Influence Process Structural Learning and the Emergence of Collective Intelligence

Abstract

Recent work [Hazy 2012] has demonstrated computationally that collectives that are organized into networks which govern the flow of resources can learn to recognize newly emerging opportunities distributed in the environment. This paper argues that the system does this through a process analogous to neural network learning with relative status playing the role of synaptic weights. Hazy showed computationally that learning of this type can occur even when resource allocation decision makers have no direct visibility into the environment, have no direct understanding of the opportunity, and are not involved in their exploitation except to the extent that they evaluate the success or failure of funded projects. Effectively, the system of interactions learns which individuals have the best access to information and other resources within the ecosystem. Hazy [2012] calls this previously unidentified emergence phenomenon: Influence Process Structural Learning (IPSL). In the prior model of IPSL, a three-tiered organizational structure was predetermined in the model design [Hazy 2012]. These initial conditions delimit the extent to which the emergence of collective intelligence can be posited because the model itself assumes a defined structure. This work contributes to the field by extending the IPSL argument for collective intelligence to a holistic emergence argument. It begins by briefly reviewing previously published work. It continues the conversation by adding two additional steps: Firstly, it shows how a three-tier organizing structure might emerge through known complexity mechanisms. In this case the mechanism identified is preferential attachment [Barabasi 2002]. Secondly, the paper shows how collective intelligence can emerge within a system of agents when the influence structure among these agents is treated as a the genetic algorithm.