Flight accidents with modern aircraft are often a result of complex dynamics of the “pilot (automaton1) - vehicle - operational environment” system. When a “critical mass” of the system’s complexity exceeds a certain level, a “chain reaction” of irreversible cause-and-effect links can be spontaneously triggered in the system behavior leading to a catastrophe. An affordable, practically tested technique is proposed to complement current methods of flight accident analysis. A generic situational model of the system behavior and a computer are employed as a virtual test article. This model includes a six-degree-of-freedom non-linear flight dynamics model, a generic situational pilot model (“silicon pilot”), models of anticipated operational factors (conditions), and a tool for flight scenario planning. Available flight recorder data are used to tune the model and reconstruct the accident. Then the model is used for in-depth examination of the accident’s “neighborhood” in autonomous “what-if” simulation experiments under actual and hypothetical conditions. The latter may include pilot errors, piloting tactics variations, onboard system’s failures and errors, and weather conditions, as well as combinations of these factors. Programming and piloting skills are not mandatory for the user. Potential applications include: flight accident investigation under uncertainty, advanced pilot training, research into aircraft practical aerodynamics, design of automatic flight control systems, and onboard AI technologies for flight safety.