The Causality of Bug Resolution: A Tale of Two Systems

Timely and effective bug resolution is a central concern of software development; more so, for systems that serve needs of real-world users. A large body of software engineering research literature involves studies on bug resolution processes. Many of these studies are correlational in nature, as they analyze historical data to uncover factors influencing bug resolution times. Understanding what causes software bugs to be resolved faster is of critical interest to developers, customers, and end users. Randomized controlled trials (RCT) – widely used in establishing causal effects in other areas such as medical research – are not feasible in a real-world software development scenario. Thus causal inference techniques are essential in such settings to isolate the influences on bug resolution times. In this paper we propose a model of factors affecting bug resolution. The model is validated on two real-world datasets, leading us to identify key determinants of bug resolution times. Our results offer valuable insights for individual developers, project managers, and development organizations.