Can AI Write Persuasive Propaganda?

Abstract

Can large language models, a form of artificial intelligence, write persuasive propaganda? We conducted a pre-registered survey experiment to investigate the persuasiveness of news articles written by foreign propagandists compared to content written by GPT-3 davinci (a large language model). We found that GPT-3 can write highly persuasive text. We investigated whether a person fluent in English could improve propaganda persuasiveness: editing the prompt fed to GPT-3 or curating GPT-3's output made GPT-3 even more persuasive, and, under certain conditions, as persuasive as the original propaganda. Our findings suggest that if propagandists get access to GPT-3-like models, they could create convincing content with limited effort.