Abstract

Advances in machine learning hold promise for corpus analysis: they have the potential to allow for more efficient and less biased analyses of text. This would be a boon for qualitative research, such as the survey research conducted by Academic Philosophy Data and Analysis. In this paper we examine the utility of automated machine learning for select survey questions, with a focus on LDA and VADER. We thus compare human and machine coding on the question of whether underrepresented philosophers are more likely to respond negatively to questions concerning diversity and inclusivity in academic philosophy. Our study has mixed results, revealing instances where automated machine learning can be used to fruitfully support and interrogate more traditional hand-coding techniques, but also instances where we were unable to glean much insight. (Contact authors for draft.)