Artificial Intelligence and Market Regulation: The Way Forward for the CCI

By: Mahima Zamindar and Parina Muchhala


INTRODUCTION

Artificial intelligence (hereinafter ‘AI’) has revolutionised the world, with algorithms taking over most manual processes. Their cost and efficiency benefits have attracted business enterprises to adopt them. Algorithms are being used by enterprises for controlling production and management functions, collecting market data, self-learning about market conditions through programmes, market monitoring, etc. Whilst they contribute to simplifying business processes, there are anti-competitive practices that may also be potentially linked to its functioning. This article seeks to examine these anti-competitive practices and how the Competition Commission of India (hereinafter ‘CCI’) can better equip itself to tackle these issues.

AI’s ANTI-COMPETITIVE BEHAVIOUR: UNDERSTANDING TACIT PRICE COLLUSION 

Juxtaposing the role of AI with the nature of the market it operates in will help understand how it facilitates anti-competitive practices. Generally, AI monitors market conditions and accordingly decides the price of products/services – also called ‘dynamic pricing’. A feature of two-sided digital markets, this involves algorithms enabling consumers and service providers to constantly view, understand and act on changing prices of commodities. In economics, two-sided markets refer to those where the presence of a platform provides a mode for the interaction of two different agents, generally being the end-user and the service provider. 

An example will help better illustrate this. The market consisting of ‘cab aggregators’ is a two-sided market since a website (the platform) like Ola or Uber links the driver to the end-use consumer. Thus, when a request to ‘book a cab’ is made, Ola or Uber’s AI will examine factors like available stock (cab drivers), immediate anticipated demand of consumers, the expected traffic and travel on a particular route at that time, an analysis of the consumer’s history and frequency of use of the platform etc. to display a price on the screen of the consumer, who may then choose to go ahead with the booking. These being variable factors, the prices are likely to fluctuate, making it ‘dynamic’. One of the major benefits of such strategies is that it fosters consumer loyalty through price personalisation. 

Whilst this in itself does not appear to be anti-competitive, there are two intertwined concerns that may arise from this: firstly, such practice may amount to ‘resetting’ the market equilibrium altogether. This practice involves the AI matching competitively charged prices and an optimal output at one point of time with higher prices and lower output at the other. This is a phenomenon we generally witness when cab prices are high during peak hours and lower at night. Such price fluctuations are not necessarily anti-competitive. However, due to the easy availability of information about other competitors publicly, AI algorithms can monitor and ‘adjust’ their prices based on competitor prices. The ecosystem that will be created will involve indirect price fluctuation/flexibility of platforms based on the actions of its competitors. Since this will not allow prices to drop down easily, the possibility of substitutability will reduce, becoming detrimental to consumer welfare. 

The regulatory challenge involved in this scenario is the absence of an express/implied agreement between these competing platforms to collectively influence market prices. ‘Tacit price collusion’, in economic literature, is when there is no express coordination/agreement between firms, but they recognise competitive behaviour in the market and change their strategies accordingly. Such practices are even more common in oligopolistic markets (as most digital markets these days are) where the survival of each firm depends on a comparative advantage it offers vis-a-vis competitors. Thus, while it may appear similar to a cartel, it demonstrates neither intent nor execution of any concerted practices towards distorting competition. 

AI adds a new dimension to this problem, insofar as it is the algorithms that detect, devise, and execute such ‘practices’. This may occur with or without the knowledge of their programmers (self-learning). Whilst the ‘intent’ to collude or the ‘actions of competitors’ (the classic prisoner’s dilemma) would become enough to warrant prosecution in the first scenario, the second one is particularly tricky. Nevertheless, this practice is increasingly being detected worldwide, demanding an urgent review of the Indian Competition Act. 

AN OVERVIEW OF THE COMPETITION ACT, 2002

The Competition Act, 2002 does not expressly envisage such conduct. The Draft Competition (Amendment) Bill, 2020 has also not provided any clarity on this topic. In the past, however, numerous pleas have been made before the CCI to recognise such conduct as ‘collective dominance’ under Section 4 of the Act. Since companies today are able to collectively influence market prices in their favour, they become akin to a ‘single economic entity’ abusing its dominant position in the market. The role of algorithms is no different, for its attempts to match/deliver prices based on competitor conduct will be undertaken across companies, facilitating a sense of collectivity amongst ordinarily competing players, which can be intentional or unintentional. Currently, penalisation remains a distant dream since the CCI has categorically expressed its inability to recognise this concept on numerous occasions, reiterating that the Act has only been drafted with the intent to envisage unilaterally dominant conduct. It appears that collective conduct thus cannot be penalised unless there is an express legislative amendment to that effect. 

Another manner is to equate such conduct to ‘cartels’. Such an argument is plausible because of the possibility of AI being used to further an agreement already entered into by its competitors. In the Competition Act, cartelisation is currently penalised under Section 3. Conventionally, the definition requires the presence of an ‘agreement’, either express or tacit. It is believed that the same can be broadly interpreted to include the conduct by AI into its ambit. However, due to the inherent inability of this provision to deal with oligopolistic markets, there is a need to ensure that there are simultaneous developments under Section 3 and 4 to effectively encompass all forms of potential anti-competitive conduct by AI. 

EQUIPPING THE INDIAN REGIME TO INVESTIGATE TACIT PRICE COLLUSION

Taking into consideration the various antitrust risks that emanate from the use of AI, it is important to explore ways in which competition regulators can effectively mitigate and/or regulate these practices. At the outset, the first step towards penalisation of such conduct entails necessary legislative or informal judicial amendments, either to Section 4 (to prosecute abuse of dominance occurring “jointly or singly”) or to interpret Section 3’s notion of ‘tacit collusion’ to include algorithmic conduct. However, since this has been briefly outlined above and discussed extensively earlier, the scope of this post is restricted to making logistical suggestions to equip the CCI to fight practical challenges associated with the investigation of such practices. 

Firstly, the CCI should constitute an internal committee of economists and data scientists to work on publishing guidelines for firms to undertake self-regulation and algorithmic transparency, in consonance with the existing competition law provisions. This would enable firms to assess and perform appropriate tests on their AI to ensure that it complies with the guidelines. It will also facilitate a much-needed indirect shift in the burden of proof to the firms to demonstrate compliance first, similar to the recent e-commerce report released by it.  Through these internal controls, businesses can ensure that algorithmic programming incorporates pro-competitive fairness metrics. 

Secondly, to aid logistical concerns, the CCI can consider constituting an AI-wing consisting of technical engineers, employees, data analysts and AI code developers. They would develop regulatory AI to assist in efficient detection of algorithmic collusion and check compliance levels of the AI, as argued above. Thirdly, the CCI can mandate that firms are required to issue a mandatory notice informing consumers the ways the data is being collected and used by the firm. This will ensure that each rational consumer is informed of the possible repercussions of their decisions, and are also themselves equipped to reasonably detect any price collusion they may encounter. 

Fourthly, the CCI may also implement specific rules that would aid in affixing liability on the third party AI providers i.e. external consultant or software developers who may develop programs that conduct coordinated algorithmic price collusion. Fifthly, firms may be mandated to undertake documented periodic review/re-evaluation of existing algorithms and produce them as evidence whenever required. This would provide firms with the option to raise ‘objective justifications’ for their conduct.  

Lastly, from an economic standpoint, many suggest that AI should be programmed to revise prices as a reaction to competitor’s prices periodically, and not regularly. However, since AI is built to enhance commercial efficiency and profits, this becomes economically inefficient. Hence, the CCI can consider formulating ‘ethical guidelines’ governing AI use. This should entail an explanation as to the constituents and contours of commercially efficient behaviour of an AI and the extent to which it would be considered pro-competitive. Some scholars suggest that firms must direct attention towards building pricing algorithms in a way that it doesn’t allow AI to collude. However, since this does not adequately address concerns like self-learning, there must be guidelines directing firms to adopt methods where they prohibit algorithms from directly conditioning on rival price alterations. Rather, only economic factors like demand and supply shocks, taxation thresholds etc. must be made permissible as it will promote pro-competitive behaviour guided solely as a response to industry and market conditions, to the advantage of consumers. 

CONCLUSION

While AI presents tremendous opportunities for businesses and e-commerce, it also poses great antitrust risks and invites regulatory uncertainty due to its multi-dimensional nature and ever-changing technological landscape. The CCI must take proactive measures, which may entail a mixture of innovative measures and steps similar to those taken by foreign regulators to ensure that it is better equipped to assess, investigate and prosecute such anti-competitive practices.


(Mahima and Parina are currently law undergraduates at Maharashtra National Law University, Mumbai. They may be contacted here and here, respectively.)

Cite as: Mahima Zamindar and Parina Muchhala, ‘Artificial Intelligence and Market Regulation: The Way Forward for the CCI’ (The RMLNLU Law Review Blog, 17 July 2020) <https://rmlnlulawreview.com/2020/07/17/artificial-intelligence-and-market-regulation-the-way-forward-for-the-cci > date of access.

Leave a Reply