The group identification literature mostly revolves around the problem of identifying individuals in the community who belong to groups with ethnic or religious identity. Here we use the same model framework to identify individuals who play key role in some sense. In particular we will focus on expert selection in social networks. Ethnic groups and experts groups need completely different approaches and different type of selection rules are successful for one and for the other. We drop monotonicity and independence, two common requirements, in order to achieve stability, a property which is indispensable in case of expert selection. The idea is that experts are more effective in identifying each other, thus the selected individuals should support each others membership. We propose an algorithm based on the so called top candidate relation. We establish an axiomatization to show that it is theoretically well-founded. Furthermore we present a case study using citation data to demonstrate its effectiveness. We compare its performance with classical centrality measures.
Group identification Expert selection Stability Citation analysis Nucleolus