Quality Assurance in Agile Safety-Critical Systems Development

In this position paper we examine how safety could be assured when increasingly complex systems are developed using agile software development methods. We first discuss the source and nature of complexity in software systems and how a probe - sense - learn approach recommended by the Cynefin Framework is appropriate for designing complex systems and a sense - analyse - learn approach is appropriate for developing a complicated system whose design has been determined. We then examine how quality assurance is incorporated into agile software development before pointing out that those characteristics of a self-managed team that produce so many benefits for software development of complex systems whose solution evolves with problem understanding, are also vulnerable to confirmation bias. This suggests that for safety critical system development, software systems developed by agile teams will need verification and validation by independent parties. We review current quality management practices for medical device software development before discussing how our earlier findings could be adopted into safety critical software quality management.