Comments of the R Street Institute
In response to the Federal Trade Commission’s request for comments dated June 20, 2018, the R Street Institute respectfully submits the following comments. Submitted in advance of the hearings planned to be held, these are intended to identify topics for those hearings, and will likely be supplemented by more detailed analysis afterward.
This comment is one of several that R Street is submitting, pursuant to the Commission’s request of a separate comment per topic. This comment relates to Topic 9 on the consumer welfare implications associated with the use of algorithmic decision tools, artificial intelligence (AI), and predictive analytics.
Continued progress in AI and algorithmic decision making holds great promise for consumer benefit and for American national security. The Commission has already begun to examine the implications of AI for competition policy, particularly within the realm of financial technology.1 However, AI and algorithms play a role in our economy far beyond the financial sector and the Commission is wise to have included a broad discussion in the upcoming hearings.
In addition to the issues already identified in Topic 9, we therefore encourage the Commission also to consider the following topics.
Dynamics of International Regulatory Competition Around AI. While the Commission has traditionally focused on domestic competition, in an increasingly globalized world, international regulatory actions by the Chinese and European Union (EU) governments, in particular, will be very relevant for the behavior of large multinational firms and startups alike in the United States. For example, the recently enacted General Data Protection Regulation in the EU contained an explainability requirement for algorithms used to make automated decisions about EU consumers.2 The Commission should pay close attention to the effect of this and similar provisions, both in the response of multi-national firms (do they chose to offer different services or different pricing models in the EU vs the US?) and in the rate of technology startup formation (do new startups choose to form or migrate to geographic regions with less restrictive AI regulations?).
The Commission should learn from these results both in terms of the regulatory proposals that should be adopted or avoided here in the United States, but also as a case study for the larger phenomenon of global innovation arbitrage.3 Just as important as having the “correct” regulatory guidelines is an awareness of the relative strengths and weaknesses of our regulatory regime and the way they affect where innovation arises and migrates.
Changes in Industrial Organization Resulting from AI. As with any new general purpose technology, advances in AI are already beginning to shape the structure of new firms. Leading economists have recently begun to study this issue, but much more analysis is warranted.4 How large are the pro-competitive effects of layering AI tools on top of distributed computing platforms? Are the returns-to-scale from data muted by the increasing importance of creative algorithmic design? Why do we not have more developed markets for data sharing? The Commission would be wise to engage deeply in this emerging conversation.
Reducing Entry Barriers to AI Development. Also of interest to the Commission was “whether restrictions on the use of computer and machine learning and data analytics affect innovation or consumer rights and opportunities in existing or future markets, or in the development of new business models.” As a closely related question, we would also recommend examining the ways in which U.S. public policy may have inadvertently created entry barriers for the development and application of AI across the economy. Restrictions on the supply of data scientists and on the supply of publicly accessible data, for instance, may have artificially bolstered the market position of leading tech firms. The Commission would be well-positioned to study this question and connect its larger implications for competition policy.
Competition in Datasets for AI research. Of particular interest to the Commission may be the lack of competition that results from insufficient competitive access to data. This should be thought of primarily along two dimensions: 1) Do the existing set of legal protections around proprietary data access need to be changed? This could take the form of intellectual property review, but also through interoperability requirements. 2) Are there ways we can make existing government databases available to the public to offset incumbency advantages?
* * *
R Street thanks the Federal Trade Commission for the opportunity to submit these comments, and recommends that the Commission pursue the above-identified areas in its ongoing work on promoting competition and innovation.
1 See, e.g., “FinTech Forum: Artificial Intelligence and Blockchain,” Federal Trade Commission, March 9, 2017. https://www.ftc.gov/news-events/events-calendar/2017/03/fintech-forum-blockchain-artificial-intelligence.
2 See, e.g., Bryce Goodman and Seth Flaxman, “European Union regulations on algorithmic decision-making and a ‘right to explanation’” AI Magazine 38:3 (2017). https://arxiv.org/abs/1606.08813; and Lilian Edwards and Michael Veale, “ Slave to the Algorithm? Why a ‘Right to an Explanation’ Is Probably Not the Remedy You Are Looking For,” Duke Law and Technology Review 18 (Dec. 6, 2017). https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2972855.
3 Adam Thierer, “Innovation Arbitrage, Technological Civil Disobedience & Spontaneous Deregulation,” Technology Liberation Front, Dec. 5, 2016. https://techliberation.com/2016/12/05/innovation-arbitrage-technological-civil-disobedience-spontaneous-deregulation.
4 See, e.g., Hal Varian, “Artificial Intelligence, Economics, and Industrial Organization,” The Economics of Artificial Intelligence: An Agenda (University of Chicago Press, Forthcoming). http://www.nber.org/chapters/c14017.pdf.