发表机构
Munich University of Applied Sciences HM; University of Edinburgh（慕尼黑应用科学大学; 爱丁堡大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究构建结合行人动力学与空气传播的智能体暴露模型，模拟沟通和社会认同对人群病原体传播的影响，发现有效沟通与强社会认同可显著减少暴露，且流程可迁移至其他调查。
AI 中文摘要
在大型集会中，公共卫生措施只有在人们遵守时才有效，而遵守程度既取决于指令的传达方式，也取决于他人的行为。我们将这些行为因素与病原体传播联系起来，构建了一个结合行人动力学和空气传播的局部尺度、基于智能体的暴露模型。该模型并非拟合参数化行为模型，而是在运行时从调查受访者（2,170名英国体育和音乐活动参与者）中抽取其遵守程度，这些受访者报告了相同的情境：我们改变管理员（stewards）的沟通有效性和榜样（role models）的遵守程度，以及智能体与他们的共享社会认同（shared social identity），假设强烈的共享社会认同会增加他们的影响力。在一个由101个智能体组成的票务检查点队列中，有效的沟通加上与管理员强烈的共享社会认同，使高度暴露的智能体数量在佩戴口罩方面减少了82%。相比之下，与不遵守的榜样具有强烈共享社会认同，导致高度暴露的智能体数量几乎是遵守榜样情况下的九倍。物理距离并非单调有益，因为遵守程度改变了智能体的移动方式：一个保持距离的传染性智能体沿着队列边缘移动，导致各条件下的结果相似。我们的流程可迁移到任何包含行为测量和情境变量的调查中。
英文摘要
Public health measures at mass gatherings work only if people follow them, and adherence depends both on how instructions are communicated and what others do. We link these behavioural factors to pathogen transmission in a local-scale, agent-based exposure model combining pedestrian dynamics with airborne transmission. Rather than fitting a parametric behavioural model, agents draw their adherence at run time from survey respondents (2,170 attendees of UK sports and music events) who reported the same context: we vary stewards' communication effectiveness and adherence of role models, as well as agents' shared social identity with each, assuming that strong shared social identity increases their influence. In a ticket-checkpoint queue of 101 agents, effective communication combined with strong shared social identity with stewards reduces the number of highly exposed agents by 82\% for mask wearing. In contrast, strong shared social identity with non-adhering role models gives almost nine times as many highly exposed agents as with adhering role models. Physical distancing was not monotonically beneficial, because adherence changes how agents move: an infectious agent that kept distance moved along the edge of the queue, leading to similar results across conditions. Our pipeline transfers to any survey that includes behavioural measures and contextual variables.
