Increasing Complexity of Agent-Based Models with Experiments – the Case of Structural Shirking

The paper explores possibilities of using experimental data to introduce additional complexity into agent-based models. The model under development was the model of structural shirking 1.0 - THE CORE, in which the manager distributes work chores among employees, who are set to complete assigned work tasks on a deadline. The manager sets the deadline on the basis of her perceptions of task difficulty and employee competence level. The employee, who knows his competences better, has an informational advantage over the manager (i.e. adverse selection). Yet, reality may outsmart both actors, who do not know the true level of task difficulty (i.e. reality fluke). The unintended consequences of the work process include shirking and overworking. Both aberrations from manager’s expectations have two dimensions: qualitative (insufficient effort and working too hard), and quantitative (insufficient working time and working too much). As the model specification and simulation results proved theoretically and empirically useful, we decided to introduce interactions among employees and discover the extent to which they influence the emergent level of shirking in the organization. The first experiment was devised in order to validate assumptions regarding learning functions of employees implemented in the core model. In addition, it provided information on descriptive statistics of main variables of interest and stimuli, which were used in research design of the subsequent stage of the study. The second experiment tested the degree to which peer influence and self-efficacy in task performance impact qualitative and quantitative shirking/overworking. Results of the second experiment were used to inform behavioural rules of agents in the model of structural shirking 2.0 - PEER INFLUENCE.