Keywords: data ethics, social theory, data science for social good
Out of necessity, data science often begins with what information is available. The resulting analyses are not only distorted by the limitations of the original data sets, but also take the "value" of the novel application for granted. This talk focuses on the ethical implications of this practice of extracting new forms of value from data. "Academic data science often represents itself as epistemologically valid according to the terms of its site of application, rather than according to criteria internal to its own practice" (Slota et al., forthcoming); the same can be said of how it presents its value commitments as a field. The act of re-purposing information as data is a political one; the current practice of data science seizes few opportunities to question whose interests are served in so doing. Using examples drawn from academic "Data Science for Social Good" projects, I illustrate that even pro-social data science does not engage with the political dimensions of its work, which necessarily intersect with (and move beyond) ethical norms and standards. These illustrative examples will highlight diverging ideas about who "data for good" is for, and the need for data science practitioners to intervene at the point of re-purposing data sets to make these commitments --to data subjects, research sponsors, and those impacted by the work-- explicit. To that end, I suggest concepts from social theory that would move data science toward more reflexive application.