Fairness in Influence Maximization through Randomization

The influence maximization paradigm has been used by researchers in various fields order to study how information spreads social networks. While previously the attention was mostly on efficiency, more recently fairness issues have taken into account this scope. In present paper, we propose use randomization as a mean for achieving fairness. general idea is not new, it applied area.
 Similar previous works like Fish et al. (WWW ’19) and Tsang (IJCAI ’19), maximin criterion (group) contrast their work however, model problem such way that, when choosing seed sets, probabilistic strategies are possible rather than only deterministic ones. We introduce two different variants of problem, one that entails over nodes (node-based problem) second sets (set-based problem). After analyzing relation between problems, show while original inapproximable, both permit approximation algorithms achieve constant multiplicative factor 1 − 1/e minus an additive arbitrarily small error due simulation spread. For node-based achieved observing polynomial-sized linear program approximates well. set-based multiplicative-weight routine can yield result.
 experimental study, provide implementations routines problems compare values existing methods. Maybe non-surprisingly, ex-ante values, i.e., minimum expected value individual (or group) obtain information, computed significantly larger (ex-post) This indicates studying via worthwhile path follow. Interestingly maybe surprisingly, observe even ex-post sampled according our routines, dominate methods many instances tested.