May I speak? Perceptions on ethical concerns and power while developing software in AI teams

This dataset contains the survey and interview questions and answers, alongside its codebook, used in the paper "May I speak? Perceptions on ethical concerns and power while developing software in AI teams", submitted to the XXIV Simpósio Brasileiro de Qualidade de Software. Paper Abstract: As Artificial Intelligence (AI) technologies increasingly shape decision-making processes, ethical concerns in their development have gained prominence—yet most studies focus on high-level decision-makers or large tech corporations in the Global North. This paper investigates how Brazilian software engineers working in AI teams at a mid-sized company perceive ethical principles, navigate ethical dilemmas, and respond to emerging concerns in their daily work. The goal is to understand how practitioners—specifically Junior, Mid-Level, and Senior developers—interpret and apply ethical frameworks in the concrete task of building AI systems such as classifiers, image generators, and object detectors. The study employs a mixed-methods approach, combining data from 18 survey responses with 8 in-depth interviews, and uses both inductive and deductive coding for qualitative analysis. Results reveal that while participants recognize key ethical principles, their ability to address ethical concerns is often constrained by organizational hierarchies, limited autonomy, and a lack of formalized ethical processes. This research contributes to a growing body of literature on AI ethics in the Global South by shedding light on the everyday ethical agency of developers in underrepresented contexts. It emphasizes the need for organizational support structures that empower practitioners to raise and resolve ethical issues throughout the AI development lifecycle.