Keywords: data ethics, data science, ethics, critical data studies
With numerous high profile scandals involving data-intensive analyses and technologies littering the news, it is clear that data science is currently facing a crisis of conscience. In this talk, I discuss how many data scientists are earnestly attempting to mitigate the possible harmful outcomes of their practice by increasing methodological rigor. This includes, for example, collaborating with subject matter experts, improving documentation, developing numerical corrections for bias, offering greater transparency, and building more interpretable models. I call this approach “data science as ethical convention,” and argue that it is necessary, and yet insufficient. I propose three alternative and complementary approaches that can help us think about data science ethics more holistically: data science as ethical interrogation, data science as ethical innovation, and data science as ethical participation.