关键词: 众智, 团队认知, 人智协同

Abstract: In today's increasingly complex and volatile work environments, teams—not individuals—have emerged as the fundamental units through which organizations navigate uncertainty, solve intricate problems, and drive innovation. Central to this shift is the construct of collective intelligence, defined as the team-level general cognitive ability that enables groups to collaboratively communicate, share knowledge, and effectively address complex tasks. Unlike the sum of individual intelligences, collective intelligence arises from synergistic cognitive processes that transcend individual limitations, resulting in emergent capabilities greater than the mere aggregation of parts. Despite its critical importance, research on collective intelligence has long been hampered by two persistent challenges: (1) conceptual fragmentation due to divergent disciplinary perspectives and the absence of an integrative theoretical framework capable of explaining the processes through which collective intelligence emerges,and (2) methodological inconsistency between measurement paradigms that either prioritize outcome-based assessment or process-based diagnosis—each with significant limitations.
To advance the field, this study proposes a unified conceptualization: collective intelligence is best understood as an emergent property generated through the dynamic interplay of three core mechanisms—shared mental models（SMMs）, transactive memory systems(TMS), and interactive team cognition(ITC). Shared mental models constitute the cognitive foundation of teamwork, providing a common understanding of tasks, roles, and procedures that enables coordination, reduces ambiguity, and enhances predictability in team interactions. Transactive memory systems reflect the distributed nature of team knowledge, functioning as a collective “memory architecture” wherein members specialize in different domains and rely on one another to access and apply expertise efficiently. This system allows teams to adapt flexibly to complex demands by leveraging the full spectrum of their distributed cognitive resources. Meanwhile, interactive team cognition emphasizes that intelligence is not static or stored solely within individuals, but is continuously co-constructed through real-time communication and interaction. It serves as the active process that animates shared mental models and transactive memory systems, transforming cognitive potential into adaptive, context-sensitive performance.
Critically, these three components do not operate in isolation; rather, they form a mutually reinforcing cycle. Shared mental models facilitate smoother interaction and more effective knowledge exchange, which in turn strengthens the transactive memory system. Efficient knowledge distribution enables richer, more informed interactions, further refining shared understanding. It is within this virtuous loop—anchored in shared cognition, distributed expertise, and dynamic interaction—that collective intelligence genuinely emerges.
Building on this integrative framework and informed by a critical analysis of the limitations inherent in the two dominant measurement paradigms, we identify three pivotal directions for future research. First, there is an urgent need to integrate and optimize measurement paradigms. Traditional assessments often sacrifice process insight for outcome validity, or vice versa. Future studies should design ecologically valid team tasks that simulate real-world complexity while systematically eliciting key collaborative behaviors. By combining performance metrics with rich process data—such as communication patterns, decision sequences, and problem-solving strategies—researchers can develop comprehensive measurement frameworks that balance diagnostic precision with external validity.
Second, the field should embrace multimodal, dynamic assessment systems. Advances in sensing technologies and computational methods now allow for the simultaneous capture of behavioral, vocal, eye-tracking, physiological synchrony, and even neurocognitive data during team interactions. Integrating these multimodal streams through methods such as machine learning can yield granular, time-sensitive insights into how collective cognition unfolds in real time, moving beyond static snapshots to capture the fluid, emergent nature of collective intelligence.
Third, and perhaps most urgently, research should expand to address Human-AI collaborative teams. As artificial intelligence becomes an integral team member in many domains, new questions arise about cognitive division of labor, mutual trust calibration, accountability, and the very nature of shared understanding between humans and intelligent agents. Developing novel theoretical models and methodological tools for these hybrid teams will not only redefine the boundaries of collective intelligence but also ensure its relevance in the age of human-machine symbiosis.

Key words: collective intelligence, team cognition, Human-AI collaborative