Empirical Validation of a Multilevel Cognitive Architecture for Collaborative Software Development

Abstract

Modern software development represents a form of collective intellectual activity in which many project decisions emerge through interactions among team members. This study presents an empirical analysis of changes in developers’ representations that occur during collaborative discussions of requirements, architectural decisions, and implementation strategies. The study is grounded in a multilevel cognitive architecture of collaborative software development that links developers’ cognitive processes, project artifacts, and the phases of the Software Development Life Cycle (SDLC). The empirical investigation is based on observations of five software development teams and the analysis of 20 episodes of collaborative interaction. The analysis identified 42 cognitive transitions, defined as changes in team members’ understanding of design and implementation decisions, traced through team communication and project artifacts. To systematize the observed changes, a typology of cognitive transitions is proposed, including alignment, refinement, reconstruction, and reinforcement of representations. The results show that the distribution of transition types varies across different phases of the development process and depends on the tasks and roles of team members: alignment and refinement dominate during requirements engineering and design, while reconstruction and reinforcement become more prominent during implementation and testing. These findings refine the dynamic aspect of the proposed cognitive architecture and provide empirical validation of its core assumptions about cognitive change in software development teams. They suggest that the software development process can be understood as a sequence of cognitive transitions emerging through collaborative interaction and lay the groundwork for future research and tool support for collaborative software engineering.