Abstract
Humans are social animals who communicate with each other even during locomotion, forming social groups within crowds. Social groups are reliably perceived based on head orientation and proximity, consistent with the social interaction field model (Zhou, Han, Liang, Hu, & Kuai, 2019). Previously, we found that a pedestrian follows a crowd by averaging the walking direction (heading) of all neighbors, with a weight that decays with distance (Rio, Dachner, & Warren, 2018). However, it seems obvious that we can follow a social subgroup (e.g. of friends). Here we investigate whether a social group is completely segmented, or affects the weight of neighbors in the group. Participants (N = 12) were instructed to “walk with” a virtual crowd of 9 neighbors while wearing a Samsung Odyssey HMD (FOV = 90˚ H). The virtual crowd contained (1) one subgroup of adjacent neighbors (A(R)=3) with their heads rotated toward each other, and (2) one subgroup of distractors (D=6) facing straight ahead. In each trial, one or both subgroups appeared and began walking; after 4-5s, the heading direction of one subgroup (A or D) was perturbed by ±10˚. The participant’s final heading direction was measured as the dependent variable. In two control conditions, both subgroups appeared with heads oriented straight ahead, and one subgroup was perturbed (A(S) or D). There were thus eight conditions, each with 10 repetitions presented in a randomized order. The results show that participants did not segment the social group from the crowd, but were influenced by all neighbors. However, the social group (A(R) with rotated heads) had a significantly stronger influence on the participant’s response (higher weight) than the asocial group (A(S) with straight heads) (p<.001). Future studies will try to incorporate the social interaction field model into the behavioral dynamics model to quantify the impact of social groups.