Opening the Black Box: Ethically Responsible Use of Big Data

Predictive modeling (PM) is a process that uses data mining and statistics to forecast outcomes. PM research within social sciences has grown considerably and this direction presents exciting and innovative opportunities for social science research (Shaw, Lee, & Farrell, 2016). First and foremost, PM makes meaningful and accurate predictions used to reallocate limited resources and focus our work where need is highest. Secondly, PM engenders a system of decision making that can be traced to clear metrics, thereby facilitating accountability, transparency, and more concrete program evaluations (i.e., tied to improved outcomes for vulnerable populations and greater equity among populations). However, legitimate concerns surround the use of PM, including the possibility of using PM to “profile” or “target” certain groups of individuals (e.g., in criminal justice, child welfare systems; Angwin, Larson, Mattu,
& ProPublica, 2016) as well as unforeseen consequences of these techniques.

This workshop aims to build an understanding of PM and our applications of it. Our exposition devotes special attention to ethical concerns and challenges, as the following topics relate to them: avoiding methodological pitfalls that can reinforce implicit biases; increasing interpretability of findings and determining appropriate communication with social work staff; privacy and data sharing; stakeholder engagement (or stakeholder buy-in); and best practices.

Participants will be asked to complete a brief pre-workshop survey so that real life exemplars about their target research/policy questions, predictive analytics aims, methods, applications, barriers, etc. can be
used to guide workshop activities. An interdisciplinary university – agency technical partnership team will facilitate the presentation, discussion, and workshop exercises.