Beyond Individual Attributes: Network Structure and Learning Perception as Drivers of Collective Intelligence at Scale

Collective intelligence, the emergent capacity of groups to perform cognitive tasks beyond the abilities of individual members, has been extensively studied in small-group laboratory settings but remains underexplored in large-scale naturalistic educational environments. This study investigates factors influencing collective intelligence among 5,347 learners across 25 class-level groups within a blended university course. Using a crowdsourced question bank activity as the collective task, we operationalized collective intelligence through four behavioral dimensions aligned with McGrath's group task circumplex theory. These dimensions included generation, negotiation, execution, and assessment. A two-stage structural equation modeling approach was first employed to validate the collective intelligence measurement model, and standardized composite scores were computed using empirically derived factor loadings. Five categories of antecedents were then examined through correlation analyses and structural equation modeling, including demographic composition, social interaction network characteristics, course organizational features, learning perception, and social presence. Results indicated that social interaction network characteristics were the strongest predictor of collective intelligence with a standardized path coefficient of 0.564, followed by learning perception with 0.282. Social presence exhibited a non-significant negative association with collective intelligence, suggesting that heightened interpersonal awareness may introduce cognitive interference in task-oriented collective activities. Demographic composition and course organizational arrangements did not significantly predict collective intelligence after accounting for other predictors. These findings highlight the primacy of group-level interaction processes over individual-level attributes in shaping collective intelligence and offer practical guidance for designing blended learning environments that support emergent collective cognition at scale.