Data mining social networks: self-selection in a Twitter contest

Abstract

Conducting Twitter contests is a method used by companies to increase their Twitter followers which in turn increases the audience for their messages. These contests are usually targeted at a certain class of users. Thus, users selectively opt-in (self-selection), unlike in random sampling or filtering where the users are randomly selected or directly sought out. In this paper, we looked for evidence of self-selection using macroscopic properties of the contest entries time series. We found that users who joined the contest have a different tweeting behavior from a typical user, which can then be observed without looking at the individual behavior of each.