Betsy Sinclair study tests AI pre-bunking for election misinformation

Washington University political scientist Betsy Sinclair has published research in Royal Society Open Science on a scalable AI-assisted framework for rapidly generating pre-bunking materials against election misinformation. Sinclair, chair of political science and Thomas F. Eagleton Professor of Public Affairs and Politics, worked with Mitchell Lineager, Sander van der Linden, and R. Michael Alvarez.

The team tested the approach in a pre-registered two-wave experiment with 4293 United States (US) registered voters, focusing on politically charged election misinformation. LLM-generated pre-bunking significantly reduced belief in election rumours, with effects persisting, though attenuated, one week later, and modestly offset declines in confidence in national election administration. The researchers reported no evidence of partisan backlash.

After the prompt was finalized, pre-bunking materials written with human feedback were no more effective than versions generated using only AI, suggesting that per-rumour human effort can be reduced once the framework is established. The team also released an interactive demonstration at electionbot.chat/article that drafts pre-bunking articles from trusted facts.