A New Approach Towards Ensuring Gender Inclusive SE Job Advertisements

A majority of software engineers are male, perhaps, because of the very way that software engineering (SE) roles are advertised is gender biased. Thus far, only word-based checking tools are available to identify gender biases (e.g., “analyst” is considered a masculine word). However, such word-based analyses end up identifying the skills required for SE job positions as masculine words, and therefore, not sufficient. In this work, we present a more nu-anced mechanism to check for gender bias in SE job advertisements by building on the GenderMag method, which has proven to be successful in gender bias detection in software interfaces. From a survey of 44 software practitioners, we identified 16 factors where male and female participants differ and based on a thematic analysis we derived three SE job applicant persona facets. We verified the facets with a small survey where SE candidates related to the descriptions of those factors. We conducted a pilot study using these facets to evaluate four SE job advertisements and identified gender related biases in two of those. The software engineering (SE) workforce is dominated by male employees. This is potentially because SE job advertisements are gender biased toward men. SE job advertisements are often checked for potential gender bias using word-based bias detection tools. However, these do not take software engineering specific words into consideration. For example the word’ analyst’ is considered male using the word-based tools, and therefore any SE analyst position would be seen (incorrectly) to be biased toward men. As a result, such word-based tools are not good enough to accurately assess for gender bias in SE job ads. We present a more nuanced mechanism to check for gender bias in SE job advertisements by building on the GenderMag method. GenderMag has proven to be successful in gender bias detection in software interfaces. This method identifies five dimensions of problem solving (referred to as facets) on which male and female software users differ. From a survey of software practitioners, we identified factors where male and female SE job applicants differ. We grouped the similar factors together and derived three facets that were important to predict job application behaviour of SE job applicants. We validated these facets with some SE candidates and used them to develop ‘new’ Tim and Abi personas from the facets and evaluated four SE job advertisements. We identified gender related biases in two of these.