Algorithmic Transparency and Welfare: The Unintended Consequences of Disclosure Mandates

Jorge Padilla, Alexander White, Cecilia Nardini, and Kadambari Prasad explore this question in a paper published on SSRN, examining the economic implications of requiring marketplaces to disclose how their algorithms work.

The views expressed in this paper are the sole responsibility of the authors and cannot be attributed to Compass Lexecon or any other parties.

Abstract

Policymakers in major jurisdictions have proposed mandates that would require marketplaces to disclose the internal logic of their algorithms. This paper argues that mandated disclosure of algorithms can undermine the goals it is intended to advance. Those who learn how an algorithm works respond to what they learn. Once an algorithm is public, rivals can copy it, which gives a marketplace less reason to build and improve it. Sellers who know the rules by which a marketplace makes them compete can optimise against those rules and, under certain circumstances, coordinate. A regulator handed an algorithm's process is prone to judging it, when what the regulator cares about, and can observe, are the results that the algorithm produces. These responses can lead to unintended consequences. Each response rests on well-established economic mechanisms that have been studied in settings that go far beyond algorithms. The paper concludes that a pragmatic approach would avoid blanket disclosure mandates and, instead, start from a specific concern and limit the scope of disclosure to the information needed to address the concern.