Collaborative Research Across Disciplinary and Organizational Boundaries Jonathon N. Cummings Massachusetts Institute of Technology and Sara Kiesler Carnegie Mellon University In Press, Social Studies of Science Key Words: scientific collaboration, multidisciplinary, universities, project size, project work, innovation, distributed work, geographic dispersion, organizational boundaries Acknowledgement: This work was supported by NSF Award #IIS-9872996. We thank Allyson Pottmeyer and Maria Ines Garcia for their excellent research assistance throughout the project. We also thank Suzanne Iacono for her helpful research suggestions and critiques of findings.

Abstract Scientific and engineering research increasingly involves multidisciplinary collaboration, sometimes across multiple organizations. Technological advances have made such cross- boundary projects possible, yet they can carry high coordination costs. This study investigated scientific collaboration across disciplinary and university boundaries to understand the need for coordination in these collaborations and how different levels of coordination predicted success. We conducted a study of 62 scientific collaborations supported by a program of the United States National Science Foundation in 1998 and 1999. Projects with principal investigators (PIs) in more disciplines reported as many positive outcomes as did projects involving fewer disciplines. By contrast, multi-university, rather than multidisciplinary, projects were problematic. Projects with PIs from more universities were significantly less well coordinated and reported fewer positive outcomes than projects with PIs from fewer universities. Coordination mechanisms that brought distant researchers together physically slightly reduced the negative impact of collaborations involving multiple universities. We discuss implications for theory, practice, and policy.

2 Collaborative Research Across Disciplinary and Organizational Boundaries Scientists have collaborated with one another for centuries (Finholt & Olson, 1997). Recently, policy makers have begun to encourage and support two or more disciplines working together in applied and basic science ��� multidisciplinary collaboration (Grinter, Herbsleb, & Perry, 1999 Teasley & Wolinsky, 2001 Chin et al. 2002). Important fields such as oceanography and cognitive science have developed out of multidisciplinary collaborations (Hesse et al. 1993 Schunn et al. 2002). Because the formal organization of science and engineering in universities and industrial laboratories usually follows disciplinary boundaries, multidisciplinary collaboration often requires crossing organizational boundaries as well. The geologist who collaborates with a computer scientist often works in another department or university as well as in a different field. In the past, dispersed forms of collaboration would have been made difficult by physical distance between scientists, which not only reduced the likelihood of collaboration, but also had a negative impact on success (Allen, 1977 Kiesler & Cummings, 2002 Kraut et al. 1990). Today, dispersed collaborations are more feasible because communication technologies allow scientists to exchange news, data, reports, equipment, instruments, and other resources (Finholt, 2002 Hesse et al. 1993 Kouzes et al. 1996). Fields such as particle physics and mathematics have relied on computer-mediated communication for several decades (Walsh & Bayma, 1996). A recent explosion in dispersed collaboration has been spawned by funding agencies, such as the National Science Foundation (NSF) in the United States and the Framework Programmes in the European Union, which aim for diverse organizational representation. Recent research suggests that, even with some signs of progress (Sonnenwald, 2003), technology has not yet conquered distance (Cramton, 2001 Herbsleb et al. 2000 Hinds &

3 Bailey, 2003 Mark et al. 1999). A major challenge for dispersed scientific collaborations is coordinating work so that scientists can effectively use one another���s ideas and expertise without frequent face-to-face interaction. Coordination is the integration or linking together of different pieces of a project to accomplish a collective task (Van de Ven et al. 1976). Although some coordination can be accomplished through project structure, for example, by creating clear lines of authority and division of labor, science is dynamic, and members of the collaboration still must talk out common problems, discuss shared resources, and monitor and review the work to make joint progress (Malone & Crowston, 1994 Kraut & Streeter, 1995). Multidisciplinary collaborations also must manage interpersonal relationships within the project. Scientists from different disciplines have usually trained in different departments, have had different advisors, publish in different journals, and attend different conferences. Their social bonds are likely to be comparatively weak (Granovetter, 1973), increasing the difficulty of developing trust and effective interdependence. Innovation in Multidisciplinary Collaborations An important claim favoring multidisciplinary collaborations is that they promote innovation. We define innovation as the successful implementation of creative ideas, tasks, or procedures (Amabile, 1988). In science and engineering, innovations are technical discoveries or insights, new ways to use existing technologies, or radical approaches to problems (Hargadon, 1998 Henderson & Clark, 1990 O'Connor & Rice, 2001 Utterback, 1994). Multidisciplinary projects should increase the likelihood of innovation due to their juxtaposition of ideas, tools, and people from different domains. As the Internet and other forms of computing have enhanced the potential for this ���distributed intelligence���, policy makers in science and engineering expect greater innovation from such projects (Zare, 1997).

4 There is tension between the benefits to innovation of working across disciplinary and organizational boundaries versus the risks that arise from the costs of coordination and relationship development in these collaborations. Dispersed science and engineering projects are forms of innovation systems that are meant to create, diffuse, and use diverse sources of knowledge (Carlsson et al. 2002). How researchers managing such projects and organize work to be productive has been the subject of much discussion over the years (Hagstrom, 1964). Some authors distinguish between the amount of bureaucracy versus participation in the scientific collaboration (Chompalov et al. 2002), whereas others focus on the extent to which work is project-based (Hobday, 2000). The existing literature provides no clear guidelines to managing coordination and relationship development in multidisciplinary collaborations. Multidisciplinary projects may require new approaches to coordination to get the work done and to foster trust. When working with other disciplines requires working across organizational boundaries, as when a biologist at one university collaborates with a computer scientist at another university, the need for coordination increases due to field differences and to geographic dispersion. The research question we pose in this paper is how collaborations involving multidisciplinary and multi-organizational relationships achieve successful coordination. Methods The authors studied a research program created by the Computer and Information Science and Engineering Directorate (CISE) of the United States National Science Foundation (NSF). The program was called ���Knowledge and Distributed Intelligence��� (KDI). Its purpose was ���to span the scientific and engineering communities . . . to generate, model, and represent more complex and cross-disciplinary scientific data from new sources and at enormously varying