Stress and Productivity Performance in the Workforce Modelled with Binary Decision Automata

Matthew Page and Daniel Ashlock
Submitted to CIG 2013

Abstract PDF eprint

This study is the third in a series developing an agent based model of stress in the workplace. Stress and stress-related health problems are a serious matter but, prior to this series of studies, quantitative modeling of stress has been substantially neglected. This model builds on earlier work, incorporating a more realistic model of the stress relief caused by weekend time off. The model also makes drug use something that can be learned spontaneously or learned from a mentor rather than being present in an endemic, fixed fraction of the population. In this study a parameter exploration is performed on the agent representation, binary decision automata. It is found that the BDA representation is highly adaptive, responding robustly to parameter changes. Parameters investigated include number of agent states, accuracy of imitation of mentors, work requirements, and probabilities of learned and spontaneous drug use. Parameter values are taken beyond reasonable ranges to examine the model’s failure modes. This study demonstrates that the model behaves in a reasonable fashion, determines its limits, and established a baseline for further investigation.