Modelling human behaviour at work: an agent-based simulation to sopport the configuration of work teams

Abstract

The modelling and implementation of human behaviour into artificial systems have attracted the interest of several people from different disciplines given the natural complexity and multidisciplinary of this phenomenon. This type of models offers a great potential in a large set of domains including military applications, training and learning, e-health, and the representation of crisis and emergency situations among others. A particular interesting application is the modelling of human behaviour within groups. More specifically, the model of human behaviour in a work team would be used to analyse the specific dynamics behind individual and work team performance. The analysis of these dynamics then can be used as additional and useful information that supports the decision-making in the formation and configuration of real work teams in companies. The work described in this thesis presents a novel agent-based simulation model that aims to support this decision-making process. This global objective is achieved through the modelling of the team candidates as software agents forming a virtual team. The configured virtual team then can be used to experiment with the internal and contextual factors that produce and direct the individual and work team performance. The real team candidates are modelled through a set of selected human characteristics that have been proved to influence human behaviour when making work activities in a work team. These characteristics include cognitive capabilities, emotional state, personality styles and social-related skills. Using the values in these attributes in combination with values represented in the attributes of project’s tasks, the model produces statistical information that represents the possible performance of the teammembers over the assigned project. The theoretical background of the model is presented through a deep review of existing (mainly psychological, sociological and organisational) theories focused in the study of human behaviour within the particular context of teams. The work presented in this thesis uses the findings from this theoretical research to implement the main characteristics of the proposed agent-based model and to further develop a usable software simulation tool. This tool has been used to validate the proposed model through the comparison of the work team’s performance results generated by the simulation tool against the performance evaluation results of a real work team from a large company. The validation results and the identified open issues to improve the model are also presented in the final part of this thesis document.