Traditional research in construction safety focused on accident data analysis, identification of root cause factors and safety climate modeling. These research efforts did not study the dynamic repetitive interaction among multiple root cause factors. Recently safety research focuses on developing accident causation models. These models attempt to explain how the interaction among multiple project factors gets translated into safety incidents. Agent based modeling and simulation (ABMS) is an appropriate technique to develop computational models of accidents causation because of its ability to model human factors and repetitive decentralized interactions. This paper presents a conceptual framework for developing agent based models of construction safety. The key components of the model such as construction crew, management, work environment, material and equipment related quality issues, project/process level complexities, interaction rules and adaptation have been discussed. Further the adaptation of agents in response to the safety culture has been demonstrated using a simple ABMS experiment.