Effects of Developers’ Networks on Software Evolution

My research focus is to understand how social interactions between project participants affect project success. Large projects involve many participants working on different parts of the project which make it prone to failures in communication, as recent technical issues with the new healthcare website illustrate. Using data from a large, software project, we studied interactions among software developers and using network analysis techniques examined how that impacted the resultant product. We compared the structure of the software with the network of interactions among the developers, tracking their activities over 10 years. Our findings indicate that only 20% of the developers were actively involved in developing key pieces of the package. Furthermore, even if people were working on the same part of the software, they rarely communicated and finally errors were only resolved after a long-time, generally as part of a major revision. These findings are detrimental for long-term project success. For example, if even a few of the key developers leave the project, it will come to a standstill. Similarly, delayed resolving of bugs will discourage users from utilizing the package and finally poor communication between developers lead to slower evolution and higher maintenance costs. Our principal outcomes are identifying factors in a collaboration that can be used to inform risk management and mitigation strategies. We developed a toolset that can provide similar analysis for other collaborative projects (not necessarily software). The recent results of this project are available at http://loki.ist.unomaha.edu/~cbharris/index.ph