Responsible Data Science

CHAPTER 6Beginning a Responsible Data Science Project

Armed with the historical, social, and technical knowledge that we have acquired over the past five chapters, we are finally ready to work through our own example data science projects. Arriving at this point may have been an unexpectedly long journey; however, each of the cases that we have discussed so far ought to have impressed upon us the need for deeper consideration of every aspect of our data science projects. We must go beyond our roles as technical experts and engage with broader questions beyond those that we can answer with code and mathematics alone. Data science is difficult to do well; doing it well and responsibly raises that bar even further.

If we want to learn to use the Responsible Data Science framework to identify potential modeling harms, then we need to become more familiar with the harms themselves. Rather than shying away from ethically charged modeling tasks (e.g., crime prediction or recidivism prediction), we will work through these tasks ourselves to learn firsthand the challenges that they present.

Over the course of this chapter, we will work through the Justification, Compilation, and Preparation stages of the Responsible Data Science framework introduced in Chapter 4, “The Responsible Data Science Framework.” In completing these stages, we accomplish the following:

JUSTIFY:

• The tractability of the modeling task, the relevance of the dataset to the modeling task, and the efficacy of the ...