Dr Arianna Andreangeli, Senior Lecturer in Competition Law, Edinburgh Law School, University of Edinburgh — Supplementary written evidence (NTL0039)

Summary

The growth of e-commerce has revolutionised how the business world works today.  The extensive use of tools such as pricing algorithms, personalised advertising and automated payment processing has brought benefits in terms of efficiency in commercial transactions, consumer choice and potentially, lower prices and better services.  However, it has also created new challenges for the agencies that are competent for the application of the rules governing the behaviour of economic entities, most notably competition law.

The expansion of electronic communications and the digitisation of economic exchanges raises important questions as to how the competition rules can be effectively applied.  The role of prediction machines, coupled with the wide availability of storage facilities for increasing amounts of data, means that commercial entities can rely on algorithms for the purpose of the supply of goods, the provision of services and the processing of transactions.  To respond to these new challenges competition agencies such as the Competition and Markets authorities have relied increasingly on artificial intelligence tools in the course of their investigation and detection of prima facie anti-competitive behaviour.  Textual analysis tools have proven extremely helpful in the “sifting” of the considerable amount of electronic evidence associated with potentially unlawful arrangements.  However, they present unique challenges for the overall fairness of investigations, to the extent that their usage can adversely affect the ability of the investigated parties to have a reasonable opportunity to understand the charges made against them, to appreciate the assessment of the evidence made by the competition agencies and ultimately to build their own defence against these allegations.

E-disclosure, namely the process through which litigants gain access to digital evidence that can either be against their claims or aid their case against the other and that the latter may possess, is regulated quite extensively in the judicial context: in civil cases, Practice Direction 31B, adopted pursuant to the Crown Procedure Rules, provides a set of procedural steps to ensure fairness and transparency when it comes to searching, sifting and identifying digital documents for the purpose of disclosure, under judicial supervision.  In criminal cases, there are also specific rules, contained in the Criminal Procedure Rules 2015 and in Annex A to the Attorney General Guidelines on Disclosure.

In the context of public competition enforcement, namely of the activity of investigation and administrative decision-making on allegations of anti-competitive conduct, instead, there are no precise rules or guidelines that explicitly discipline how the Competition and Markets Authority deploys textual analysis algorithms or similar automated search software in the context of its analysis of the evidence gathered in the course of the proceedings.  It is however argued that the lack of express procedural guidelines in this area sits uncomfortably with basic requirements of procedural fairness, enshrined in Article 6(1) of the European Convention on Human Rights, as interpreted by the European Court of Human Rights.  It appears also to be difficult to reconcile with the commitment to openness to which all public authorities are bound on the basis of the Seven Nolan Principles on Public Life.

Accordingly, it is suggested that the CMA should include, in its Guidelines on Cartel Investigations (CMA 8), express guidelines as to how it intends to deploy textual analytics and other artificial intelligence tools in the sifting of digital documents, for the purpose of the latter being included in the case file and analysed as evidence of prima facie anti-competitive conduct.

These new guidelines could encompass: 

-       An express commitment to open discussion with investigated parties of the possibility that automated search tools, such as analytical methods and other textual analysis software, may be used;

-       If these search tools are used, the CMA officers should be transparent with the investigated parties as to their approaches in sifting through and identifying potentially relevant documents.  These exchanges could take place at state of play meetings and should involve a discussion of the methods being used in this context—for instance, the nature of software used, how document samples have been identified for the purpose of training the system, what key words have been used and how documents have been ranked and, on that basis, identified for the purpose of being examined and potentially acquired into the file.

-       Once a decision is adopted, the latter should provide a statement detailing how documents have been searched for the purpose of facilitating the judicial scrutiny of the decision in all its aspects, including those of an evidentiary and procedural nature. 

13 September 2021