When humans and agents form a collective: a framework for collective intelligence in persistent organizational human-AI teams

Abstract

Artificial systems increasingly participate in organizational work as recurrent contributors to interdependent activity, yet research remains fragmented across system capabilities, individual responses, episodic interactions, and constructs derived from human-only teams. This article develops Organismic Human-Artificial Intelligence (OHAI) as a bounded integrative theoretical framework for studying collective intelligence in persistent organizational human-AI teams. Its focal unit is a bounded collective of distinguishable human and artificial contributors whose interdependent activity extends across successive cycles and whose prior collective activity can affect later artificial participation. The framework distinguishes system configuration, collective functioning, and collective intelligence (CI) as analytical objects, with CI as its primary theoretical target. CI denotes a general group-level capacity manifested across diverse tasks within a specified and empirically qualified task universe. Task performances may provide indicators for estimating CI or serve as external criteria, while the framework separately examines broader employee, team, and organizational consequences, including changes in human working relationships and employee wellbeing. Its central inferential rule is that a property implemented in an artificial component does not by itself establish a corresponding team-level phenomenon. Collective functioning must instead be demonstrated in the enacted organization and use of distributed contributions. Partner modeling, transactive-memory-system-related coordination, and shared-mental-model-related representational alignment provide selective lenses on such functioning. Persistence makes collective history consequential, bringing recursive updating, overlapping timescales, reconfiguration, and collective-intelligence trajectories into view without presuming adaptation or improvement. Human-only teams serve as secondary epistemic references for bounded tests of observable-pattern and functional correspondence, while relational-pattern correspondence remains a limited future methodological possibility. The empirical program distinguishes four nonmandatory stages of research maturity: configuration and measurement qualification, local relations, longitudinal team-level phenomena, and integrated multilevel studies. The framework thereby provides a disciplined basis for cumulative research on how heterogeneous contributions become collectively consequential and change over time.