Spectrum of Agreement: Stances on Ethical Issues across Software Domains

While software drives the modern lives of many, it also introduces a number of ethical issues. Users perceive issues with privacy, fraudulent activity, or censorship in a number of applications from different domains. Yet, users do not always agree on these issues, sometimes even supporting actions that others perceive as unethical. So far, there is no research on how stances on ethical issues relate to each other and how similar software domains are regarding these issues. To fill this gap, we establish a big picture of these relationships by collecting 65769 Reddit posts about applications from 11 popular software domains and manually analyzing 1035 of them. We analyze the prevalence of specific stances according to software domain and type of ethical issue, pointing out similarities between domains. We further probe the potential of knowledge graph representation learning on these complex relationships for automated classification. For example, we find higher likelihoods for support of unethical practices in the business software domain and contextual overlap in posts across all software domains. Relationships inferred from our knowledge graph between software-related online communities, ethical issues, applications, and companies reveal similarities matching those derived from textual context through text embeddings. Our results lead to a better understanding of the nature, prevalence and interconnectedness of ethical issues in software, and inform practices to identify ethical issues at scale.