Ebook: Responsible Data Science

This is the first book on ethical data science that provides hand-on practical technical steps practitioners and managers can take to fix the ethical and fairness issues arising from large data sets in data science. The book sets the stage with a review of the types of ethical challenges posed by the increasing use of data science methods to make decisions previously made by humans. Some of the topics covered include: Types of ethical challenges posed by data science including overtly bad examples like the Chinese modification of the credit score as "social worth" as well as grey areas such as increase surveillance from law enforcement through consumer devices (Ring, Alexa, Nest). Behavior manipulation (Cambridge Analytica) and deepfakes as well as unintentionally bad consequences such as mortgage discrimination and biased cash bail systems Review of “black box” models and how their usage can aggravate issues of model transparency, bias, and fairness Approaches for making black box models interpretable and identifying issues of bias or fairness Using statistical methods to analyze the effects of models and mitigate bias The book will describe interesting real cases in focused, readable and practical terms suitable both for managers with some technical ability, and for practitioners, steps to address key ethical issues in data science.