A Simple Computational Theory of General Collective Intelligence

• Published in: Topics in cognitive science 2019-04, Vol.11 (2), p.374-392
• Main Author: Krafft, Peter M.
• Format: Article
• Language: English
• Subjects: Collective agency; Collective intelligence; Collective rationality; Computational social science; Computational theory; Intelligence; Multiagent systems
• ISSN: 1756-8757; 1756-8765
• DOI: 10.1111/tops.12341

Researchers have recently demonstrated that group performance across tasks tends to be correlated, motivating the use of a single metric for the general collective intelligence of groups akin to general intelligence metrics for individuals. High general collective intelligence is achieved when a group performs well across a wide variety of tasks. A number of factors have been shown to be predictive of general collective intelligence, but there is sparse formal theory explaining the presence of correlations across tasks, betraying a fundamental gap in our understanding of what general collective intelligence is measuring. Here, we formally argue that general collective intelligence arises from groups achieving commitment to group goals, accurate shared beliefs, and coordinated actions. We then argue for the existence of generic mechanisms that help groups achieve these cognitive alignment conditions. The presence or absence of such mechanisms can potentially explain observed correlations in group performance across tasks. Under our view, general collective intelligence can be conceived as measuring group performance on classes of tasks that have particular combinations of cognitive alignment requirements. What are the conditions under which groups of agents will perform well across multiple tasks? The author establishes a set of “alignment conditions” that enforce identity of beliefs and desires across agents. These conditions are necessary and sufficient for ensuring that a multiagent system behaves as if controlled by a rational centralized controller. Several widely observed social phenomena can be understood as facilitating the alignment conditions.