The authors are Madhav Tripathi and Prateek Yadav, third year students at RMLNLU.
The injection of big data into multiple industries and consumer lifestyle paradigms through pricing algorithms has altered the competitive landscape. It has changed how corporations make commercial and strategic decisions. There is an increasing proclivity for businesses to employ pricing algorithms that speedily react to market conditions, improve pricing models and tailor services by studying emerging trends such as the ones used by major airlines, taxi operators (OLA, Uber) and online retailers such as Amazon.
These algorithms use large sets of collected data coupled with autonomous machine learning to digitally fix the prices without any interaction between the executives of these enterprises, making it near impossible to prove the presence of any anti-competitive agreements or intent on the part of these firms.
Under India’s current competition regulation framework, abuse of dominance and price-fixing are problematised. India’s competition law functions within a narrow ambit of harm that considers price distortion practices and abuse of dominance as significant determinants of consumer welfare. Therefore, it penalises any cartel that fixes the price.
The authors discuss these algorithm-based challenges for competition law enforcement and market regulation in India. This article addresses whether there is a need to revise the traditional concepts of agreement and tacit collusion for antitrust purposes. It studies the regulatory design and capacity to evaluate the preparedness of the competitive regime to investigate algorithms. Furthermore, it juxtaposes the same with landmark developments from that of the European Union (“E.U.”), the United States (“U.S”.) and the United Kingdom (“U.K.”). Finally, we propose recommendations for India’s competition regulators to develop a comprehensive strategy to address algorithmic harms and the position best suited to deal with rapid technological advancements in the digital market.
Algorithms in Play –The Modus Operandi
The Digital Market is the new normal in markets and economics, and the angelic hand of technology is the new market tactic. The marketplaces like Amazon and Flipkart are running so successively because of these algorithms. Amazon uses its pricing algorithm “A09” and “A10” to rank the products of their hundreds of sellers based on various factors to try and let the customer get his desired product in no time. The sellers use the algorithm to shuffle their product’s prices as per demand and supply, which is a commendable pro-competitive step. Other examples of pro-competitive steps are cost reduction, quality improvement and better allocation of resources.
But , the issue arises when the competitors profit by colluding with each other by entering into “automatised virtual agreements” through these Pricing Algorithms. As per the current debate on algorithmic collusion, these pricing algorithms facilitate a price agreement between competitors.
Let us objectively define the problem. For instance, there are five competitors in a particular digital market. Each uses a different pricing algorithm without any allegedly explicit knowledge of other competitors using that and no alleged intention of collusion. The evidence of this could be no record of physical or virtual or accidental or indirect communication or exchange of thoughts.
But, when they all deploy the algorithms, by the very nature of the algorithms, all those algorithms would assess the market by breaching the boundary of their own business, unlike in a pro-competitive scenario. In this design, the companies are canonically harmonised. Therefore all those players’ commodity prices ( for example) get the same, which is a little lower than the cheapest of those who do not use the algorithm. From this, it can be observed that – the algorithms would not make a “price war” among themselves even when no communication has been done among the players. It is happening by virtue of the self-learning capacity of A-I based algorithms.
The question that finally churns out of the problem is – how can the current law punish these digital cartels, which were formed owing to the self-learning algorithm without any alleged human intervention?
The Legal Position in the Present Regime
Section 3 talks about “Anti-competitive Agreement”, reading with section 3(1), “Section 3(3) states that “ Any agreement entered into ………………………similar trade of goods or provision of services, which—
(a) directly or indirectly determines purchase or sale prices.”
When an agreement is formed and charged under section 3(3)(a), two ingredients that must be present are “the agreement” and “the impact of the agreement – which is the “determination of price.” Here in the case, we are dealing with, we can see the “determination of price”, but the moot question remains: Is there an agreement when the market players deploying the pricing algorithms?
The Notion of agreement: Does it need revisiting?
The definition of agreement in this section comes from section 2(b) of the act, which states that- “agreement” ……………………………….by legal proceeding.”
So, the question now becomes very precise, which is –“When the companies deploy these algorithms with the basic knowledge that these algorithms may be able to collude without their order, intention and intervention as happening here”, can that be called as “tacit meeting of mind and does that meet the criteria of section 2(b) and consequently of section 3(3)(a)?
To answer such a question, the length of the arms of section 2(b)‘s definition must be checked. In the case of Director, Supplies & Disposals, vs Shree Cement Limited, the CCI explained that establishing a “tacit meeting of mind” also requires a “nod or wink”. The standard of proof required to prove an understanding or an agreement would be based on 'preponderance of probabilities' and not 'beyond reasonable doubt'. In Jyoti Swaroop Arora vs The CCI, the Delhi HC observed, “Tacit agreement is proved by circumstantial evidence, and the circumstantial evidence should point to the higher preponderance of probability.” Before moving to the suggestions from the authors, it is crucial to look into the foreign jurisdiction where these challenges arose first.
Experiences in the foreign Jurisdictions
So far, the cases in the jurisdictions like The U.S.A. and The U.K. have been around algorithmic problems which involved some human intervention, like the very famous Poster Case in the USA for instance .In these cases , there was a human intervention and of no use for this article.
If we look beyond the case-laws to many rules and regulations the foreign jurisdictions have so far formed to tackle this challenge, including the broad definition of agreement. In the case of Bayer vs AG, the definition given by the European courts was – “an agreement reflects “a concurrence of wills between economic operators on the implementation of a policy, the pursuit of an objective, or the adoption of a given line of conduct on the market, irrespective of the manner in which the parties’ intention to behave on the market in accordance with the terms of that agreement is expressed. “In other words, there has to be a meeting of minds irrespective of any method”.
In the United States, section 1 of the Sherman Act uses multiple terms to refer to an agreement, including “contract”, “combination in the form of trust”, and “conspiracy”. According to the U.S. Supreme Court in Interstate Circuit Inc v United States, an agreement does not necessarily need to be explicit or formal as long as it involves “a unity of purpose or a common design and understanding, or a meeting of minds”.
Section 5 of the F.T.C.- is “principle-based” rather than “rule-based”. During the time when it was drafted, it was suggested that statutes like Section 5 could be used to tackle algorithmic collusion if the agency could show that, when developing the algorithms, defendants were either motivated to achieve an anticompetitive outcome or were aware of their actions’ natural and probable anticompetitive consequences.
A New Lens to the Old Definition
So, now after learning the new interpretations, if we return to our moot problem- In our current situation, where the competitors know about the “Self-learning mode of algorithms” and have the idea that in the same market, there could be other “Self- learning Algorithms” working, the slight knowledge of these two factors, does it show a silent meeting of mind? Here the defence of the accused would be that – Firstly, It's essential to use the algorithms for the business. Secondly, it is impossible to stop others from using other Self-learning algorithms. If they interact – it is automatic. Mens rea is missing.
So far, to constitute a “cartel”, we needed a malaise meeting of mind to distort competition by determining the sale or price of a product. And to prove “that meeting of mind” in the Indian jurisprudence, what is required is “mere preponderance of probability “, which means there just needs to be a good possibility of the “meeting of the mind”
Now, having such jurisprudence in the background, to punish this “Digital tacit collusion”, the agencies need to add a new lens or perspective to the agreement definition. Till now, the interpretation of agreement has been that- to enter into an agreement, two parties have to be present at the same time to signal or wink. But neither of the two parties is present while the algorithms wink (metaphorically) in the digital space. This must also be interpreted as an agreement between those two parties because these two algorithms work as agents for the competitors. In this situation, the competitors may not know when exactly. Still, they will know that these two algorithms will meet because the competitors know about the nature of their algorithms and also have the knowledge about the others using it.
Now a defence from the accused could be that just because of the knowledge of these factors cannot be an agreement. To counter this, we now have to add not a new lens but a new layer of interpretation to the definition of agreement. Taking inspiration from the suggestion which was given during the drafting of section 5 of FTC (aforementioned) , it was for application during the forming of the algorithm. We will apply for those who use the algorithm.
So, the new interpretation that must be read under the definition of agreement is that – When suppose two (allegedly) unknown competitors know the natural and probable anti- competitive consequences of their algorithms that is, they can mingle with the other algorithm to form a tacit collusion, and still deploying in the same market, it is quite “probable” that they might have colluded to do the collusion in this way.
By this interpretation the agency will be able to punish “tacit collusion” and those competitors who have delicately left no evidence of a physical meeting to escape the jaws of laws of punishing the collusion.
Conclusion
So, if the authorities use this interpretation of agreement, then, such ‘without the biological presence of a human’ agreement will qualify as agreement as defined under section 2(b) and, therefore, will qualify the requirements of section 3(3)(a) of the competition act enabling the agencies to punish the “Digital Cartels” or “Digital Tacit Collusion”. These new interpretations are required because of the abundance of the digital economy and its combination with A. I. will present many complex challenges before the competition regime. This area is highly complex and uncertain, where lack of intervention and over-regulation could pose severe costs to society, especially given the potential benefits of algorithms. Whatever actions are taken in the future should be subject to in-depth assessment and a cautious approach.
Comentários