents observing the same situation reported in a survey. Every two seconds the agents revise their decisions. In our scenario, part of the crowd acts as role models for the others. In the role model group, either everybody wears a mask or nobody does. Among the other agents we observe how their behaviour adapts to what they observe: In the absence of role models, they switch from complete non-adherence to near-complete adherence within seconds, purely by following each others lead. If they see role models wear masks the switch is faster, reaching the same final adherence and exposure levels. The only way to change this outcome is if the role model group do not wear masks. Then, the larger this group is, the more agents will be highly exposed.