Study: Conspiracy rabbit holes spread like interacting social contagions

A new study from Fabian Tschofenig, Paul Vicinanza, Henrich Greve, Hayagreeva Rao, and Douglas Guilbeault (INSEAD, Berkeley) examines why people fall into conspiracy theory rabbit holes. Rather than focusing solely on psychological or algorithmic causes, the researchers argue that rabbit holes can emerge from interactions among conspiracy theories spreading as social contagions. Using 7.6 million tweets from 7,416 X users during the first wave of COVID-19, they identified public endorsement of 15 conspiracy narratives with prompt-tuned large language models. Sequential hazard models revealed that adopting a conspiracy theory elevates the risk of sharing subsequent theories—and this effect grows and persists longer with each additional theory shared. Transitions between theories cluster among semantically similar narratives, suggesting semantic interactions mediate the spread.

The study introduces the "settler effect": a user's entry into a new semantic region is slow, but once inside, subsequent adoption of related theories accelerates. Agent-based models show that an "ecology-of-contagions" framework—where belief systems reshape after each endorsement—best reproduces these dynamics. Counterfactual network simulations found that preventing the first public endorsement of a conspiracy theory can rival high-detection shadow banning in effectiveness, and outperforms week-long read-only lockouts. The findings offer practical implications for platform moderation, suggesting early intervention at the first sign of engagement can be disproportionately impactful.