The paper "Predicting Positive and Negative Links in Online Social Networks" by Jure Leskovec explores the prediction of positive and negative links in online social networks, where relationships can be either positive (e.g., friendship) or negative (e.g., opposition). The study uses datasets from Epinions, Slashdot, and Wikipedia to investigate the patterns of signed links and their generalizability across different platforms. The authors develop a machine-learning framework to predict the sign of edges in these networks, using features based on node degrees and two-step paths involving third parties. The results show that the learned models outperform previous methods and generalize well across the datasets, indicating that there are underlying principles governing the formation of signed links. The study also compares the learned models to social-psychological theories of balance and status, finding that while both theories agree in certain aspects, there are also discrepancies, suggesting that the learned models capture more nuanced effects. Additionally, the paper discusses the global structure of signed networks, exploring whether the theories of balance and status can be applied to predict the overall pattern of signs in the network.The paper "Predicting Positive and Negative Links in Online Social Networks" by Jure Leskovec explores the prediction of positive and negative links in online social networks, where relationships can be either positive (e.g., friendship) or negative (e.g., opposition). The study uses datasets from Epinions, Slashdot, and Wikipedia to investigate the patterns of signed links and their generalizability across different platforms. The authors develop a machine-learning framework to predict the sign of edges in these networks, using features based on node degrees and two-step paths involving third parties. The results show that the learned models outperform previous methods and generalize well across the datasets, indicating that there are underlying principles governing the formation of signed links. The study also compares the learned models to social-psychological theories of balance and status, finding that while both theories agree in certain aspects, there are also discrepancies, suggesting that the learned models capture more nuanced effects. Additionally, the paper discusses the global structure of signed networks, exploring whether the theories of balance and status can be applied to predict the overall pattern of signs in the network.