29th March 2016

Rise of the Machines – the future of Electronic Disclosure

At a glance

For litigation cases involving large amounts of electronic disclosure, predictive coding could save countless hours of document review time. But although the technology is well-established in the US, a recent High Court case marks the first time the English Courts have approved the use of predictive coding. Michael Axe explains what this means for the future of electronic disclosure in English litigation and how it could signal a greater willingness to allow automated software to carry out the majority of the document review process.

Electronic disclosure (or e-disclosure) is the umbrella term used to describe the disclosure of electronically stored information (ESI). Whereas a decade or two ago, most or even all of a company’s documents that needed to be disclosed in Court proceedings would be paper documents, today the vast majority of a business’ documents are held electronically. And whereas when the documents were paper-based there may, in most cases, have been a few hundred documents to review to sort the relevant from the irrelevant, with ESI there are often several million documents that need to be reviewed.

Early attempts to use specialist software to assist with the “human” part of the ESI review process centred around “de-duplication” technology (which, as the name suggests, would remove duplicate copies of emails, Word documents and so forth), and around “keyword searches” (where the software would search for documents containing one or more keywords suggested by those knowledgeable in the facts of the case). The keyword searches were helpful in reducing the number of potentially relevant documents that then needed to be manually reviewed by a human, but the searches were not always particularly sophisticated. As a result, many irrelevant documents would still be included in those for manual review.

So what is “predictive coding”?

Predictive coding is essentially the next evolution of keyword searches.

Instead of drafting a list of keywords for the software to search, the human reviewer reviews a sample set of documents and then categorises (or codes) them based on a number of factors, such as their relevance to the issues in dispute, the author of the document, where the document was stored, and its date of creation. The coding for the sample data is then entered into the software, which uses the coding to “learn” which documents are likely to be relevant and which are not. This sampling process can be repeated several times with other sample sets of documents and each time the software is likely to increase the accuracy with which it can predict the degree of relevance of the remaining documents in the whole set.

The software is then able to “score” the documents for their likely relevance to the issues involved in the case and the human reviewer can limit their manual review to only those documents which the predictive coding software considers are likely to be relevant (ie scored highly), thereby saving considerable document review time.

Pyrrho Investments Ltd v MWB Property Ltd

The Pyrrho case was the first reported instance of the English High Court permitting the use of predictive coding in an electronic disclosure exercise and it is a useful illustration of how these technologies work in practice. It is not, however, the first use of predictive coding on English soil, as the technology has been around for quite some time.

In the Pyrrho case, one of the parties originally identified 17.6 million potentially relevant documents that needed to be reviewed. This figure was reduced to 3.1 million by using de-duplication technology, but it would still have taken countless man-hours to manually review the remaining documents. The Court therefore approved the use of predictive coding to sort through the 3.1 million remaining documents for the following reasons:

  1. Past experience in other jurisdictions (most noticeably America and Ireland) had suggested that predictive coding was a useful tool in appropriate cases (emphasis added)
  2. There was no evidence that predictive coding was any less reliable than either human/manual reviews or keyword searches (and in fact, there was some evidence that predictive coding was more reliable)
  3. There would be greater consistency in the review process if it was carried out by the software applying the coding of one senior lawyer from a sample set, rather than a manual review by dozens, if not hundreds, of junior lawyers each applying their own independent assessment of relevance
  4. Predictive coding would allow the documents to be reviewed at a cost proportionate to the value of the dispute, whereas the cost of a full manual review of the documents could never have been justified in this case
  5. Both parties had agreed (subject to the Court’s approval) to the use of predictive coding, and as the trial itself was still over a year away, there would still be the opportunity to consider other review methods if for some reason the predictive coding method proved unsatisfactory.

The future of e-disclosure

The Pyrrho case is important, as it is the first time that the English High Court has been asked to consider, and ultimately permitted, the use of predictive coding in an e-disclosure exercise, and it signals a greater willingness to allow automated software to carry out the majority of the document review process.

Jonathan Maas, a Senior Director at Consilio and a leader in the field of electronic disclosure, takes this view:

“There are a number of interesting points to take away from this judgment:

  1. The parties had reached agreement as to how to proceed
  2. The parties were confident enough with the technology to advance cogent arguments to the Court in favour of predictive coding
  3. The judge was open-minded enough to accept those arguments
  4. The technology used on this case is not the only type of predictive coding available
  5. The workflow set out by the judge in his judgment is appropriate in this case with the types and number of documents involved
  6. Not all cases benefit from the application of predictive coding.

This judgment advances the cause of the sensible use of appropriate technology to keep the costs of litigation to a proportionate level. It also shows how clients can benefit when the overriding objective is allowed to override.”

However, the Courts are arguably still somewhat behind the times. Predictive coding technology has been around for many years and, while the software continues to become more and more sophisticated, it may well soon be replaced by the next evolutionary step in e-disclosure technology.

For more information on this or any other issue relating to electronic disclosure, please contact Michael Axe or speak to your usual contact in the Commercial Disputes Team.

This document is provided for information purposes only and does not constitute legal advice. Professional legal advice should be obtained before taking or refraining from taking any action as a result of the contents of this document.