E-disclosure: Training Technology

April 16, 2012

Are you one of the many lawyers trying to reduce your client’s review costs who have considered using computer assisted review but who have been deterred by the lack of judicial endorsement for the technology? If so, you may have been afforded some comfort from the statement in Da Silva Moore v Publicis Groupe & MSL Group 11 Civ 1279 by US Magistrate Judge Andrew Peck in February 2012 that ‘Counsel no longer have to worry about being the “first” or “guinea pig” for judicial acceptance of computer assisted review……Computer assisted review now can be considered judicially approved for use in appropriate cases’.  

In the UK, Practice Direction 31B was introduced in 2010 to encourage parties to manage electronic documents ‘efficiently in order to minimise the cost incurred’ and to use a broad range of evolving technologies to assist in this process (Practice Direction 31B, 6(1) and (2)). Predictive coding or computer assisted review is one such technology. Put simply, predictive coding is a process by which software is trained by a senior lawyer (the ‘expert’) who is familiar with the issues of a matter. The expert reviews a representative sample of the documents and registers his or her decisions upon the relevance or otherwise of each document reviewed on the system. In this way the expert ‘trains’ the technology to recognise responsive patterns.  

Judicial acknowledgement in the UK has been forthcoming, albeit falling short of direct endorsement. For example, in the case of Goodale v Ministry of Justice [2009] EWHC B41 (QB), Senior Master Whittaker said ‘..when it comes to review I am aware of software that will effectively score each document as to its likely relevance and which will enable a prioritisation of categories within the entire document set .’  

In the US, the situation up until very recently was summarised by Magistrate Judge Andrew Peck who recognised that ‘‘Until there is a judicial opinion approving (or even critiquing) the use of predictive coding, counsel will just have to rely on [this article] as a sign of judicial approval. In my opinion, computer assisted coding should be used in those cases where it will help ‘secure the just, speedy, and inexpensive’ (Fed.R.Civ.P.1) determination of cases in our e-discovery world’ (see L Tech News, October 2011 at 25, 29). 

The momentum towards judicial acceptance of computer assisted review has continued to build on both sides of the pond. Moreover, a willingness to test alternative working methods which might reduce client concerns about spiralling review costs and enable lawyers to comply with their obligations to carry out disclosures efficiently and proportionately in compliance with their legal obligations has ultimately created a perfect storm.  

That perfect storm resulted recently in the opinion and order by Magistrate Judge Peck in Da Silva Moore v Publicis Groupe & MSL Group. In this case, the parties had agreed to the defendant’s use of predictive coding technology. However, disputes arose as to the scope and implementation of the technology. The court was required to rule upon the dispute, and also therefore upon the acceptability of the use of computer assisted review. Judge Peck formally approved computer assisted review for use in appropriate cases and referred to his comments in an article in which he prescribed how the parties might ensure that the use of predictive coding is defensible in the following terms:  

 ‘…I may be less interested in the science behind the “black box” of the vendor’s software than in whether it produced responsive documents with reasonably high recall and high precision. That may mean allowing the requesting party to see the documents that were used to train the computer assisted coding system…’  

The plaintiff has appealed Judge Peck’s decision on several grounds and the case has raised a number of interesting issues:  

1. Methodology:  

Predictive coding techniques are applied by different providers in different ways and this leads to confusion. Some require the experts to review documents which have been randomly selected by the software while others require the expert to feed a sample of relevant documents into the technology. Once the technology is trained, it applies the results of the sampling across the entire collection of documents, producing a percentage responsiveness rating for each of the documents to be reviewed. The benefit of using predictive coding software is that it is able to prioritise a large body of documents, thereby presenting those documents which are most likely to be relevant to the front of the review queue, at a fraction of the time and cost of a traditional linear document review.  

However, the arguments raised by the plaintiff’s appeal in Da Silva Moore (questioning, for example, whether the final random sample size of the ‘deemed-irrelevant’ documents was statistically significant and would enable meaningful estimates of the system’s recall to be assessed) exemplifies the continuing debate which exists about the correct methodology which should be deployed when training the software and establishing a defensible approach.  

2. Rule change required?  

Prior to the plaintiff’s appeal, the Da Silva case suggested that the question in the US was not whether parties should use predictive coding technology in appropriate cases – that was a given – but rather how parties should use it. If Judge Peck’s decision is upheld, argument will no doubt ensue in the US as to the efficacy of the requirement to disclose to opponents both relevant and non relevant documentation used to train the software.

If there was a similar judgment in the UK, it is anticipated that, in the absence of a rule change, any such requirement would encounter stiff resistance. For there is currently no obligation upon UK litigants who are subject to a standard order for disclosure to disclose anything other than those documents on which the part relies, documents which adversely affect the party’s case and documents which adversely affect or support another party’s case. The idea that, in order to utilise predictive coding technology, litigants must volunteer to extend their disclosure obligations and share documents with opponents which, whilst irrelevant to the litigation in hand may be relevant to future or adjacent proceedings is likely to be an anathema to litigation lawyers and the judiciary may need to revert to the use of their overriding discretion in order to impose the requirement. Indeed it may inadvertently delay acceptance and use of technology which would otherwise assist litigants to reduce legal review costs and increase efficiency – the very object of Practice Direction 31B and Lord Justice Jackson’s Review of Litigation Costs.  

This is by no means the end of the debate. In the US, the plaintiff’s objection to Judge Peck’s ruling has been lodged and we await Judge Carter’s ruling with interest. However, in light of the judicial focus upon increasing efficiencies in disclosure – both in the US and the UK – it is likely that we will see a number of judgments in the coming months which seek to clarify the means by which litigants can demonstrate that they have employed defensible processes when using predictive coding technology. 

Greg Wildisen is International Managing Director at Epiq Systems: www.epiqsystems.com