Who Speaks for Ethics? How Demographics Shape Ethical Advocacy in Software Development

TL;DR

The paper addresses how practitioner demographics shape ethical advocacy in software development by surveying 217 practitioners across roles, industries, and regions, using a validated taxonomy to categorize ethical concerns. It finds that marginalized groups report higher frequencies of ethical issues and greater empowerment to address them, yet overall organizational support for ethics remains limited, with market pressures and perceived neutrality impeding action. Onboarding, prioritization, and intervention are differentially influenced by nationality, residence, gender, race, and ability, and intersectional effects reveal complex patterns across groups. The work underscores the need for reforms in education and development processes to incorporate diverse perspectives and uphold responsible computing in a rapidly evolving technology landscape.

Abstract

The integration of ethics into software development faces significant challenges due to market fundamentalism in organizational practices, where profit often takes precedence over ethical considerations. Additionally, the critical influence of practitioners' individual backgrounds on ethical decision-making remains underexplored, highlighting a gap in comprehensive research. This is especially essential to understand due to the demographic imbalance in software roles. This study investigates ethical concerns in software development, focusing on how they are perceived, prioritized, and addressed by demographically different practitioners. By surveying 217 software practitioners across diverse roles, industries, and countries, we identify critical barriers to ethical integration and examine practitioners' capacity to mitigate these issues. Our findings reveal pronounced demographic disparities, with marginalized groups - including women, BIPOC, and disabled individuals - reporting ethical concerns at higher frequencies. Notably, marginalized practitioners demonstrated heightened sensitivity to ethical implementation and greater empowerment to address them. However, practitioners overall often lack the support needed to address ethical challenges effectively. These insights underscore the urgent need for reforms in software education and development processes that center on diverse perspectives. Such reforms are essential to advancing ethical integration in software development and ensuring responsible computing practices in an increasingly complex technological landscape.

Figures (4)

• Figure 1: Responses to research questions Q1.1 through Q2.3, with the Full sample, then stratified by marginalized demographic groups, including BIPOC, Global South (based on nationality and residence), women (gender), individuals with disabilities, socio-economic status (middle class), participants with less than 5 years of experience, and those identifying as female. The demographic groups presented were selected based on statistically significant results. The x-axis represents the research sub-questions, while the y-axis shows the percentage distribution of responses across Likert categories.
• Figure 2: Responses to research question Q2.4 to Q2.7, with the Full sample, then stratified by demographic groups, including BIPOC, Global South (nationality and residence), women, and individuals with disabilities. The x-axis represents the demographic stratifications, while the y-axis displays the percentage distribution of responses across Likert categories.
• Figure 3: Responses to research questions Q3.1 through Q3.3a, with the Full sample, then stratified by demographic groups, including BIPOC, Global South (nationality and residence), women, participants with less than 5 years of experience, and those identifying as queer. The demographic groups presented were selected based on statistically significant results. The x-axis represents the research sub-questions, while the y-axis shows the percentage distribution of responses across Likert categories.
• Figure 4: Collinearity heatmap depicting the relationships between demographic variables measured by Cramér's V. Higher values indicate stronger associations between variables.