The PLS agent : agent behavior validation by partial least squares

Abstract

Agent-based modeling is widely applied in the social sciences. However, the validation of agent behavior is challenging and identified as one of the shortcomings in the field. Methods are required to establish empirical links and support the implementation of valid agent models. This paper contributes to this, by introducing the PLS agent concept. This approach shows a way to transfer results about causalities and decision criteria from empirical surveys into an agent-based decision model, through processing the output of a PLS-SEM model. This should simplify and foster the use of empirical results in agent-based simulation and support collaborative studies over the disciplines.