Data Science Ethos Lifecycle: Interplay of Ethical Thinking and Data Science Practice

Abstract

This article presents the Data Science Ethos Lifecycle, a tool for engaging responsible workflow developed by an interdisciplinary team of social scientists and data scientists working with the Academic Data Science Alliance. The tool uses a data science lifecycle framework to engage data science students and practitioners with the ethical dimensions of their practice. The lifecycle supports practitioners to increase awareness of how their practice shapes and is shaped by the social world and to articulate their responsibility to public stakeholders. We discuss the theoretical foundations from the fields of Science, Technology and Society, feminist theory, and critical race theory that animate the Ethos Lifecycle and show how these orient the tool toward a normative commitment to justice and what we call the “world-making” view of data science. We introduce four conceptual lenses—positionality, power, sociotechnical systems, and narratives—that are at work in the Ethos Lifecycle and show how they can bring to light ethical and human issues in a real-world data science project.

Keywords:

• Data science lifecycle
• Ethics in practice
• Narratives
• Positionality
• Sociotechnical systems
• World-making
• Power

Acknowledgments

This article describes an interactive Ethos Lifecycle tool that is being developed by the authors and members of the Academic Data Science Alliance (ADSA) Ethics Working Group. We wish to thank in particular Stephanie Shipp, Cathryn Carson, Micaela Parker, Steve Van Tuyl, Maria Smith, Tiana Curry, Anna-Maria Gueourguieva, Carlos Ortiz, Eva Newsom and Anika Cruz who have engaged and debated the ideas presented here and have provided the vision, design and institutional support to put them into practice in the tool. This article benefited from Margarita Boenig-Liptsin’s fellowship at the Paris Institute for Advanced Study (France), with the financial support of the French State, programme “Investissements d’avenir,” managed by the Agence Nationale de la Recherche (ANR-11-LABX-0027-01 Labex RFIEA+).

Notes

1 Currently, there is no standard data science lifecycle with a set number of steps that is accepted by everyone in the data science community. For example, Stodden (2020) discusses a lifecycle with 11 steps, Janeja (2019) uses a lifecycle with 6 steps, but these do not correspond exactly to the steps we have identified.

Additional information

Funding

The Academic Data Science Alliance provided a gift of financial support to UC Berkeley undergraduates to contribute research and design work on the Data Science Ethos Lifecycle tool described in this article.