Abstract
Social media platforms are facing declining user engagement and increasingly turning to artificial intelligence (AI) to re-activate users. One strategy is deploying socially intelligent AI to generate humorous content—a powerful driver of online interactions. Yet research shows that identical content is often judged less favorably when attributed to AI, raising doubts about its effectiveness. Drawing on benign violation theory, we investigate how users respond to humorous responses from AI versus those from humans. An observational study of 18,601 posts on a Chinese social media platform shows that AI-generated humorous comments consistently attract more engagement by stimulating direct interactions with the comments than human-generated humorous comments. We then outline a complementary experiment to test causal mechanisms and boundary conditions. Together, these studies advance understanding of bias in human–AI interaction and offer guidance for AI-driven engagement strategies.
Recommended Citation
Lan, Xinmiao; Tang, Jie; and Li, Ting, "When the Humor Comes from a Bot: User Engagement with Socially Intelligent AI on Social Media" (2025). SIGHCI 2025 Proceedings. 16.
https://aisel.aisnet.org/sighci2025/16