Troll Farms

Abstract

Political agents often aim to influence elections through troll farms -- organisations that disseminate messages emulating genuine information. We study the behaviour of a troll farm that faces a heterogeneous electorate of partially informed voters, and aims to achieve a desired political outcome by targeting each type of voter with a specific distribution of messages. We show that such tactics are more effective when voters are otherwise well-informed, for example, when the media is of high quality. At the same time, increased polarisation, as well as deviations from Bayesian rationality, can reduce the negative effect of troll farms and restore efficiency of electoral outcomes.

Keywords: social media, disinformation, persuasion, elections, polarisation, bounded rationality JEL codes: D72, information design

JEL Classification: D72, D83, D91