Generalizability Across Platforms, Periods, and Contexts

Ascertain whether the small increases in following and liking news content induced by GPT-2–based bot replies on Twitter would replicate on other social media platforms, in different time periods, and in different sociopolitical contexts.

Background

The experiment was conducted on Twitter, a platform with distinctive political discourse dynamics and user demographics. The authors caution that these features may limit external validity.

They explicitly state uncertainty about whether similar effects would appear elsewhere or under different conditions and encourage replication on other platforms.