In the face of the vast scale of software-intensive systems needed today, modern development environments fail dramatically, typically leading to information overload, an inability to deal with the highly dynamic nature of both the systems and the organizations that develop them, and failure to support collaboration across organizational boundaries. The overarching aim of this project is to provide a scientific foundation for human-centered environments that make large-scale and distributed project awareness, communication, and coordination as effortless as in a small team. It accomplishes this by (a) performing empirical studies of real-world large-scale high-complexity software projects to understand how task coordination occurs in and contributes to organizational context, (b) developing an underlying theory of coordination in context, which will motivate and guide (c) the design of new coordination technology that explicitly addresses information overload, dynamism, and organizational boundaries.

Intellectual merit: The research will result in four contributions: (a) a sound theoretical basis that captures how task coordination and organizational context interplay at scale; (b) theory-driven empirical studies of in-context coordination; (c) knowledge about how to achieve improvements in productivity, quality, and development speed; and (d) a suite of design principles, tool prototypes, and interaction techniques for collaboration at a very large scale. These outcomes will transform the landscape of coordination technology by squarely addressing the issue of scale, moving from coordination within a team to coordination across many developers, across many teams, and across multiple geographical and organizational boundaries.

Broader Impacts: As society enters the era of "ultra large scale" software-intensive systems, coordination at such scales is a major unsolved problem, persistently hampering development and advances in vital domains such as healthcare, security, defense, eGovernment, and energy. The outcomes of this project will not only provide major economic benefits, but also major societal benefits in the form of the new systems that now can be developed. Through close collaboration with industry partners, the results will quickly find their way into practice. The project will also increase involvement of women in computer science through workshops and mentoring activities.

Project Report

Groups are increasingly used in organizations to accomplish tasks. Argote, Aven and Kush examined factors affecting group performance, with particular attention to transactive memory systems and social networks. A transactive memory system (TMS) is a shared understanding of who knows what within a group. The effects of centralized and decentralized communication networks and member turnover on transactive memory system development and group performance were investigated in a laboratory study. An experimental platform was developed in which groups performed a realistic collaborative computing task. An advantage of the experimental design is that it enables one to make causal statements about the effects of variables. Both network structure and the interaction between network structure and turnover were found to affect transactive memory system development. Decentralized groups had stronger transactive memory systems than centralized groups when turnover did not occur while the transactive memory systems of centralized groups were not affected by turnover. Further, decentralized groups performed better when there was no turnover while centralized groups performed better when turnover occurred. Transactive memory systems accounted for the effect of network structure on performance when groups did not experience turnover but not when they did. Our results, which indicate that network structures and turnover have powerful effects on transactive memory systems and group performance, have important implications for organizations. In dynamic environments with high degrees of member turnover, organizations would benefit from encouraging centralized communication structures. By contrast, when attrition is low and organizational group membership is stable, decentralized communication networks are preferable.

National Science Foundation (NSF)
Division of Information and Intelligent Systems (IIS)
Standard Grant (Standard)
Application #
Program Officer
Kevin Crowston
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
Carnegie-Mellon University
United States
Zip Code