Improving the Research Software Review Process of License Scanner Output

Abstract

Software has become an integral part of modern research. In more and more disciplines, researchers develop their own software to conduct their scientific work. This research software rarely stands alone but builds on numerous open-source packages, whose licenses impose legally binding obligations whenever the software is distributed. Since researchers frequently want to share their software alongside their results, they must comply with the licenses of all included packages. As domain experts rather than licensing specialists, they face a demanding task, and the Research Software Engineers (RSEs) who support them are a scarce resource in research organizations. License scanners promise relief by automating the detection of licenses in source code, but their raw output still requires extensive manual review, and existing tools offer little support for this step. This thesis addresses that gap with the License Scanner Review Tool, a prototype of an interactive review layer on top of the OSS Review Toolkit. Grounded in interviews with practitioners experienced in research software license compliance, the tool enriches each license finding with the context reviewers need to make a decision and feeds those decisions back into the automated analysis pipeline. We evaluated the prototype in a usability study with ten participants. It achieved a mean System Usability Scale score of 69.5, and nine out of ten participants agreed that it would reduce their effort compared to their current approach. These results indicate that a dedicated review tool can reduce the effort of license analysis for research software, easing the load on scarce RSEs.