Exploration of Discriminatory Language in Software Artifacts and Developer Communications

Discriminatory terminology has been identified as a potential contributing factor to low diversity in the software developer population. Terms including whitelist, slave, and dummy value have been commonplace in the technical vernacular of software developers for decades. There has been significant movement in the software industry within the last ten years to deprecate such terminology in software. The purpose of this thesis is to explore the impact and usage of discriminatory language in software development. A collection of open-source software projects were analyzed for this terminology from GitHub. Developer communications were also collected from Gitter. Questions from Stack Overflow were collected from the Stack Exchange database from a ten year period between 2010 and 2020 for the purpose of examining the trend of the terminology in question. Finally, a survey was distributed to professional software developers and computer science students to collect information on the impact this terminology may have on individual and team performance. The survey found that gender and ethnic minorities express a higher sensitivity to the identified terminology. Most demographics surveyed experienced a below average negative workplace impact within their organization due to the usage of discriminatory terminology. This experience varies by demographic, with female respondents stating a negative impact on self-esteem and Hispanic and Latinx respondents stating a negative impact on diversity. Analysis of Gitter channels and project source code found that most projects had a lower number of occurrences of the identified terminology in the source code than in the Gitter chat channels. The Stack Overflow analysis found usage of whitelist/blacklist has steadily risen in questions over the last decade, and tend to be associated with web-based technologies and Android applications. Usage of slave on Stack Overflow has steadily declined since 2016, and is highly associated with Java questions.