2024/05/28 by Johannes B. Gruber, Mariken Anna Catharina Geertruida van der Velden · 1 voice
Decision Sciences · #Scientific Computing and Data Management
paper · pdf · doi:10.31219/osf.io/a3kwf
openalex publication_date 2024/05/28 · openalex created_date 2024/05/29 · openalex updated_date 2026/07/14
In the past decade, the social sciences have undergone a transformation, embracing computational methods with increased data and tool availability. However, the adoption of these methods remains uneven, rooted in the discipline’s discrepancies between dominant and marginalized groups. Studies reveal that underrepresented groups are less likely to use computational methods, attributing it tolimited methodological and technical guidance and documentation. To address that shortage, this paper proposes an approach to software documentation that employs readily available Large Language Models (LLMs). Rather than just urging (academic) software developers, who are already shouldering the brunt of the work in bringing the tools and approaches to the field, to spend even more time and effort on their tools, we demonstrate that LLMs like GPT-4 can help produce (more) inclusive documentation with minimal effort. Specifically, the approach is exemplified through the documentation of the R sentitopics and cookiemonster packages. By promoting inclusive computational documentation tools, the aim is to bridge existing gaps and ensure equal opportunities for researchers in the field of social sciences, while not slowing down software development or discouraging researchers from packaging their code as tools with further demands.