Attitudinal Impact of the Bot-Based News Nudge Intervention

Determine whether the two-week Twitter intervention using GPT-2–generated bot replies that linked to ideologically balanced news outlets and encouraged users to follow those outlets produced any changes in users’ political attitudes.

Background

The paper deployed 28 GPT-2–based bots to reply contextually to users’ non-political tweets with a message that included a link to a topic-relevant section of a reliable, ideologically balanced news outlet and an encouragement to follow that outlet’s Twitter account. The intervention produced small increases in following and liking news content, especially from bots presented as female, but the paper did not measure downstream attitudinal outcomes.

The authors highlight that a growing literature finds that algorithmic and platform interventions often change on-platform behavior without clear corresponding changes in attitudes. They explicitly note that their design focuses on behavioral outcomes and thus cannot establish whether their intervention affected political attitudes.