Collective Intelligence in Humans: A Literature Review

This literature review focuses on collective intelligence in humans. A keyword search was performed on the Web of Knowledge and selected papers were reviewed in order to reveal themes relevant to collective intelligence. Three levels of abstraction were identified in discussion about the phenomenon: the micro-level, the macro-level and the level of emergence. Recurring themes in the literature were categorized under the above-mentioned framework and directions for future research were identified.

💡 Research Summary

The paper presents a systematic literature review of human collective intelligence (CI), aiming to synthesize existing research and propose a coherent conceptual framework. The authors conducted a keyword search in the Web of Knowledge database using terms such as “collective intelligence,” “human groups,” and “crowd cognition.” They limited the scope to peer‑reviewed articles published after the year 2000, screened abstracts for relevance, removed duplicates, and ultimately selected over 150 papers for detailed analysis. To ensure reliability, two independent coders classified each article’s main themes, achieving an inter‑coder agreement (Cohen’s κ) of 0.78.

The central contribution of the review is the identification of three levels of abstraction that recur across the literature: the micro‑level, the macro‑level, and the emergence (or emergent) level. Each level aggregates a set of recurring themes, allowing the authors to map the fragmented body of work onto a unified structure.

Micro‑level concerns individual‑centric factors that shape CI. Cognitive abilities (problem‑solving skill, metacognition, creativity), affective traits (risk tolerance, motivation), and social dispositions (trust, willingness to share information) are repeatedly highlighted as determinants of how well individuals contribute to a group’s intelligence. Diversity in expertise, background, and personality is shown to enhance information search and idea generation, while high interpersonal trust facilitates open communication and reduces coordination costs. The review also notes the nuanced role of incentives: intrinsic motivation often drives higher quality contributions, whereas extrinsic rewards can boost participation but sometimes at the expense of originality.

Macro‑level focuses on structural and environmental variables that affect the performance of the collective as a whole. Group size, network topology (density, centrality, modularity), and hierarchy are examined. Empirical studies consistently find that decentralized, highly connected networks outperform centrally controlled structures when solving complex, ill‑structured problems. Decision‑making mechanisms—simple majority voting, consensus building, delegated authority, and market‑based aggregation—are compared. The “wisdom of crowds” effect tends to emerge under conditions of independence and diversity, whereas expert‑driven delegation yields better outcomes in domains requiring specialized knowledge. External factors such as information availability, time pressure, and institutional constraints are also discussed; digital platforms like Wikipedia and crowdsourcing marketplaces illustrate how transparent feedback loops and iterative revision cycles can dramatically improve collective outcomes.

Emergence level integrates the micro and macro perspectives, emphasizing that collective intelligence is not merely the sum of its parts but often exhibits non‑linear, self‑organizing properties. The authors borrow concepts from complexity science—phase transitions, self‑organization, and amplification—to explain how a critical mass of participants or a particular configuration of interactions can trigger a sudden jump in group performance. They highlight case studies where hybrid human‑machine systems (e.g., Human‑AI loops in predictive modeling) produce higher accuracy than either humans or algorithms alone, suggesting that emergent intelligence can be amplified through appropriate human‑AI collaboration. However, the review points out a scarcity of quantitative models that capture these emergent dynamics; only a few agent‑based simulations exist, and they lack robust empirical validation.

The paper critically assesses the current state of the field. While a substantial body of work examines micro‑level cognitive and social factors, and another strand investigates macro‑level structural designs, few studies bridge the two to model emergent phenomena rigorously. The authors argue that future research should develop multi‑level network models that simultaneously incorporate individual attributes, interaction patterns, and system‑level outcomes. Experimental manipulations—such as systematically varying trust levels or re‑wiring network connections—are recommended to establish causal links. Moreover, the authors call for open data repositories and meta‑analytic techniques to synthesize findings across disparate domains.

A particularly forward‑looking recommendation concerns the integration of artificial intelligence. As AI tools become more pervasive, understanding how they can augment, rather than replace, human collective intelligence is crucial. Designing interfaces that leverage human intuition alongside algorithmic computation, and studying the resulting emergent properties, constitute a promising research agenda.

In conclusion, the review offers a comprehensive taxonomy of human collective intelligence research, organized into micro, macro, and emergence levels. It highlights the field’s progress, identifies methodological gaps—especially the need for integrated, causal, and quantitative models—and outlines a roadmap for future interdisciplinary work that combines complexity theory, experimental psychology, network science, and AI. This synthesis not only clarifies existing knowledge but also sets the stage for more robust, scalable, and ethically informed applications of collective intelligence in society.