Decision-Making in Human Crowds: Nonlinear Competition Dynamics
Decision-Making in Human Crowds: Nonlinear Competition Dynamics
复制标题
人群决策：非线性竞争动态
DOI：
10.1167/jov.21.9.2566
复制
发表时间：
2021
影响因子：
1.8
通讯作者：
Warren, William H.
中科院分区：
文献类型：
--
作者：
Wirth, Trenton D.;Free, Brian;Warren, William H.
当一群人分成两组朝不同的方向走时，个人如何决定跟随哪一组？我们之前发现(VSS 2019)，参与者对视场中所有邻居的前进方向进行平均，组间角度差高达40，这与Rio，Dachner&Warren(PRSB 2018)的邻里模型一致。在这里，我们研究哪些规则管理具有较大角度差异的个体决策(≥80)。参与者(N=10)被指示戴着三星奥德赛头盔(110°FOV)与虚拟人群(8或16个虚拟人)“同行”。在2-3s后，一组做了一个小转弯(左或右20)，另一组做了一个大转弯(60、100、140或180)。我们操纵大转弯人群的百分比(0%，25%，50%，75%，100%)，并测量参与者的行走方向。结果显示，对多数人(p<；0.001)有很强的吸引力，而不受总人群规模的影响(p=0.27，与法定人数相反)。然而，这种效应与对小转弯的吸引力相抵消：随着大转弯幅度的增加，参与者更有可能跟随少数人进行更小、更不费力的转弯(p<；0.001)。有趣的是，存在个体差异：虽然大多数参与者遵循这一权衡，但少数人总是倾向于多数或小转弯。为了对这些数据进行建模，我们在邻里模型中引入了非线性竞争动力学，使得个体的决策是由人群比例的相对权重和群体之间的角度差异驱动的。我们对每个参与者的比赛参数进行了拟合，并在所有的实验测试中模拟了航向。新模型占数据集中个人决策的86%。
When a crowd splits into two groups walking in different directions, how does an individual decide which group to follow? We previously found (VSS 2019) that participants average the heading directions of all neighbors in the field of view up to an angular difference of 40 between groups, consistent with Rio, Dachner & Warren’s (PRSB 2018) neighborhood model. Here we investigate what rules govern individual decisions with larger angular differences (≥ 80). Participants (N= 10) were instructed to “walk with” a virtual crowd (8 or 16 virtual humans) while wearing a Samsung Odyssey HMD (110º FOV). After 2-3s, one group made a small turn (20 left or right) and the other group made a large turn (60, 100, 140, or 180) in the other direction. We manipulated the percentage of the crowd making the large turn (0%, 25%, 50%, 75%, 100%), and measured the participant’s walking direction. The results show a strong attraction to follow the majority (p< 0.001), with no influence of total crowd size (p= 0.27, contrary to a quorum). However, this effect trades off with an attraction to the smaller (20) turn: as the magnitude of the large turn increased, participants were more likely to follow a minority making the smaller, less effortful turn (p< 0.001). Interestingly, there were individual differences: while most participants followed this trade-off, a few always preferred either the majority or the small turn. To model this data, we introduced nonlinear competition dynamics into our neighborhood model, such that an individual decision is driven by the relative weight of the crowd proportion and the angular difference between groups. We fit the competition parameters to each participant, and simulated heading in all experimental trials. The new model accounted for 86% of the individual decisions in the data set.