The “Algorithmic” Factor in Competition Law

Date : November 05, 2019

The periphery of ‘human superiority’ has been breached, what we are living in now, is the Fourth Industrial Revolution which is symbolised by the dependence of man over Artificial Intelligence (AI). The Competition Act in India which is not a very old member of the legal fraternity is already affected by the challenges of the Fourth Industrial revolution. The major challenge being that of online pricing Algorithms.

Algorithms
The term Algorithm has not been defined clearly; however a general understanding is that it is the application of reasoning by an AI for the purpose of finding a sequence or solution.[1] One of the aims of competition is to ensure that sellers set a competitive and not collusive price, to protect consumer and economic welfare. However, the Competition Commissions across the Globe have been tackling this very problem of collusive pricing on account of AI based pricing Algorithms.[2] Recently, the European Commissioner for Competition, Margrethe Vestager, expressed concern over the ills of companies trying to hide behind the garb of computer programming when faced by collusion charges.[3] Use of Algorithms is not new but, it is the increase in the use of pricing algorithms that has made it possible for competitors to collude their prices.[4]

Pricing Algorithms analyse and collate a pool of data to determine the price of a product or service, thus enabling the sellers to leave the pricing burden on algorithms.[5] The new age development in this area is that of Learning Algorithms, i.e., those which have tendencies of learning by observation and thus may cause a non-human induced collusive behaviour.[6]

Algorithmic Collusion

Whenever dealing with Algorithm based collusion it is necessary to keep in mind the fact that it is after all an artificial person that is involved.[7] In the age of e-commerce, the task of Algorithms is to dance to the music provided by humans, thus acting as aids to behavioural actions.[8] The agreements which result in collusive behaviour are anti-competitive agreements however, when the collusion is induced by an algorithm i.e. without an agreement, the illegality of the act is questionable.[9]

With “agreements” as the key, an agreement to follow a particular formula or pricing algorithm may be said to be the anticompetitive act.[10] Thus it could be deciphered that, the role of the regulators begins from the finding of such agreements. Ariel Ezrachi & Maurice E. Stucke had analysed the categories of algorithmic collusion as follows-[11]

Messenger Scenario

Here the players in the market use the medium of computers or a single algorithm for the purpose of colluding.[12] Such a situation was seen in the case of U.S v. David Topkins[13] wherein the conspirators, agreed to sell their posters (product) by using a single pricing algorithm so as to assist in the likeness of prices amongst them. This was found to be an unlawful agreement. Similarly, in 1994, it was found that six airlines had been sharing a computerized online booking system which enabled collusive price setting and thus it was found to be anticompetitive.[14] Thus the messenger scenario is where the conspirators agree to use the algorithm to implement collusion.[15]

 Hub and Spoke Conspiracy

This form of conspiracy refers to an agreement between either vertical or horizontal players (spokes) via the medium of a platform (hub), thereby being a sort of indirect agreement.[16] In this case, as in the messenger scenario, a single pricing algorithm is used however, it is the algorithm developer that causes the players to collude.[17] Thus, it is the use of the hub that becomes the subject of agreement.

Predictable Agent

In this category, the players develop separate pricing algorithms however, they are programmed in such a way that all will react in the same way to similar situations so as to result in a tacit collusion.[18]

Autonomous Machine

This is considered to be the most advanced category of algorithmic collusion as the AI learns by doing and thus it becomes self-sufficient in predicting the necessary behaviour for colluding.[19]

Conclusion

The main problem is that though we are aware that the Algorithms “may” pose a problem to the competition in the market, the watchdogs of competition may not fare so well in dealing with the posed problems. The regulators are yet to understanding the amplitude of data collection and misuse,[20] thus calling for greater research and interpretation of emerging laws such as data protection, privacy etc.

It can be observed that the problem per se is not with the use of Algorithms but instead it is with the agreement to conspire to use a particular algorithm or Algorithm developer for the purpose of collusive actions. The unexplored problem remains to be the enigma of AI taking control over our lives by learning by doing what its master has taught.

Author: Rhea Abraham  3rd year B.A.L.L.B (Hons.) Maharashtra National Law University, Nagpur, intern at IP and Legal Filings  and can be reached at support@ipandlegalfilings.com.

References:

[1] See, Francisco Beneke& Mark-Oliver Mackenrodt, Artificial Intelligence and Collusion, 50 IIC (Jan. 2019).

[2]Noel Beale and Sandra Mayenda, Competition law and ecommerce: 'it wasn't me, it was the algorithm!'Burges Salmon(Nov. 26, 2018)https://www.lexology.com/library/detail.aspx?g=6b55b42b-c225-4407-8065-eb561227a019.

[3]VaibhavChokse,Why digital cartelisation will be a new challenge for the anti-trust regime, Financial Express (Aug 13, 2018), https://www.financialexpress.com/opinion/why-digital-cartelisation-will-be-a-new-challenge-for-the-anti-trust-regime/1278723/.

[4] Bill Baer Sonia KuesterPfaffenroth, Pricing Algorithms: The Antitrust Implications, Arnold&Porter(Apr 17, 2018), https://www.arnoldporter.com/en/perspectives/publications/2018/04/pricing-algorithms-the-antitrust-implications.

[5]Id.

[6] See, Id.

[7]PallaviGuniganti, US DOJ deputy: algorithmic cartel requiresagreement, GCR (Feb. 05, 2018) https://globalcompetitionreview.com/article/1153380/us-doj-deputy-algorithmic-cartel-requires-agreement.

[8]Supranote 1.

[9]Supranote 4.

[10]Algorithms and Collusion - Note by the United States, DAF/COMP/WD(2017)41, OECD(Jun 2017), https://www.ftc.gov/system/files/attachments/us-submissions-oecd-2010-present-other-international-competition-fora/algorithms.pdf.

[11]Ariel Ezrachi& Maurice E. Stucke, Virtual Competition: The Promise and Perils of the Algorithm-Driven Economy 46 (Harvard Univ. Press 2016).

[12]Maurice E. Stucke&Ariel Ezrachi, How Pricing Bots Could Form Cartels and Make Things More Expensive, HARVARD BUSINESS REVIEW (Oct 27, 2016), https://hbr.org/2016/10/how-pricing-bots-could-form-cartels-and-make-things-more-expensive.

[13] United States v. Topkins, No. CR 15-00201 (N.D. Cal. Apr. 6, 2015) (Information and Plea Agreement).

[14]Id.

[15]Jay L. Himes &Tianran Song, “Welcome to the Hotel California”:

The Beast of Algorithmic Pricing, Cpi Antitrust Chronicle(Feb 2019), https://www.labaton.com/hubfs/CPI-Himes-Song.pdf.

[16]Roundtable on Hub-and-Spoke Arrangements – Background Note, DAF/COMP(2019)14, OECD (Oct. 2019), https://one.oecd.org/document/DAF/COMP(2019)14/en/pdf.

[17]Supranote 6 at 573.

[18]Madhavi Singh, Algorithmic Collusion in Flight Pricing in India, Law School Policy Review (Nov. 29, 2018) https://lawschoolpolicyreview.com/2018/11/29/algorithmic-collusion-in-flight-pricing-in-india/.

[19]Supranote 14.

[20]Supra note 11.