Decision support systems perform complex computations to provide suggestions on decision making and problem solving. Recent analyses reveal limited user acceptance of such systems. One reason is that these systems work as a black box so that the users cannot fully understand how the solutions were derived from these systems.;To address this issue, three types of more advanced explanations were identified (contribution 2). A comprehensive methodology was proposed to apply three machine learning techniques for generating these explanations from software release planning project data, and to utilize the generated explanations in justifying existing solutions and providing guidance on backward planning for generating solutions that potentially achieve predefined target values in the objective space (contribution 3).;As a proof of concept, a prototype tool called EXPLAINER for automating the tasks proposed in the methodology was developed (contribution 4). Two case studies were conducted on a real world release planning project to illustrate and evaluate the application of the advanced explanations and the related methodology (contribution 5).;A first series of replicated experiments was conducted to evaluate explanations as a part of decision support for software release planning (contribution 1). In particular, manual planning versus black box versus white box usage of ReleasePlanner(TM), an intelligent decision support system for software release planning, were investigated. A major finding from the empirical investigations was that existing explanations provided to the tool (white box view when using the tool) were insufficient to increase user trust in the tool and confidence in the solutions produced by the tool.
展开▼
