Validating and Improving a Knowledge Acquisition Approach for Architectural Decisions

Abstract

Software architects make architectural decisions such as choosing architecture patterns or frameworks. Capturing architectural decisions reduces evolution costs for software systems. Existing approaches overlook the challenge of capturing tacit knowledge about architectural decisions. Previously, we proposed the REGAIN approach to capture tacit knowledge about architectural decisions. REGAIN is based on the Repertory Grid technique, a powerful knowledge acquisition approach from knowledge engineering. However, REGAIN lacked industrial validation. Therefore, in this paper, we initially present a study to ensure that REGAIN meets the needs of industrial architects. We interviewed 16 architects who indicated REGAIN advantages such as systematic decision-making support. Also, architects indicated improvement opportunities, in particular tool support and the possibility to prioritize concerns that are used to evaluate decision alternatives. Therefore, we conducted an additional study to evaluate two approaches for prioritizing concerns: pairwise-comparisons and the hundred-dollar approach. We conducted an experiment with 30 graduate students to compare the two prioritization approaches. Based on the results of the experiment, we added the hundred-dollar approach to REGAIN. Moreover, based on the study with architects and the experiment with students, we implemented open source tool support for REGAIN, which architects can use to capture architectural decisions.