Abstract of Meeting Paper
A Stochastic Model of Human Error During Software
Development. M. Stutzke, M. Agarwal, and C. Smidts,
University of Maryland at College Park
Existing software reliability models estimate the software
failure intensity function during integration and acceptance
testing. While these models are useful engineering tools, they
cannot be applied to earlier lifecycle phases (e.g., requirements
analysis, preliminary and detailed design, coding) and do not
provide insight into the nature of human errors that occur during
software development. Two broad categories of human error occur
during software development:
- development errors made during design and coding
activities, and
- debugging errors made during attempts to remove faults
identified during software inspections and dynamic
testing. Based on Markovian methods, a stochastic model
has been developed that relates the software failure
intensity function to development and debugging error
occurrence throughout all software lifecycle phases.
Various influencing factors (e.g., developer education,
experience, domain expertise) and the software
development schedule are used to predict human error
rates. The model’s results generally agree with
observed data.