In this dissertation, we focused on software product uncertainties, such as modeled for CEquencer software, though process uncertainties were treated briefly (e.g., path selection criteria in Table ). Despite the emphasis on artifact uncertainties, we firmly believe that process uncertainties should also be modeled. We contend that software-process modeling formalisms must be augmented to include uncertainty values; that an environment for supporting definition and execution of process models should include capabilities for representation and interpretation of belief values and should allow for Bayesian updating of those values; and that Bayesian updating procedures must be carried out during process execution, such that belief values and confidence levels are continuously updated as new evidence arrives.
The provision and update of belief values may be greatly enhanced in software process frameworks that include process measurement capabilities. Such capabilities constitute a rich source of information regarding the current state of various elements and support the collection of statistical and empirical data that may significantly improve the accuracy of prior belief value estimation.
We expect that by modeling software process uncertainties, one may achieve a more realistic representation of the process, enable automated belief revision by means of Bayesian updating, and support prediction and guidance of future development activities.