I use the term “predictive coding” in this article because that is the term used a) in the relevant judgment, b) by BLP whose successful use of the technology is the subject of the article, and c) by FTI Consulting, whose Ringtail eDiscovery software was used. I wrote about the varied terminology (TAR, CAR, etc) in my article on Triumph Controls. Whatever you call it, we are talking of a system which “harnesses human judgments ….on a smaller set of documents and then extrapolates those judgments to the remaining document collection [Grossman and Cormack, TAR Glossary, 2013].
2016 saw two judgments in the courts of England and Wales in which the use of predictive coding was upheld, both from disclosure applications rather than at trial. The first was Pyrrho which, since the parties consented to the use of the technology, is more an example of the wisdom of cooperation than a clear sign of court approval – once the parties had agreed and explained their positions cogently to the court, it was unlikely that the master would have declined to convert their agreement into an order.
The second was BCA Trading which drew less attention than Pyrrho but which is more significant because BLP’s proposed use of predictive coding was hotly contested. The registrar inevitably drew on Master Matthew’s careful analysis in Pyrrho and said “I reach the conclusion based on cost that predictive coding must be the way forward”.
BCA Trading is back in the news again, following a 12 day trial at which BLP’s client was successful. BLP has written about it in an article called BLP wins case for BCA using predictive coding in disclosure. It says that “the disclosure provided was thoroughly tested at trial and the judge relied heavily on it for his findings”. I have not seen the judgment, but I understand that there is little in it about the discovery process – nor would one expect there to be in a judgment after trial.
To save me reciting the facts, I can refer you to articles by Litigation Futures and Artificial Lawyer. Both take the line that the use of predictive coding all the way to trial will encourage wider use of the technology; phrases like “a watershed moment for the technology in the UK”, “machine-learning breakthrough”, and “big future for predictive coding” are among the phrases used in various articles. I don’t disagree with any of that, but it is worth digging a bit deeper than saying “the winning party used this technology, so this technology is great”. It is great – but what makes it so? What did predictive coding do for BLP and its clients that other tools might not have done?
The key points lie in BLP’s own article. Some quotations:
The petitioner, having resisted [the use of predictive coding] throughout, could find only minor ways to challenge its output, resulting in discrete further searches being performed by BLP.
[the case contained] broad allegations of unfair prejudice and bad faith. These issues did not allow the formulation of a sensible set of key words for a traditional disclosure review.
[The algorithm produced] documents that were highly relevant to the issues, many of which would be unlikely to have been retrieved using a proportionate set of key words in a traditional process.
the computer was able to “learn” from the sample reviews performed by senior BLP lawyers in order to “educate” it.
These points lie at the heart of any debate about the propriety of using predictive coding (assuming it is used properly, as it would be at a firm like BLP). Surviving challenges from a resisting opponent matters more than any theoretical arguments before disclosure; finding documents which would have defied a keyword search is the best answer to challenges; the tests were established by the work of senior lawyers – not just humans but skilled ones who used FTI’s technology to augment their work, not to be a substitute for it. These points are more significant than the headlines imply.
BLP is one of the few firms which consistently emphasises its use of technology to bring down costs and help its clients succeed. What is interesting here is not that the firm actually uses the technology – many others do that – but the way in which BLP makes use of that fact to promote (in this case) its Litigation and Corporate Risk (LCR) practice.
Does it matter whether the work is done within the firm or by an external provider, perhaps FTI itself, whose reputation lies as much in its consulting service as in its software? BLP thinks it does. Its article about BCA Trading says:
Our ability to handle all e-disclosure work in-house enables us to shape the exercise in a more streamlined way than when third parties are involved,
Nick Pryor is Head of LCR Client Technology, not a role found in every firm, and I spoke to him about the case and about BLP’s use of technology. He said:
“This result is a great success for our client, and validates BLP’s strategic commitment to keeping the tools, people, and processes in-house. Having our technical and legal experts working alongside each other was key to getting the most out of the predictive coding tools on this case.”
We heard much the same from Eversheds Sutherland (I wrote about them and their use of OpenText here), and there is no doubt that the message is compelling for many clients. That should not put off firms who have no taste for the investment (in people as much as in technology) or the responsibility. Particular firms, and BLP is clearly one of them, can attract clients with their technology tools and skills, but many clients don’t mind where the work is done or with what technology. Bruce Braude, Head of Strategic Client Technology at BLP, was quoted in last year’s FTI / Lawyer Global Litigation 50 (I wrote about that here) as saying that he was not seeing clients necessarily expressing a preference as to what the tools are – “they’re concerned with the output rather than the model”.
The outcome of the BCA Trading case is important for its own sake, for the reasons I identify above. It is important also, for taking us back to the 2016 disclosure application which is, in many ways, more important than Pyrrho because it was contested and because of the registrar’s observations about proportionality and cooperation.
If you want to use predictive coding (or whatever you care to call it) you will rely on this case to persuade your clients, your opponents and the court that you should be allowed to. If you oppose its use, the case will present hurdles which you must get over. In a sense, the argument has moved on, as it has in the US. The question is not whether you use predictive coding but how you do it, how you articulate the benefits, how you draft your protocol and, not least, how well you set out the costs and proportionality arguments.
There is another aspect to all this which is nothing to do wth court approval or the outcome of applications and trials. At the end of BLP’s article about BCA Trading is this paragraph:
Our disclosure technologies are now being harnessed to offer clients early case assessments so that they can obtain an indication at the outset of an instruction whether the underlying documents are likely to support or undermine their case and adjust their strategy accordingly.
This again echoes a point made by Eversheds Sutherland – once you own the technology and you have the staff and the training, then you can apply them to any matter, not just to those which seem big enough to warrant the time and money, and you can do so at once. You can give the client a very quick appraisal of prospects, again by adding human judgment to the facts as they appear from the documents. The visual analytics in FTI’s Ringtail are designed for this purpose.
If that is an advantage, you can replicate it, or come close to replicating it, by relationships with outside providers. By “relationships”, I mean that you know and trust them, and perhaps have master agreements with them, and need waste little time in getting access to their first-pass output without the investment in your own systems and people.
Ask yourself this: if a big new case – litigation, a regulatory investigation or an internal investigation – comes in tomorrow requiring near-immediate reaction, what will you do and who will you call – what is the firm’s playbook? If you don’t know, then it is time to take some of those calls from providers.
And when you have sorted that out, make sure the marketing department knows about it. Firms like BLP and Eversheds Sutherland are using their investments not just to do the work but to sell their ability to do the work.