This article, available here, surveys EU case law on the role of anticompetitive intent in abuses of dominance, with the goal of understanding how intent can be relevant to the assignment of liability for anticompetitive algorithmic outcomes.
The role of subjective intent in EU antitrust analysis remains controversial. Some argue that evidence of intent is an invaluable tool in the antitrust arsenal, allowing agencies and litigants to address anticompetitive conduct where facts are ambiguous or evidence of harm to competition inconclusive. Others warn against relying on intent. First, ‘sales talks’ encouraging employees to beat – and indeed eliminate – competitors is common and merely indicative of a (competitively desirable) aggressive business strategy. Secondly, banning any exhortation to compete aggressively would encourage firms to deploy more subtle forms of inducement when engaged in anticompetitive conduct, while favouring those with the resources to develop such strategies.
The law seems to follow a middle path in this debate, suggesting that the notion of subjective intent can only be validly relied upon where, on balance, other evidence already indicates that there is sufficient likelihood that the defendant’s intended conduct will have anticompetitive effects. In other words, subjective intent is only relevant in the context of an objective analysis of effects, and there must be a sufficiently close connection between subjective state of mind and anticompetitive outcome. In the algorithmic context, however, it is tempting to rely on subjective intent, since the outcome of algorithms is often unpredictable and their operation can be hard to understand from the outside. As such, the author develops a ‘qualified intent’ solution that would avoid overreliance on subjective intent in those situations where undertakings may be liable for algorithmic design choices.
Section 2 reviews the role of intent in abuse of dominance cases.
According to Hoffman La Roche, abuse is an objective concept, so one may argue that intent plays no role under EU law. However, this would be mistaken. Article 23(2) of Regulation 1/2003 clearly indicates the need to establish intention or negligence in order to impose a fine on an undertaking – in line with the recognition by the European Court of Human Rights that competition fines are criminal in nature, and hence trigger criminal process guarantees such as the presumption of innocence. This implies that both knowledge (actual or constructive) and control over one’s anticompetitive conduct must be established before a fine can be imposed. It would thus seem that while no proof of intention to harm competition is needed, this is so only insofar as that result is a likely and foreseeable consequence of conduct which could have practically been addressed by the undertaking (were it not for its negligence). This would align with the Court’s oft repeated holding that the requirement to establish intention or negligence is satisfied if the undertaking concerned ‘cannot be unaware of the anti-competitive nature of its conduct, whether or not it is aware that it is infringing the competition rules’.
A different question is what role evidence of subjective intent should play in competition cases. Recent cases – in particular Akzo, Tetra Pak II, Compagnie Maritime Belge, Astra Zeneca and France Telecom, which the author analyses in detail – have clarified this, although not always consistently. The author ultimately reads the case law as creating a distinction between objective and subjective intent. Objective intent is attributed to a dominant firm by way of inference from business decisions it made. It is obtained through an inductive process that builds on a rational homo economicus, as illustrated by widely deployed unilateral conduct tests such as the ‘no-economic sense’, the ‘profit-sacrifice’ and the ‘as-efficient competitor’ test. As such, it focuses on the presumed state of mind of a reasonable person (rather than the defendant specifically) in committing a particular act or omission. It is objective intent, and objective intent alone, which is relevant to determine whether the mens rea element of competition infringements is met.
Subjective intent, on the other hand, refers to the defendant’s actual state of mind – i.e. its intentions. Subjective intent is typically only relevant when it supports findings of objective intent that are also evidenced by other means. However, subjective intent gains particular salience where the effects of a business conduct are ambiguous, where it serves to clarify the competitive nature of the investigated factors. Under the case law, when subjective intent is used to clarify whether the dominant firm sought to anticompetitively exclude a competitor, the evidence must establish that subjective intent beyond reasonable doubt (or, as the EU courts have confusingly put it in France Telecom, on the basis of objective factors).
Section 3 examines the roles played by subjective and objective intent in the Commission’s Google Shopping decision.
The author considers that the Commission has misunderstood the case law, both in its Art. 102 Guidance Paper and in the Google Shopping decision. In particular, the Guidance Paper’s states that subjective evidence can be used to interpret conduct by a dominant firm. Such a formulation is overbroad, to the extent that it could be taken to mean that evidence of subjective intent may prevail over conflicting evidence of objective intent. This would run against the principle set out in the case law that the subjective and objective components of intent must be aligned, with the latter being a necessary element to establish an infringement.
The impact of this interpretation is apparent in the recent Google Shopping decision. As noted by others, the theory of harm developed in this case – self-preferencing – is rather vague, and it is unclear how previous case law is relevant to its particular factual scenario. In doing so, the Commission failed to clarify what types of ‘self-preferencing’ conduct breach the special duty incumbent on dominant firms and, hence, are abusive. The reference in the remedial order to ‘not engaging in any conduct or act having equivalent object or effect” may well be interpreted as preventing algorithmic decisions that have a merely theoretical capability of favouring Google’s own services despite the absence of any materialised, or indeed likely, effects. However, that would appear to be in tension with the rationale of requiring negligence to be established before imposing a fine – namely that the dominant company must have at the very least infringed a duty of care in connection with the foreseeability and preventability of the harmful effects of its conduct. Taken at face value, the Commission’s decision implies that a dominant company having developed or used an algorithm is de facto strictly liable for any possible anticompetitive (in particular, self-favouring) effects derived therefrom.
Against this, it might be said that the Commission relies very heavily on internal documents demonstrating subjective intent to leverage market power. The question is whether this suffices to establish intent or negligence, particularly when the case law requires them to be assessed by reference to objective intent. Specifically, the Commission took issue with the outcome of Google’s algorithmic choices without proving that its design choices violated a duty of care – i.e. that such choices led to foreseeable and preventable anticompetitive harm – from which negligence could be deduced. In doing so, the Decision may have infringed the presumption of innocence, the principle of legality and the principle of proportionality in relation to the remedy.
Section 4 offers a possible solution to avoid overreliance on intent, with particular regard for situations in which undertakings may be liable for algorithmic design choices.
Unless self-preferencing is prohibited per se, one needs to develop a limiting principle for the concept of ‘anticompetitive self-preferencing’, which provides legal certainty for undertakings offering algorithmic selection or ranking services. This article proposes a ‘qualified intent’ doctrine to this end, which flows from existing case law. Under this approach, the requisite intent should be grounded on three basic principles: (1) the anticompetitive outcome is foreseeable by the dominant company, based on its knowledge or reckless disregard of the consequences of the action; (2) that outcome is an immediate consequence of the dominant company’s purported conduct; (3) intent is grounded upon a set of facts which, in the context of the entire body of evidence, make the achievement of an anticompetitive outcome more likely than not. The main difference of this approach, when compared to existing case law, is that it moves from a standard of ‘knowledge or negligence’ to a standard of ‘knowledge or reckless disregard’. The consequence of such a move is that the inquiry becomes subjective, rather than objective, thereby enabling authorities to take into account the superior knowledge of certain dominant firms over a reasonable, but only averagely knowledgeable, market participant.
Such an approach could still chill business conduct, however, since algorithms may select the provider’s own product regardless of whether the provider sought this outcome or not. To avoid such chilling effects, the author suggests a mechanism under which antitrust liability for algorithmic result is excluded for those providers that: (i) are able to demonstrate the testing and mitigation of possible bias in their algorithms; (ii) have in place a dedicated procedure to receive notices of discriminatory treatment and respond timely with an explanation; and (iii) submit to an independent dispute resolution system to resolve any controversy arising from such notices.
I really enjoyed the intricate – and rather sophisticated – dissection in this article of the various meanings and roles that intent can play in establishing a competition infringement. I cannot do the discussion justice here, largely because it involves going in detail over various judgments which are ambiguous – and at times outright contradictory – on this matter, and trying to make sense of them. Inasmuch as this is one of the main, if not the primary, role of legal scholarship, the paper is very good, even if I think the argument could be expressed more succinctly.
At the same time, I was slightly confused by an argument developed during the discussion of Google Shopping. There, the author seems to be saying that: (i) the decision seems to create strict liability for algorithmic design; and (ii) that the decision was based on evidence of subjective intent that does not fulfil the objective element necessary to establish negligence. The latter may indeed amount to an error of law on the part of the Commission, even if I do not think this is one of the arguments being made by Google to avoid being subject to a fine in the event the European courts find that its conduct was anticompetitive. However, the decision is either an instance of strict liability or was adopted on the basis of subjective intent – I fail to understand how these two propositions can coexist.
I would have thought that, instead, the problem regarding demonstrating negligence is that this requires a mental state towards adopting a conduct that infringes the law – which is hard to do when the theory of harm is not clear. From this perspective, the problem with the Commission decision – assuming there is one – would not relate to intent at all, but would be instead an incidental consequence of the Decision’s ‘original sin’ of being unclear on what the infringing conduct actually was.
This, in turn, goes to another distinction that I think the author could have drawn more clearly: even if a business conduct is ‘objectively’ abusive, for it to be subject to sanctions the infringing party must at the very least also have been negligent. The paper at times gives the impression that intent and negligence are necessary for any practice to be deemed abusive – and, in effect, the author seems to argue in support of this approach early in the paper, in particular by linking it to a duty of care not to breach the special responsibility incumbent on a dominant company – while I am not sure this is necessarily the case. In any event, I would have liked to see this question discussed in more detail.