Simulating the actions and decisions of occupants in buildings is necessary to evaluate energy performance and understand its causes in a more reliable manner. While various models exist, all recent state-of-the-art reviews on occupant behavior models identified unresolved scientific issues. Here, these issues are classified, and we present the SMACH multi-agent model of human activity in buildings (based on cognitive ergonomics) and its application to assessing residential load curves and energy consumption. We show how the model can be used to address each of the unresolved scientific questions and present the validation process that we carried out. Key Innovations This work presents: (i) a condensed summary of currently unresolved scientific issues regarding occupant behavior modeling, (ii) a multi-agent architecture to provide a holistic modeling approach of human activity in buildings, and (iii) a demonstration of how the model can be used to address the scientific issues identified in (i). Practical Implications Occupant activities affect building energy consumptions. It is possible to use multi-agent systems in an effective way to generate more realistic annual consumption data. Simulating realistic activity can help sizing systems by understanding the coincidence of local energy generation and demand.