Predictive coding, a type of machine learning technology, has the potential to dramatically reduce the cost of the e-disclosure process. It has been gaining wider acceptance in the UK, although it has been used for several years in the US.
The first formal judicial endorsement of the use of predictive coding in e-disclosure came from Master Matthews in his High Courtruling in the Pyrrho Investments v MWB Property Ltd case. Noting the excessive cost and time required to manually review documents, the judge approved predictive coding as acceptable for navigating the millions of electronic documents involved in the case. This landmark decision is likely to impact e-disclosure costs and proportionality arguments in UK courts, and increase the willingness of legal teams to use predictive coding in documentreviews as part of a wide range of disputes and investigations.
With document sets growing exponentially, lawyers are struggling to get through the data in an efficient and defensible manner. Predictive coding works by analysing the coding decisions made on a sample document population, and extrapolating those decisions across the wider population. Uses for this technology range from review prioritisation to the automated culling of non-responsive documents. While predictive coding can help make review calls which are traditionally made by humans, lawyers must select tools that allow control and provide transparency into how results are obtained.
Another tool in the armoury
Now that the UK courts have accepted the use of predictive coding software, it is crucial that lawyers understand how to apply it and determine which type of case will benefit most. It is not a one-size-fits-all solution, and may need to be bypassed or used in conjunction with other tools and processes for certain cases.
Itis one of many technologies that can streamline document review. However, when appropriate, it can offer a great deal oftime and cost savings due to the reduction in overall billable hours.
In the Pace Law School paper, ‘Inconsistent Responsiveness Determination in Document Review: Difference of Opinion or Human Error?’, Maura Grossman and Gordon Cormack conclude that inconsistent responsiveness coding regarding the same documents is due in large part to human error. By contrast, with predictive coding, the algorithm can be relied upon to make the same call time after time, based on the guidelines it has been given. Its ability to automate much of the document review process can increase efficiency, improve accuracy and quality control, and reduce overall review expense.
Inside the black box
Many lawyers fear that predictive coding is a black box, so whichever platform is being used, it must have strong visualisation capabilities and transparent workflows to aid in the understanding of the document population and process. Document review platforms, such as FTI Technology’s Ringtail, can deliver visual analytics that quickly summarise the content of the included and excluded document sets and provide full transparency throughout the process, with each iteration tracked and displayed.
It is equally important to involve experts who can offer technical experience, guide the legal team through the process and help them become comfortable with the technology. Predictive coding is a process that requires legal, technical and statistical expertise. The right combination of people and transparent processes will ensure defensibility and visibility into the results. Involving an expert third party can assist with decision making around which issues in the case would benefit from predictive coding, assigning the number of documents to review in the training set and designing iterative workflows. These professionals can support the team and testify to the defensibility of the process credibly.
Document review is evolving
While predictive coding is gaining acceptance in the UK, it received court approval in the US in 2012, and its adoption has been growing ever since, particularly on matters involving large document populations. As with all new technologies, it is important that the people using this software have a broad understanding of how it works, its benefits and how best to achieve the desired goals.
Originally published in The Lawyer (April 25, 2016)