Analyzing a User's Contributive Social Capital Based on Acitivities in Online Social Networks and Media

To improve the quality of communication in Online Social Networks and Media (OSNEM), we envision a system that models a person's contributive social capital (CSC), which encompasses their competence, trustworthiness, and social responsibility. Having the CSC score available may inspire social behavior and mutual support. The system is based on three pillars: the analysis of OSNEM activity, interactions in virtual social capital market systems, and personal endorsements. In this paper we present our investigations regarding the first pillar. To obtain a dataset, we ran an experiment where 165 participants interacted on a custom social networking platform and assessed each other. Ground truth data was derived from these assessments. The dataset shows characteristics that are similar to larger OSNs. With different machine learning algorithms we investigated the hypothesis that contributive social capital can be extracted from network properties and networking activity, which were assessed with features such as the number of contributions of each participant. The prediction of contributive social capital showed an improvement over the baseline. A ranking of the participants following their predicted CSC scores showed a moderate correlation with the ranking according to the ground truth assessment. We also investigated the relative importance of the features for the analysis, and the effect of excluding inactive users to better understand network dynamics on a micro level. The selected features are also available in most other OSNEM platforms, like Facebook and Twitter. This allows a large-scale application of our investigations.