Conceptualizing Collaborative Agentic Information Systems through Shared Cooperative Activity
Year: 2026
Type: conference-paper
Abstract: Autonomous multi-agent systems introduce a new organizational reality: humans and multiple AI agents pursuing shared goals, negotiating roles, and adapting responsibilities in real time.Yet existing frameworks for human-AI collaboration remain largely anchored in dyadic settings.Drawing on Bratman's Shared Cooperative Activity theory, a systematic literature review, and expert consultation, we derive three collaboration principles for Collaborative Agentic Information Systems: dynamic goal alignment, specifying how shared objectives are continuously adapted; mutual orchestration, addressing how roles and decision rights must be dynamically negotiated; and adaptive action and co-learning, specifying how participants proactively support each other while preserving human oversight as agent autonomy increases.We illustrate these principles through a Model Context Protocol, suggesting how resource, tool, and sampling layers provide architectural primitives for human-multi-agent collaboration.These principles address how delegation, accountability, and human oversight can be sustained as organizations transition from dyadic human-AI interaction to multi-agent collaboration.
Authors: Kozachek, Diana, Janson, Andreas, Mahei Li
Institution: University of St.Gallen
Language: English
Subfield: Management Information Systems
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7220782018