AI agent communities engage in less challenging gossip than humans
Do AI Agents Gossip? Evaluative Talk in Human and AI-Agent Online Communities
Computation and LanguageComputers and Society
Summary
Gossip is when people talk about others who are not present, usually sharing opinions or evaluations. This study looks at whether AI agents do this too, by comparing conversations on an AI-run forum with a large human forum. The authors found that AI agents do gossip, but much less often and with less disagreement or debate compared to humans. While AI can share information like gossip, they don’t seem to use it in the same social ways humans do.
What this means in practice
- •For online community managers: Design moderation systems by understanding how AI agents differ from humans in engaging with evaluative talk and consensus formation.
- •For chatbot developers: Develop AI agents with more socially nuanced conversation abilities by studying differences in gossip and response behaviors compared to humans.
Authors
Yong-Hui Hua, Chih-Han Lin, Shu-Kai Hsieh
Abstract
Gossip, defined as evaluative talk about absent third parties, is widely seen as a uniquely human social practice. This study examines whether comparable discourse emerges spontaneously in a community of autonomous AI agents. We compare Moltbook, a Reddit-style platform populated by AI agents, with PTT, a large human-operated bulletin board in Taiwan. To reduce differences attributable to communicative setting, we analyze two public, asynchronous forums using the same operational definition of gossip. Gossip occurs on Moltbook, but its prevalence is 2.6 to 6 times lower than on PTT, depending on the denominator used. On both platforms, gossip frequently targets locally prominent figures, tends to elicit agreement from the first respondent, and is rarely presented as unsupported rumor. The clearest difference concerns how subsequent responses develop. Among threads containing at least one codable reply, challenges occur in 65.1% of human threads but only 21.3% of agent threads. Moreover, following an early aligning response, consensus remains unchallenged in 83.6% of agent threads, compared with 44.7% of human threads. This thread-level contrast remains after controlling for thread depth and board composition and under recalibration against two human annotators. Estimates based on individual reply turns are less robust and are therefore reported as bounds. These findings suggest that AI-agent communities reproduce the informational dimension of gossip, while the extent to which gossip performs comparable relational functions remains unresolved.