Towards gender-inclusive job postings: A data-driven comparison of augmented writing technologies

Job advertisements are often worded in ways that might pose discrimination risks leading to the exclusion of certain groups of applicants, particularly in relation to their gender. Especially in male-dominated professions or leadership roles, the specific linguistic formulation of job postings acquires relevance if more women are to be attracted to apply. Various technologies have emerged that offer automated text screening, some of them even suggesting alternative formulations to increase gender inclusivity. In this study we analyze four software providers on the German market using a corpus of ∼160, 000 job ads from three different platforms. We identify the relevant social psychological research on gender and language that is at the scientific core of these technologies. We show that, despite sharing a common foundation, the four tools assess the potential for exclusion in job postings in a considerably divergent way on multiple levels of comparison. We discuss the levers in the software pipeline of all four technologies, as well as the potential effect of certain implementation decisions, such as string-based vs. semantic approaches to computational processing of natural language. We argue that the 'technological translation' of research is extremely involved and further studies of its use in practice are needed to assess the potential for more gender equality.