Date of Award
Fall 11-7-2022
Degree Type
Dissertation
Degree Name
Doctor of Philosophy in Analytic and Data Science
Department
Statistics and Analytical Sciences
Committee Chair/First Advisor
Ying Xie
Committee Member
Sherrill Hayes
Committee Member
Herman Ray
Committee Member
Jennifer Priestley
Committee Member
Michael McBurnett
Abstract
Ethics can no longer be regarded as an add-on in data science and analytics. This dissertation argues for the necessity of formalizing a new, practically-oriented sub-discipline of AI Ethics by outlining needs, highlighting shortcomings in current approaches, and providing a framework for Ethical Analytics, a field concerned with the study of the ethical issues surrounding the development, deployment, and/or dissemination of ML/AI systems and data science research, as well as the development of tools and procedures to mitigate ethical harms. While data science and machine learning are primarily concerned with data from start to finish, ethical analytics is concerned primarily with people – moral agents, the groups and societies they comprise, and the world they inhabit. It studies the issues that arise in holistic sociotechnical environments, and it seeks to develop concrete solutions or interventions where possible – from the mathematics and algorithms to procedures and protocols.
In addition to the framework, this dissertation includes 2 additional contributions to the field. One applies ethical analytical problem solving to issues of trust and transparency for consumers (and lenders) in credit risk modeling, leading to the enumeration of 3 minimum requirements that explanations should meet in order to satisfy both regulatory and ethical considerations of transparency in the United States socio-historical and legislative context. It is then demonstrated that differentiable nonlinear models (i.e., neural networks) can be made to satisfy these requirements. The final contribution introduces a procedural approach to jointly conducting ethical foresight analysis and assessing principle alignment for prospective ML/AI technologies and data science research along with an interactive dashboard for visualizing principle-specific ethical risk. The byproduct of this procedure is an auditable document called an Ethical Assessment Sheet (EAS).