Social influence maximization under empirical influence models.

Area of Science:

• Network Science
• Computational Social Science
• Information Diffusion

Background:

• Social influence maximization seeks minimal seed nodes for maximal diffusion.
• Empirical evidence highlights network assortativity and influence/susceptibility distributions as key factors.
• Current models often neglect these empirically observed network features.

Purpose of the Study:

• To specify and study empirically motivated influence models.
• To evaluate their implications for influence maximization in diverse networks.
• To compare their performance against traditional models.

Main Methods:

• Development of empirically motivated social influence models.
• Testing models on six synthetic and six real-world social networks.
• Analysis of influence propagation and seed set identification.

Main Results:

• Traditional models underestimate influence propagation by 21.7% on average.
• Overlap between traditional and empirical seed sets is only 19.8%.
• Empirical models select less central nodes with more cohesive ties.

Conclusions:

• Ignoring network assortativity and influence distributions leads to inaccurate maximization.
• Empirically motivated models offer more realistic influencer identification.
• These models can enhance intervention designs for information or behavior change.