Abstract

The persistent gender gap in Software Engineering (SE) and related software development fields necessitates a thorough examination to expose the root causes and advance women’s engagement in technological innovation. This disparity presents both societal and technical challenges, perpetuating implicit gender biases in technology due to the limited representation of women. Online forums provide insight into women’s experiences with sexism in technical environments, but the unstructured nature of this data complicates the extraction of such specific instances. Our research aims to address this issue by analyzing online narratives from women software developers illustrating their experiences with sexism in technical teams. We initiate this study by constructing a taxonomy to identify various forms of sexism. Subsequently, we apply conventional data extraction techniques, such as static keyword-matching, and advanced artificial intelligence (AI) methods, including semantic similarity, to identify sexist experiences in the online dataset. Lastly, we evaluate the AI model’s effectiveness with Equity, Diversity, and Inclusion (EDI) experts to ensure alignment with nuanced human understandings of sexism. Our results reveal the development of a taxonomy encompassing four distinct classes of sexism, supported by definitions, anchor examples, and lexicons. We observe that while semantic similarity techniques are proficient in extracting narratives of sexist experiences, the model encounters difficulties in accurate classification. Furthermore, our results highlight the intricate challenges of trying to align AI systems with human interpretations of sexism as defined in our taxonomy. Additionally, our findings reveal three previously overlooked instances of sexism. Based on our outcomes, we propose a code of conduct for practitioners to mitigate sexism within technical teams, enhancing women’s participation in SE and technological innovation.