Understanding the Diverging Dynamics of Conspiracy Theories on Twitter

The final speaker in this AoIR 2019 session is QUT DMRC PhD graduate Dr. Jing Zeng, whose focus is on the automated dissemination of conspiracy theories on Twitter – including suggestions that celebrities like Justin Bieber, industry leaders like Mark Zuckerberg, and royals are actually shape-shifting lizards; that planes spread mind-controlling chemtrails; that the Earth is flat; or that the California wildfires were started by a new energy weapon created by the U.S. government.

Such conspiracy theorists are experts at providing apparently simple explanations for complex phenomena. They also clusters together to support each other’s explanations with self-reinforcing theories that paint all opponents as part of the conspiracy. But ‘conspiracy theory is often either used as a generic term that classifies all such theories as equivalent, or focusses only on very specific cases. It would be more useful to develop a better typology of such theories and their dynamics, and this aim is what this project pursues.

The project works with some 100,000 tweets by 34,000 Twitter accounts over a period of several months during 2018 and 2019; it examined the clusters in the retweet networks between them, and they key actors in these clusters. Such clusters focussed variously on anti-vaccination, flat-earth, anti-agenda 21 (a UN agenda for sustainable development), chemtrails/illuminati/lizard-people, pizzagate/911/climate-change and directed energy weapon conspiracies. Anti-flat earth and pro-vaccination groups were more separate from this.

The project also calculated the Botometer score for all of these accounts, and it turns out that bots are considerably more prominent in the politically tinged and U.S.-focussed conspiracies. Further manual analysis of the 120 most influential accounts in the network provides some additional insights: first, there are some fairly inactive yet highly influential human accounts on either side of these conspiracy theories.

Second, there are a number of highly active bots that contribute especially into U.S.-based conspiracy theories, are mostly expressing support for Donald Trump, include three star emoji in their account names, and post to the #TrumpTrain hashtag that helps promote other pro-Trump accounts; indeed, some 20% of all accounts in the underlying dataset self-identify as Trump supporters. It is difficult to establish whether all of these are actually bots, however; some hardcore Trump supporters post so frequently to appear as bots to Botometer.