By Rob Mitchum // May 18, 2015
Academic research, for all its aspirations of collaboration and openness, remains a very closed community. Data and software are rarely shared, findings are published in static, paywalled journal articles, and collaborations are typically “invite-only.” But as more researchers turn to computational methods to power their work, many are looking to the culture of software programming as a potential model for a more open world of science. In his talk at the Computation Institute on May 15, GitHub’s Arfon Smith presented several lessons that academia can borrow from the thriving world of open source software development.
GitHub is one of the pillars of the open source community, providing a public space for programmers to share their code and work together on projects. It’s culture is inherently collaborative — users can “fork” any public project, receiving a copy of the source code that they can then modify for their own purposes, or share back with the project owners for potential integration. These practices create a society where anyone can join a collaboration, frequent contributors receive public credit, and decision-making about improving software takes place in an open forum.
Smith, a former astronomer, argued that all of these characteristics would be just as valuable for science as they are for software. Instead of closed collaborations that share code and make decisions over e-mail, an open academic collaboration promises a more inclusive, low friction way of pushing research forward, as researchers can work in parallel on code and publicly discuss new additions and modifications. Such a platform could also create a new “culture of reuse” in science, where absorbing and reproducing methods can be freed from the shackles of traditional journal articles and new metrics on usage and quality can provide more appropriate credit for the best, most popular scientific software and its creators.
Promoting these practices will take more than just a “GitHub for science,” Smith said, though efforts such as figshare, Dataverse, and Plotly have made promising first steps. In many cases, open source science requires serious cultural shifts in how researchers approach their work, including changes in how credit is awarded, the trust between researchers, and the discoverability of software tools. But by moving away from the scientific literature model that has dominated research for hundreds of years, Smith feels that open source approaches can actually bring research back to its true first principles.
“The whole point of doing research and sharing our work is to explain what you did so that others can repeat, with the idea being that everybody learns,” Smith said. “I think open source does this much, much better.”
You can watch Smith’s full talk here (fast forward to 3:00 to skip the setup).