Computational Models for Social Influence Analysis

Social influence occurs when one’s opinions, emotions, or behaviors are affected by others, intentionally or unintentionally. In this article, we survey recent research progress on social influence analysis. In particular, we first give a brief overview of related background knowledge, and then discuss what is social influence. We try to answer this question in terms of homophily and the process of influence and selection. After that, we focus on describing computational models for social influence including models for influence probability learning and influence diffusion. Finally, we discuss potential applications of social influence. Preliminaries First we introduce some basic knowledge for social network analysis, including related theories in sociology, fundamental models underlying social networks, and algorithms or measures for quantifying social influence. For social theories, we introduce social balance theory [10], social status theory [21], structural hole theory [6, 22], and two-step information flow theory [19]. For social network models, we introduce the Erdös-Rényi (ER) model [11], Small-World model [28], and the Barabási-Albert (BA) model [4]. For algorithms, we review several fundamental problems in graph theory and the corresponding algorithms to solve them including the Ford-Fulkerson algorithm for maximum flow in a flow network [12], the push-relabel maximum flow algorithm for finding k-densest subgraph [13], and the greedy algorithm for the set covering problem [1]. We will also go through standard measures and concepts of social networks and their connection to