There isn’t a lot of gender diversity in software engineering (SE). Most engineers are male. In previous articles I already discussed how this can have a detrimental effect on the quality of communication, can result in designs that might not take the needs of women sufficiently into account and contribute to the gender wage gap.
The gender imbalance in our field may take years or even decades to fix. It’s not enough to get girls to choose STEM subjects at school: they also need to apply for jobs in STEM. This week’s article looks at job advertisements for SE roles, and how we can make sure that they don’t inadvertently discourage female candidates from applying.
According to an Australian report from 2021 as many as 58% of women believe that gender discrimination is present in the tech industry, while 38% of women believe that gender discrimination exists in hiring practices. For instance, the requirements and the language used to describe those requirements are often biased towards male candidates.
Part of the reason why gender bias exists in the first place is that the majority of SE roles are currently performed by males. Therefore, when companies look for new hires, they are more likely to look for traits that are typically associated with men.
There have been attempts to remove gender bias from job advertisements. Most methods are based on word-based language checking, which simply involves counting the number of “male” and “female” words. This approach doesn’t work very well for SE job advertisements however, as much of the terminology within the field is (seen as) masculine.
Some hiring managers ask peers or HR colleagues to read the job advertisement before publishing it, but as far as we know there are no systematic approaches to check for gender biases in job advertisements.
GenderMag (Gender Inclusiveness Magnifier) is a method that uses the cognitive walkthrough technique to evaluate problem-solving software from a gender-inclusiveness perspective.
There are five facts of problem solving, for which there is empirical evidence that they differ between the two genders: motivation, risk averseness, information processing style, computer self-efficacy, and learning style.
The differences that arise from these five facets are captured in three : Abi (with facet values mostly seen in females), Tim (with facet values mostly seen in males), and Pat (who combines facet values for both genders).
The personas are then used to perform cognitive walkthroughs on problem-solving software to identify potential gender biases. In every walkthrough, a so-called evaluator performs actions on the software based on the way they answer questions about those actions, while taking the facet values of each persona into account.
In this study, the GenderMag method is modified so that it can also be used to evaluate SE job advertisements. This is done in several steps.
Reading job advertisements is clearly not the same thing as using problem-solving software, so not all of the GenderMag persona facets can be reused. The researchers conducted a survey among SE job candidates to identify the facets that are related to SE job seeking.
They found 10 factors that can be used to predict SE job application behaviour. The researchers grouped these factors into three facets: SE views, career goals, and job application behaviour.
|Importance to job application
|SE team expectations
|What characteristics they expect in the SE team they will join
|Frustration in SE
|What within SE frustrates them
|Short term career goals
|Where they want to see them in next five years
|Long term career terms
|Where they want to see themselves in 10–15 years
|Perceived challenges in achieving those goals
|Planned measures to overcome the challenges
|Job application behaviour
|Whether they read job advertisements selectively or sequentially
|Most attractive information in jobs advertisements
|Most important information in job advertisements
|What factors do they look for to decide to apply
These facets were analysed and used to develop new personas that take the differences and similarities in thinking processes between the into account.
To validate the newly created job seeker personas and their facets, the researchers conducted a second survey in which they asked respondents whether they relate to Abi, Tim or neither when reading SE job advertisements.
The final versions of the two personas can be found below (reproduced verbatim).
Finally, the researchers took the updated versions of Abi and Tim for a test drive by performing a cognitive walkthrough on four randomly selected SE job advertisements for project manager, programmer, analyst, and tester roles. Their results were compared to those of 10 actual SE job candidates who were asked to review the same job advertisements.
The results show that actual SE job candidates often come to a different conclusion (yes, no or maybe) than Abi and Tim, likely due to personal differences. However, the researchers claim that their (unnamed) version of GenderMag can be successfully used to detect gender bias in SE job advertisements.
I suggest that you try it out yourself and see if it works for you.
GenderMag is a method that uses cognitive walkthroughs with personas to evaluate gender-inclusiveness of software
This study applies the GenderMag method to detect gender bias in software engineering job advertisements
The identified facets and personas can help managers design better SE role requirements and job advertisements