Railway system design in the light of human and machine unreliability

In modern railway systems there is an increasing use of automation and artificial intelligence, but humans are retained as intelligent agents. This means that safety depends on the reliability of both human and non-human agents. The traditional approach to safety involves a probabilistic estimate of reliability. We suggest a new approach which supports the design of safe systems by the use of causal models of unreliability. This approach provides strong support for the analysis of the interaction of human and non-human intelligent agents in automated railway systems