CS-RAD: Conditional Member Status Refinement and Ability Discovery for Social Network Applications

In a social network environment, member status represents a member's social value in the network. A member's abilities represent the potential of a member projecting his/her social values to others, and also represent the level of credibility and authority for a member to hold certain status. Therefore, the concepts of status and ability are deeply related, and should be consistent with each other. In this paper, we establish the consistency models among different member status and their abilities through analyzing member data and integrating domain knowledge. We use these models to help our members refine their inconsistent status, at the same time, identify ability gaps. To reliably refine a member status, we introduce a practical and human-in-the-loop methodology to build status hierarchy. Conditioned on the hierarchical structure, our modeling process exploits the associations between status and abilities. We applied the technique to LinkedIn member titles -- one of the major types of the member status, and member skills -- the main ability representations at LinkedIn. We showed that our models are intuitive and perform well. The skill gaps identified are actionable and concise. In this paper, we also discuss the aspects of building such systems, and how we could deploy the models in production.