Part of my role is to help those responsible for eDiscovery / eDisclosure to identify the products and services which they might consider when deciding on the “tools and techniques” (as the UK Practice Direction 31B puts it) to be used to reduce document volumes to a manageable size and to review them. I don’t give system selection advice, for obvious reasons, but I do like to identify the product descriptions and ancillary materials – articles, papers etc – which help to throw light on the wider subject beyond the product itself. We are seeing some good materials coming out of providers at the moment, and I am working my round some of them; you would not thank me if I served them up all at once.
FTI Consulting, and its technology segment FTI Technology, have just announced their new Predictive Discovery solution. That is interesting enough, but what makes it more so is the material which FTI has published at the same time in order to make the subject more comprehensible to lawyers. Part of the value in the surrounding materials (which goes beyond their application to FTI’s own solution) is the way they address the “black box” problem – the glib expression from lawyers who don’t feel up to the intellectual effort of trying to work what what this kind of software does and how it can benefit their clients and themselves.
Let us look first at FTI’s new Predictive Discovery solution. It is a managed service, founded on new technology, which brings together lawyers, statisticians, technical experts and a set of processes which together work with the client to produce a result which is accurate, defensible and cost-effective. The consultative approach is important: the FTI consulting team works with the lawyers to review a subset of a document collection, making relevance and privilege decisions to develop a training set which is used to score documents for those attributes. The lawyers can then use the advanced analytics in FTI’s Ringtail software to verify the results. That same iterative approach continues as the review progresses, with a mixture of statistical sampling and by-eye verification from the lawyers. This iterative element is critical – humans have every opportunity to cross-check software conclusions and software functions can monitor manual decisions to identify, for example, inconsistent conclusions reference to other things known about documents of the same kind.
The service can be customised to suit the case; prioritisation based on relevance is one obvious function, but the technology and processes can be used as a means of validating decisions made by other processes, to cull down obviously irrelevant material and to check incoming productions, amongst other things.
A press release gives limited scope for explanation, and FTI has been working hard to foster understanding of predictive coding generally as well as writing about the scope of their service. This output includes an article in the FTI Journal by Senior Managing Director Joe Looby called Taking Predictive Coding out of the Black Box. It is an extremely helpful article, neither neither requiring existing knowledge nor patronising the knowledgeable reader, and it is illustrated with helpful diagrams.
That is backed by a webcast which I promoted at the time of its live broadcast and which is still available for download. In addition to Joe Looby, the speakers are the well-known Jason Baron, Director of Litigation at the Archives and Records Administration and Daniel Slottje who, in addition to being a professor, economist and statistician at the Southern Methodist University is a Senior Managing Director in FTI Consulting’s Economic Consulting Services practice. The article and webinar together serve as a comprehensive review both of the technology aspects of predictive discovery and of the recent developments in the courts.
Lastly on this subject, Joe Looby recently gave an interview to Metropolitan Corporate Counsel which explains in more detail what FTI’s offering consists of and expands on the point that there are multiple use cases which lawyers should think about when considering predictive coding.