Dynamic model and crowd entropy measurement of crowd intelligence system
杰 罗, 鑫 姜, 炳晖 郭, 宏威 郑, 文峻 吴, 卫锋 吕 · Scientia Sinica Informationis · 2021
The phenomenon of crowd intelligence widely existing in nature has attracted extensive attention from researchers worldwide. Along with the fast development of network and artificial intelligence techniques, it enables that the crowd intelligence can be formed by stimulating the intelligence of individuals and converging the intelligence of the crowd. As a result, how to recognize and construct such a crowd intelligence system becomes a hot topic. This paper explores the theory and method for measuring crowd intelligent systems, so as to enrich the understanding of forming mechanisms for crowd intelligence systems. We propose the basic properties of the crowd intelligence system and the incentive-convergence model for its forming. Actually, the crowd intelligence system is considered to be a complex nonlinear dynamic system, which includes three core dynamic attributes: micro individual excitation, macro group collaboration, and global group intelligence convergence. Based on this model, we demonstrate the basic properties of crowd entropy which is a measurement for the crowd intelligence system. Finally, we perform a case study on crowd-based graph search and demonstrate the incentive and convergence processes of the crowd. During these two processes, the crowd entropy effectively reflects the change of crowd intelligence behaviors, which verifies the effectiveness of crowd entropy as the metric for crowd intelligence behaviors.