Social network modeling and applications, a tutorial

Synthetic networks are valuable tools for studying real-world problems when access to actual social media data is limited by proprietary restrictions or privacy concerns. Researchers can use generative models to create representative networks that allow study of issues like data privacy, spreading dynamics, algorithmic bias, and ranking inequalities without requiring access to sensitive user data.

Social network analysis has broad applications across multiple disciplines and societal challenges, including understanding employment inequality, managerial performance, and epidemic spread. The tutorial emphasizes that these problems can now be studied at global scale through digital footprints, making network science relevant to responsible AI discussions about fairness and societal impact.

Data accessibility remains a critical barrier to responsible AI research. The paper identifies that proprietary data ownership and representativeness issues prevent scientists from studying important societal questions, suggesting that more transparent data sharing practices (with appropriate privacy protections) could advance research on algorithmic bias and other responsible AI concerns.

Understanding network formation mechanisms is essential for building trustworthy systems. The tutorial highlights network selection analysis as a method to characterize how edges form in real-world networks, which is important for identifying and mitigating biases that may be embedded in algorithmic systems built on social network data.