LLMs Can Fuel Extremist Attitudes Using Universal Moral Framings.

Résumé

Social media influence campaigns are thought to sway public opinion, particularly during election campaigns and national crises. These campaigns are often based on generative artificial intelligence technologies that flood the internet with polarizing content. How these social media influence operations change public opinion and ignite extremist attitudes is not well understood. We evaluate whether short, large language model (LLM)-generated arguments using universal moral framings impact extremist attitudes. In two studies with Democrats and Republicans in the United States (N = 951), we find that universal moral concerns related to welfare, rights, and fairness predict perceptions of political stances as absolutist moral obligations, known as sacred values, and explain extremist attitudes in defense of these values. We also find that short LLM-generated arguments that appeal to individual rights and fairness increase people's willingness to fight and die and justify violence to defend political stances. This effect is partially mediated by increased perceptions of the stance as having an absolute value. Our findings shed light on the potential of social media influence campaigns in polarizing society through LLM-generated messaging.