Simulating Collaborative Innovation in Volunteer Groups: A PLS-Agent Based Model with Multiple Aggregation Levels

Abstract

The lack of analysis of complex phenomena and results emerging from volunteer events organized by public entities leads to very limited understanding of how these collaborative events innovate public services or solutions to social problems. With this chapter, we contribute to expand this understanding by proposing a method to simulate how participants’ perceptions of their collaborative groups outcomes form. Using the PLS-agent methodology over a sample of original data (n = 219), we have been able to simulate the discussion rounds in work groups of volunteer events in the LivLab Simulator project. The method departs from a partial least squares on structural equations model (PLS-SEM) that aggregates data into three hierarchical levels. The PLS path model allows to calculate each criteria scores per simulation run and per agents’ interaction. This method simplifies the architecture of the artificial agents, the modeling of their environment and relationships between the criteria, and clarifies the agent model outputs visualization. This is important because the resulting longitudinal data provide ways to answer questions at the micro level (e.g., how an event should be organized if the objective is to improve relations in a community), and facilitate others at the meso level (e.g., how to improve public innovation through co-creation with citizens). Keywords: Artificial agents, social problems, Simulator project, public services