The Importance of Representing Cognitive Processes in Multi-agent Models