Experimental Design Details
The experiment will consist of the creation of fictitious accounts on Twitter that signal their preferred candidate in the 2022 Brazilian election (Luis Inácio Lula da Silva or Jair Messias Bolsonaro) and their preferred Brazilian soccer club. The bot accounts randomly follow Twitter users who share or not each identity (political and soccer club preference) with it. After five days of activation, we compute the number of follow-backs and blocks obtained by each bot. The experiment will be run in waves during the second semester of 2022. Overall, we plan on running 44 experimental waves between July and December, 2022.
The subject pool is composed of Twitter accounts from Brazilians who tweeted or re-tweeted a status containing a pro-Lula or pro-Bolsonaro hashtag between May and July 2022. From these accounts, we identify the ones that also signal a preferred soccer club in their profile. This is our final subject pool.
In each experimental wave (with a time span of 5 days), we will activate 8 bot accounts that differ in their preferred soccer club and/or political candidate. Each account will randomly follow approximately 100 accounts from the subject pool. We perform block-randomization to determine treatment assignment, stratifying by subject's political and affective identity, as well as by whether the subject is above or below the sample's median number of followers. Each subject may be treated (i.e., followed by a bot) more than once, but never in subsequent waves: after being treated in a wave, a subject only returns to the subject pool after 3 waves.