The UK Competition and Markets Authority (CMA) made a splash with the publication earlier this year of its paper ‘Algorithms: How they can reduce competition and harm consumers'. The paper was published alongside a call for information (now closed).
While we wait for the response, the CMA's originating paper is worth a fresh look.
Glass half empty?
The paper focuses on the harms the algorithms can (potentially) cause to competition and consumers. A number of “direct harms to consumers” are identified, including:
- Personalised pricing: advertising different prices to different people and using practices which achieve the same effect, e.g. discounts to selected customers
- Complex and opaque pricing techniques: using a wide range of consumer data to inform themselves of their willingness to pay
- Personalised rankings: firms may use information about the user beyond the search query to decide which results to display and in what order
- Recommendation and filtering algorithms: systems affecting choice architecture by pre-determining the range of options available to consumer
- Manipulating user journeys: using variety of techniques to predict likely rating that user would give
- Unfair ranking and design: firms not transparent about criteria they use to organise options to best meet consumers' needs
A number of these harms amount to so-called “dark patterns” which influence consumer journeys through the interface design choices used by particular websites. The CMA is particularly focussed on where such design choices are deceptive and “trick” the user, and gives the example of scarcity messages which can create a (false) sense of urgency on the part of consumers.
However true it is that such harms do arise, one potential risk of the CMA's overall approach is that it treats as a given that algorithms can also do an array of useful and highly innovative things – many of which are now fundamental to the way we live and work. The CMA's ultimate approach needs to ensure that those positive attributes are recognised and respected.
Theories of harm, not legal theory
Another feature of the paper is that it focuses on the factual issues which may cause harm to consumers, rather than analysing in detail how such harms will contravene any existing laws. This is not an uncommon approach, including for other agencies and in other sectors – for example, the Commission's 2009 Pharmaceutical Sector Inquiry also focussed on the factual situation on the market rather than on the application of competition law, which came subsequently. (Of course, this paper has not been conducted under the CMA's formal powers to carry out market studies.)
Nevertheless, the CMA does focus primarily on harms that can be addressed through the enforcement of consumer and competition law. In particular, the paper lists a number of potentially exclusionary and collusive practices (some of which rely on Commission decisions which remain subject to EU appeal). The limited remit of the paper is noted, with the CMA stating that “the use of algorithmic systems can cause harms to people in their role as citizens, and not just in their role as consumers or business owners”.
The consideration of the topics relating to collusion, a central concern of competition law, are predominantly future-facing, with reference being made to “simulation studies” and “theoretical concerns” of autonomous tacit collusion. It remains an open question whether or when it could be legitimate for a competition authority to rely on observations of market outcomes to infer that algorithms have “colluded” and if they have, what liability the companies making use of those algorithms should bear.
Ultimately, many of the topics under discussion may require new regulation to resolve – for example, a “fairness by design” requirement for software could seek to avoid the dark patterns mentioned above, and anti-collusion mechanisms may be required if it emerges that algorithms really are capable of learning to collude without the input of the companies which install them.
In the meantime, the CMA's response to the consultation when it is published will be a guide to the views of industry on the topics addressed in the algorithms paper.