This dissertation offers a path that, if followed, helps address software development problems for which uncertainty plays an important role, specifically with respect to human judgment and decision making. We strongly believe that software uncertainty modeling - using Bayesian belief networks - will help managers to better plan, schedule, estimate, and evaluate software projects. It will also help developers to better trace and navigate interrelated artifacts in large software spaces, understand and discover system qualities and properties, and, ultimately, achieve better judgment and make better decisions.