Predictive coding is one of the more recent developments in technology assisted review ('TAR') in the context of large-scale electronic disclosure which has become increasingly common in modern litigation fuelled by the digital revolution and constant generation of digital records. Keyword searching (a more established form of TAR) has been used in electronic disclosure exercises for many years.
The recent cases of Pyrrho Investments v MWB Property Ltd  EWHC 256 (Ch) and Brown v BCA Trading Ltd  EWHC 1464 (Ch) suggest that predictive coding will play an increasing role in e-disclosure over the coming years.
Predictive coding is an umbrella description given to a number of different forms of technology and workflows designed and promoted by different providers and applied in alternative ways. At its heart however, predictive coding involves the application of technology review software and sophisticated algorithms to an initial core set of human reviewers’ decisions in order to “train” the system to conduct an automated review of the documents. The application of predictive coding can be extremely helpful in reducing from the universe of documents collected, large numbers of irrelevant documents, whilst identifying those which are more likely to be of potential relevance (or privileged, confidential or sensitive) which can be fed through to a linear human review.
Whilst predictive coding is a familiar form of TAR in the United States, it is only recently that its use has been endorsed by the High Court in England, in the Pyrrho case in February 2016.
In that case, the court confirmed that:
- predictive coding can be helpful in appropriate cases and is used in other jurisdictions;
- there was no evidence to suggest that it was less accurate than other more established forms of TAR;
- there was nothing in the CPR to prevent its application;
- it can often (and in that case did) result in cost-saving and ensure that the electronic disclosure exercise remains proportionate by reference to the case overall; and
- the parties had agreed to the use of predictive coding.
The more recent case of Brown in May 2016 goes one step further in confirming that predictive coding may be ordered despite a party’s objection to its application. The court’s decision was influenced on this occasion by the relative cost savings which were said to be achievable through the application of predictive coding over the traditional application of keyword searches. The court also took particular consideration of the views of the party on whom the disclosure obligations were likely to be most heavily felt: in that case (an unfair prejudice petition) the Respondents, who held the majority of the documents. Whilst their views would not be determinative, the court would certainly take them into account in directing what process would be most reasonable and proportionate.
The familiarisation of practitioners and the courts with predictive coding and other forms of TAR is welcome. Electronic disclosure is, regrettably, generally the scourge of modern litigation and absent any wholesale reform of the disclosure process (which looks unlikely for the immediate foreseeable future), assistance through TAR and predictive coding in particular is essential to ensure that modem large scale litigation and electronic disclosure can be properly managed and at reasonable and proportionate cost.