Abstract: In 1975, the Atomic Energy Commission initiated the landmark Rasmussen study that led to the advent of Probabilistic Risk Assessment (PRA) in the nuclear industry. Over the years, PRA has grown into a well-established technical discipline with a wide range of applications in diverse industries for the design, operation, and regulatory oversight of complex technological systems. In both the research and the application of PRA, the nuclear industry continues to be the leader. Last year's disaster at the Japanese Fukushima Daiichi Power Plant is an unfortunate reminder of the vital importance of comprehensive PRA in the nuclear industry. Classical PRA primarily uses statistical data and expert opinions to estimate the probabilities of basic events in Fault Trees (FTs) and Event Trees (ETs) in order to calculate system-wide risks. Recently, and due to a broader role envisioned for PRA in various risk-informed applications, there has been a gradual shift in interest towards modeling the underlying failure mechanisms of the elements of risk scenarios. As a result, PRA is expanding its horizons into such fields as reliability engineering, cognitive psychology, organizational behavior, and computer science. This talk will describe the two emerging causal extensions of PRA models suitable for constructing a hybrid PRA. The first extension results from a study on the incorporation of human, social and organizational failure mechanisms into PRA. The proposed modeling techniques and theories are applied for maintenance safety. The second extension relates to a research study on the integration of physical failure mechanisms into PRA and its application for nuclear power safety and for the advancement of the treatment of dependent failures (i.e. Common Cause Failures). These two lines of research provide some of the building blocks for developing a fully hybrid PRA capable of covering the dynamic interactions of diverse risk-contributing factors. The talk will offer paradigms of the state-of-the-art combination of deterministic and probabilistic modeling techniques for the incorporation of underlying failure phenomena into risk assessments of complex systems.