Agentic Sociology: Relations and Institutions for Productive Agent Systems

We argue that Persistent Agent Organizations and Agent Societies should be central targets of agent research. Organizations preserve workflows, responsibilities, and review protocols across tasks and member changes. Societies connect independently managed activities through lasting relationships, shared norms, and institutions. We call the applied study of these systems agentic sociology. Roles, relationships, and shared rules should be first-class design elements alongside models, tools, and memory. Architectures shape how cooperation can develop; social arrangements then change what members know, can do, and must do. A five-level framework connects individual action to persistent organizations and a wider society. Building on existing research, we develop three organizational questions about combining capabilities, adapting rules, and expanding capacity, and three societal questions about specialization, continuing agreements, and public authority. Language lets participants propose, negotiate, and revise these arrangements. The goal is reliable, efficient collective work across changes in members, tasks, and partnerships, with coordination and recovery costs included.