ARTIFICIAL INTELLIGENCE IN DISPUTE RESOLUTION
Stories about the inevitable and relentless disruption artificial intelligence ("AI") will cause to established industries are commonplace. A recent Financial Times article was entitled: "Artificial Intelligence closes in on the work of junior lawyers".
However, before aspiring lawyers start wondering whether their hard-earned skills will soon be redundant in the modern world, it’s worth thinking about the effects that AI has already had on running large-scale disputes.
Disclosure is typically the most expensive phase of large-scale litigation cases. At the appropriate stage of the proceedings, a party to a Court dispute has a duty to carry out a reasonable search for, and then allow the other parties to inspect, documents which support or adversely affect their own or another party's case (i.e. documents which are "relevant" to the dispute).
Emails, chats, recorded telephone lines and more sophisticated electronic storage systems have exponentially increased the amount of material which is potentially relevant to a dispute. Further, when there are tens or hundreds of millions, or even billions, of pounds at stake in a dispute, judges have not been sympathetic to arguments that extensive searches for relevant documents are not reasonable in the circumstances.
Trainees and junior lawyers in busy dispute resolution departments could easily spend a large proportion of their time reviewing emails, chats, telephone calls and other communications to determine whether they were relevant to the dispute and therefore needed to be disclosed. This is a crucially important task but, as the number of documents potentially relevant to a large dispute could be in the millions and each needs to be manually reviewed by a trainee or junior lawyer, it could be an incredibly time-consuming process.
“Predictive coding is a form of technology assisted review which helps identify relevant documents based on an algorithm, which uses machine learning to suggest the relevance of each document.”
However, recently AI – in the form of predictive coding – has been used to ease some of the burden. Predictive coding is a form of technology assisted review which helps identify relevant documents based on an algorithm, which uses machine learning to suggest the relevance of each document.
How it works is that senior team members with extensive case knowledge (typically senior associates) will review a sample set of documents. The algorithm will then promote, to a specified confidence level, the documents it believes are likely to be relevant (e.g. it might have 95% confidence in its categorisation of one document but only 20% in another). As junior members review these documents, the machine continuously learns, changes its assessments of the other documents and ascribes relevance with higher levels of confidence. Once sufficient iterations have been undertaken, it is possible to exclude documents, which the coding suggests with a high level of confidence, are not relevant. Properly trained predictive coding has been shown to be more accurate than human reviewers and the costs for clients are significantly lower – particularly in large, complex disputes.
This technology has been available for some time now. HSF has been using the technology for several years, either for early case assessment (identifying the most relevant documents early on in a case) or to check whether there are potential errors in the categorisation of documents. However, recently (in May 2016) the English High Court ordered that predictive coding should be used for the purposes of disclosure despite one of the parties arguing that it should not be used. This comes after a previous English High Court endorsement for the technology earlier that year. Courts in the US have been even more proactive in ordering its use for discovery (the US equivalent to disclosure). It seems inevitable that predictive coding will continue to become more widespread in disputes, especially as the technology continues to become more and more accurate.
However, this does not mean extinction for the junior lawyer. Predictive coding will help sift out the most irrelevant material so that junior lawyers do not have to waste their time. But the most important legal work – finding the documents which help or harm your case and using them to craft the evidence and the advocacy – will still need to be done by someone with sound legal skills and judgment. To the contrary, it could be argued that the rise of predictive coding has in fact improved the quality of work that junior lawyers perform on large cases – by spending less of their time on reviewing irrelevant documents and more time conducting analysis, legal research and drafting evidence or argument.