Stirring Ripples Across the Lake: The Impact of AI Social Member's Comments on the Opinion Diversity

Current studies on social bots have primarily focused on their functional advantages (such as moderation and information retrieval) or human-AI dyadic interactions and underexplored their more extensive social functions. We fill this gap by examining a new situation in which bots act like regular members in social groups, termed AI-bots as Social Members (ASMs), within online communities. Using data from a leading Asian social platform, we demonstrate that ASM participation significantly increases subsequent opinion diversity among human members. The characteristics of ASM-generated comments and ASMs' social identity jointly shape this effect. Specifically, subsequent human opinions diverge more when ASM's comments are emotionally neutral and consistent with prior human opinion. In contrast, human-generated comments with similar neutrality and consistency reduce opinion diversity, highlighting a fundamentally distinct social influence of ASM relative to human contributors. We also find that this effect is stronger when ASM participates autonomously and occupies a higher rank within the community. Additionally, ASM-induced opinion divergence is more pronounced in discussions centered on subjective topics and in communities characterized by higher modularity. We also step further to consider the consequences of ASM-induced opinion diversity. We find that while such opinion diversity fosters higher-quality content and innovation, it also increases offensive language. This research advances the understanding of human-AI interaction within social contexts and how bots can influence the dynamics of opinion formation.