Big Data Analytics: The Use of Recommender Systems as Viewed Through an Ethical Filter
School of Public Service
Policy decisions are increasingly dependent on abundant volumes of data. This dependence is highly controversial: for example, offender risk scores used by the courts to predict recidivism often influence bail and sentencing decisions, but have been shown to be racially biased. A major obstacle to addressing these controversies with appropriate policy measures is the lack of ethical frameworks to guide policy development for predictive scoring and recommender systems. Dimitris Paraschakis has developed one of the first ethical frameworks for recommender systems, relying in part on a privacy calculus to balance owner goals and user privacy. I will discuss his recommendations in the context of Utilitarianism. The similarities between the privacy calculus and the hedonic calculus in Utilitarianism suggest that this framework may be vulnerable to the ethical trade-offs of Utilitarianism. Through research team discussions and a targeted literature review, I found that policy recommendations relying upon the Utilitarian framework alone may suffer from a tendency to marginalize specific minority groups. In the context of recommender systems, this tendency can perpetuate racial and socioeconomic injustice. Recommender systems need a stronger ethical framework with which to bind themselves. I end with an alternative social epistemological framework to augment the Utilitarian framework.
Turner, Gabrial; Lippitt, Frances; and Gardner, Kimberly, "Big Data Analytics: The Use of Recommender Systems as Viewed Through an Ethical Filter" (2018). 2018 Undergraduate Research and Scholarship Conference. 3.