The authors discuss the method of using F-formations to detect dominant sets in a crowd. According to the authors, “A dominant set is a form of maximal clique which occurs in edge weighted graphs.” Note that this research is applied to a real human crowd--a social network of people gathered in a public setting, not the social networks found online.
The authors define an F-formation as “a group of people who have easy ... access to the same shared space, around which they can communicate for a prolonged time.” The F-formation is modeled as graph G = (V, E, w), where V is a vertex set, E is an edge set, and w is a positive edge weight function. People are modeled as the set of vertices; connections among people are modeled as the set of edges; and the affinity is implied by the edge weight function w. Function w measures the overall relative affinity between a vertex i (a person) and the rest of the vertices (the rest of the group), weighted by the average affinity of the group. If the value w for a vertex i is above a certain threshold, the person represented by i is said to be in the dominant group. The function w takes into consideration the distance between any pair and face orientation.
The dataset used in the study “consists of real footage of over 50 people who met to present scientific work during a poster session.” The three-hour event was videotaped. The authors annotated 82 images (of about 1,700 people) and found different F-formation sizes. The data was analyzed after the singletons (one-person groups) were removed. The authors then used the distance between a pair (two meters) and face orientation as parameters to measure the affinity among the people in the crowd. The results show that F-formations are effective with regard to finding dominant sets. Compared with the modularity cut method, F-formation shows significant and more stable performance improvements.
The paper is of interest to researchers in sociology who study crowd behavior. It will also help those in the online social networking field look at the problem from a different perspective.