University of Haifa
When a Small Change Makes a Big Difference
A growing body of scholarship is addressing the risks of opaque analyses as well as the fear of hidden biases and discrimination that may come along with automated decision-making. This work mirrors the growing public anxiety about the influence and lack of transparency of Big Data analysis practices. The current literature and related debate struggles with defining and revealing unfairness. As a possible strategy to respond to, and possibly preempt the fairness concerns, algorithmic examinations might seek out differences between inputs and outputs. This Article will focus on one form of such assessment – seeking out whether small changes in inputs produce large changes in the outputs. These are instances in which a “Small Change Makes a Big Difference” (or “SCMBD”). This simple form of analysis (which is both contemplated and already conducted) could prove to be an important innovation in building trusted and trustworthy AI systems and goes beyond merely examining the ex-post (under)representation of groups in the process’s outcome. To explore the nature and merit of SCMBD testing, this Article will proceed as follows. Beyond our introduction we will move, in Part II, to define SCMBD and confront challenges in attending to this concept’s scope. In Part III we will explain why SCMBD dynamics might be deemed unfair and therefore used to flag problematic algorithmic processes. In Part IV, we identify factors that would suggest that SCMBD is not problematic. In Part V, we conclude with policy recommendations to be applied in situations where we will find that SCMBD testing is appropriate. These include institutional and practical recommendations.
Prof. Tal Zarsky is the Vice Dean of the University of Haifa's Faculty of Law. His research focuses on Information Privacy, Cyber-Security, Internet Policy, Social Networks, Telecommunications Law, Online Commerce, Reputation and Trust. He published numerous articles and book chapters in the U.S., Europe and Israel. His work is often cited in a variety of contexts related to law in the digital age. Among others, he participated in the Data Mining without Discrimination project, funded by the Dutch Research Council (NWO) as well as other national and international research projects. He has advised various Israeli regulators, legislators and commercial entities on related matters. He severed on a variety of advisory boards and is a frequent evaluator of articles and research grants for various international foundations. Prof. Zarsky was a Fellow at the Information Society Project, at Yale Law School and a Global Hauser Fellow, at NYU Law School. He completed his doctorate dissertation, which focused on Data Mining in the Internet Society, at Columbia University School of Law. He earned a joint B.A. degree (law and psychology) at the Hebrew University with high honors and his master degree (in law) from Columbia University.