Author: Rainer Mühlhoff
Publisher: Ethics and Information Technology
Publication Year: 2021
Summary: The following paper focuses on predictive analytics (PA), where a model is used to predict the actions of a target individual/group based on large sets of data about other people. The author also defines the term “predictive privacy” as the ethical concerns that arise when sensitive information about a person is predicted based on proxy data gathered from others. The paper discusses how this particular subcategory of privacy concerns impacts all aspects of building a predictive model. In particular, the concept of a “prediction gap” is discussed in which a statistical inference (which is generally a range of possible outcomes for a population) is reduced down to a single prediction about how an individual will act. The author concludes that predictive analytics are inherently designed to discrepant and argues that serious political and social debate is needed to determine if PA is something we want in our society