Out-Law / Your Daily Need-To-Know

High Court orders in favour of e-disclosure predictive coding in first contested case

Out-Law News | 25 May 2016 | 12:28 pm | 2 min. read

The High Court has backed the use of predictive coding in a litigation disclosure exercise, in what is being reported as the first use of the technology without the consent of all parties.

The order, details of which have been published in the legal press, comes shortly after the High Court first sanctioned the use of the technology. In February, Master Matthews set out ten factors in favour of the use of predictive coding, "almost all" of which applied in this case, according to the law firm that made the case for it.

Litigation expert Richard Dickman of Pinsent Masons, the law firm behind Out-Law.com, welcomed the order. Disclosure "can be one of the most expensive and time-consuming stages in High Court litigation", particularly in cases requiring review of many thousands of electronic documents, he said.

"Any technology which assists in managing disclosure more efficiently and cost-effectively will therefore be welcomed by companies having to bear the costs of litigation," he said. "The Pyrrho case in February demonstrated how the judiciary can encourage the use of such technology without being too prescriptive. However, there the parties had agreed to the use of predictive coding. Here, they had not."

"What is perhaps most surprising is that it has taken until this year for the English courts to address the use of predictive coding. After all, the technology has been available for many years, and is already widely used in the US. Lawyers and companies who fail to embrace this technology may find it imposed upon them. However, predictive coding is not a panacea – it involves significant upfront costs, although it should save costs later. Its use has to be assessed on a case-by-case basis," he said.

In particular, Dickman suggested that there was a risk that the "pool" of documents requiring review in any given case could end up growing "on the basis that the technology can carry out the review at a fraction of the cost of human reviewers" as use of predictive coding became more widespread.

"This could potentially undermine the very costs and efficiency savings the technology offers," he said.

Predictive coding refers to the use of software to review the majority of electronic documents in a disclosure exercise, rather than human beings. The software is programmed to account for findings made by senior lawyers who have reviewed a sample of the documents, who then 'train' the document review platform on what documents are likely to be relevant.

Having been 'trained', the software can then analyse the remaining documents and identify their relevance to the issues in the case. Senior human reviewers will then review further samples as a quality control process, to ensure that the software has correctly identified the relevant documents and further 'train' the software until an acceptable level of accuracy is reached.

Predictive coding has been used for e-disclosure purposes in the US. Last year, the High Court in Ireland approved the use of the technology in a contested case.

Writing for Lexology, the law firm promoting the use of the technology said that the cost of traditional document review in this case would be more than two and a half times that of predictive coding, given the number of documents to be reviewed and its ability to run the predictive coding workin-house. Its client is seeking a buy-out of his minority shareholding in  an unfair prejudice petition, something that is "strongly contested" by the respondent, the law firm said.

Litigation expert Michael Fletcher of Pinsent Masons said that one of the interesting things about this case was that the court had backed the use of a form of predictive coding which "requires significant training of the system up-front".

"There is a separate form, where the review platform continuously learns from the reviewer's decisions and uses those suggestions to propose further documents for human review. This does not involve the coding of documents – human reviewers do that, but they code documents which the machine proposes. This form may actually be more robust as the machine constantly adapts to the decisions being taken, rather than relying on the initial training, until it can find no further potentially relevant documents for human review. That the court has now approved the first form may therefore encourage use of either format," he said.