Adolescent’s Collective Intelligence: Empirical Evidence in Real and Online Classmates Groups

Abstract

Humans create teams to be more successful in a large variety of tasks. Groups are characterized by an emergent property called collective intelligence, which leads them to be smarter than single individuals. Previous studies proved that collective intelligence characterizes both real and online environments, focusing on adults’ performances. In this work, we explored which factors promote group success in an offline and online logical task with adolescents. Five hundred and fifty high school students participated in the experiment and faced Raven’s Advanced Progressive Matrices, first by themselves using the computer, then in a group. Groups interactions could have been computer-mediated or face-to-face, and the participants were randomly assigned to one of the two experimental conditions. Results suggest that groups perform better than singles, regardless of the experimental condition. Among adolescents, online groups performance was negatively affected by participants’ average perception of group cohesion, the difficulty of the problem, and the number of communicative exchanges that occur in the interaction. On the contrary, the factors that improve their performances were the average intelligence of the teammates, their levels of neuroticism, and the group heterogeneity in terms of social abilities. This work contributes to the literature with a comprehensive model of collective intelligence among young people.

g versus c: comparing individual and collective intelligence across two meta-analyses

“…Yet the g- factor, which is thought to be situationally robust (e.g., Schmidt and Hunter 1981 1998 ) and highly stable across the human lifespan (e.g., Deary et al 2013 ), should, in principle, retain certain aspects of these characteristics at the group level of analysis such that the g- factor, even when averaged across group members, shares a significant relationship with true G g (see Kozlowski and Klein 2000 , for arguments in support of this view). In addition to Bates and Gupta ( 2017 ), another study that directly tested and supported this hypothesis was conducted by Imbimbo et al ( 2020 ) who asked 550 high-school students to complete alternative (odd/even) sets of Raven’s Advanced Progressive Matrices, first as individuals, then as 110 randomly allocated groups comprised of five members each. A generalized linear mixed model revealed a strong relationship between the probability function of selecting the correct answer at the individual level and selecting the correct answer at the group level using a majority vote method.…”