From Swarms to Hyperswarms: A New Methodology for Amplifying Group Intelligence

Swarm Intelligence (SI) is a biological phenomenon that enables social species to reach group decisions by converging as real-time systems. Studied in fish schools, bird flocks, and bee swarms, SI has been shown to greatly amplify group intelligence. Artificial Swarm Intelligence (ASI) is recently developed method that enables networked human groups, moderated by real-time swarming algorithms to achieve similar benefits. While ASI can significantly boost the accuracy of group forecasts, estimation, and decisions, the process can be brittle if a large majority supports an inaccurate view, even if their average confidence is low. Thus, a major focus of ASI research has been to increase resilience to low-confidence majorities. This paper introduces a new structure called a hyperswarm that enables a confident minority to more easily sway an unsure majority, enabling convergence on solutions that would otherwise be inaccessible. The approach involves dividing a population P into a set of overlapping subgroups (H1, H2 … HP) such that each member is only exposed to the real-time behaviors of their subgroup. And because each subgroup overlaps multiple other subgroups, interactions can quickly propagate across the full population. In this paper we simulate hyperswarm dynamics, showing that a confident minority can sway a less confident majority in real time, even when more than 65% of the participants initially harbor the majority view, so long as the minority has higher confidence. In addition, we derive theoretical guidelines to predict hyperswarm success based on population size, subgroup size, and confidence differential.