The emergence of Autonomous Vehicles (AVs) has prompted extensive research into their potential to change transportation systems by improving efficiency, reducing environmental impact, and enhancing traffic safety. However, concerns remain regarding the capabilities of replicating human behavior as machines do not have the intuition that is intrinsic in human drivers. This paper investigates several different applications of modeling and understanding road user behavior in traffic conflict scenarios using Inverse Reinforcement Learning (IRL). This study explores two distinct IRL approaches: a single-agent Gaussian Process IRL and a multi-agent Adversarial IRL. Utilizing diverse datasets, including cyclists, pedestrians, vehicles, and motorcyclists across heterogeneous urban environments, the multi-agent frameworks have accurately simulated and represented road user behavior. Furthermore, applying the framework in different environments indicated that road user behavior is highly dependent on local conditions, and considering agents of one location in another might be associated with increased risk levels. Insights derived from this research offer crucial implications for the advancement and deployment of AV technologies. Nevertheless, challenges persist, such as the development of intuitive decision-making algorithms, the refinement of collision avoidance systems, and the establishment of societal trust in AV technologies.