Meta AI helps WhatsApp admins create group rules, study finds

Researchers from Cornell and IIT conducted a speculative design study with 20 WhatsApp group admins in India to examine how they would collaborate with an AI assistant—specifically Meta AI—to create and enforce group rules. The study, published as a CSCW 2026 paper on arXiv, involved two phases: first, admins co-created rules with the AI; second, they responded to probes illustrating potential AI-assisted moderation features like automatic rule suggestions and enforcement alerts. The findings reveal a nuanced picture: admins valued the AI's ability to identify rules they hadn't considered (e.g., prohibiting forwarded messages during peak hours) and appreciated reduced manual workload. However, they were highly sensitive to relational trust (fearing the AI might misread tone in culturally specific jokes), data privacy (worrying about message scanning), and loss of social nuance in automated responses.

The study also found that the type of group (e.g., professional vs. family) and the admin's personal style (authoritative vs. democratic) strongly influenced how willing they were to delegate authority to the AI. For instance, family group admins resisted AI enforcement entirely, while professional group admins were more open to automated warnings. The researchers identified limitations of current chatbot interfaces for collaborative rule-making, noting that admins wanted more control over editing rules iteratively rather than accepting AI-generated drafts wholesale. They concluded with design implications for building moderation tools that center human judgment, relational nuance, and contextual adaptability—arguing that over-automation could harm group cohesion. The work directly addresses WhatsApp's lack of built-in moderation infrastructure, suggesting a hybrid human-AI approach as a viable path forward for the platform's billions of users.