The Right to Algorithmic Transparency

Abstract

Abstract Chapter 9 considers a disclosure policy that would require transparency with respect to the nature, the uses, and the consequences of algorithms in consumer markets. In various areas of regulatory law, transparency of certain kinds is mandatory, largely on the theory that sunlight can be a kind of disinfectant, helping consumers to make better choices and potentially deterring certain practices. This chapter explores the possibility of designing a “right to algorithmic transparency” that would uncover and mitigate the kinds of practices that concern this book. Consider algorithmic price discrimination. Suppose, for example, that a seller’s algorithm divides consumers into four categories corresponding to their income and wealth; suppose too that wealthier consumers are charged higher prices. Companies might have to disclose that (not particularly alarming) fact. Or suppose that an algorithm uses data on a consumer’s borrowing and saving behavior to identify myopic consumers (who tend to borrow more and save less) and then offers such consumers low introductory prices and high long-term prices. Or suppose that the algorithm learns to identify consumers who would likely overestimate the benefit from the firm’s product, and then sets higher prices for these consumers. A transparency requirement, if implementable, would force firms to disclose that their algorithms are searching for myopic consumers or for consumers who suffer from an overestimation bias, and offering different prices to these consumers. It is easily imaginable that transparency could deter some of the practices on which we have focused here.

Ähnliche Arbeiten