Links do Matter: Understanding the Drivers of Developer Interactions in Software Ecosystems

Studies of collaborating individuals engaged in collective enterprises usually focus on the individuals, rather than the links supporting their interaction. Accordingly, large scale software development ecosystems have also been examined primarily in terms of developer engagement. We posit that communication links between developers play a central role in the sustenance and effectiveness of such ecosystems. In this paper, we investigate whether and how developer attributes relate to the importance of the communication channels between them. We present a technique using 2nd order Markov models to extract features of interest of the links and apply the technique on data from a real-world project. Our statistical models - developed on records involving 900+ software developers, exchanging 20,000+ comments, across 500 units of work - offer surprising insights on factors associated with link importance, even after controlling for known effects. These results inform a deeper appreciation of the importance of links in large scale software development along with a number of practical implications.