This study contributes to human behaviour (decision rule) modelling in the agent based simulation, by improving the existing data collection methodologies and comparing their benefits. Improving data collection methodologies can help in developing a more realistic agent’s decision rule and increasing the validity and credibility of the final model. This study uses a dairy supply chain case because the actors in this context can have one to one correspondence with the agents in the simulation.
This study begins by presenting a literature review on the applications of agent-based simulation in the agri-food supply chain. This literature review highlights existing agent-based modelling practices in the agri-food supply chain such as the scope of the modelling, data collection, validation and sensitivity analysis techniques. This study then proposes some improvements to the existing data collection methodologies namely questionnaire survey and role-playing game. This study proposes the use of a scenariobased questionnaire to improve the benefits of a questionnaire survey for decision rules calibration. While to extend the usefulness of role-playing game this study propose the use of the design of experiment, and game scaling based on empirical probability distribution.
The improved data collection methods are then used to calibrate a base model that was developed from the previous literature. Primary data from 16 villages in Indonesia is used to elicit empirical decision rules in this calibration process. The result from simulation experiments shows that the improved data collection methods can produce models with higher operational validity. This study is concluded by evaluating the advantages and disadvantages of each data collection methodology.