Disclosure of electronically stored information can be a time-consuming and costly exercise unless technology is on your side, writes Charles Orton-Jones
Disclosure is a process which many lawyers find difficult. It’s understandable. The sheer proliferation of digital documents, the rise social media and challenge of getting to grips with cutting-edge technology, such as predictive coding, leave many legal practitioners pining for the days when everything was done on paper.
Just consider the smorgasbord of places which need searching. There’s WhatsApp. The Libor subterfuges took place via online chatrooms. And how about Snapchat, where the messages are erased after a few seconds – does that need to be considered? Add in Facebook messenger, recorded Skype conversations and loud storage, and it’s a tough ask to cover everything.
Get it wrong and there is hell to pay. In West Africa Gas Pipeline v Willbros Global Holdings Inc., the respondent failed to identify a complete set of relevant documents at the outset. An additional 47,197 documents were provided two years after the start of the proceedings, resulting in an unamused Mr Justice Ramsey ordering the respondent to pay half the applicant’s costs incurred reviewing these documents.
So what tips can disclosure experts pass on to practitioners who want to raise their game?
Theo Solley, partner at Davenport Lyons, says the initial conversation is vital. “The West Africa Gas Pipeline problem occurred because the lawyers failed to communicate which documents were going to be relevant. Disclosure begins with picking up the phone and having that conversation,” he says.
As well as discussing what constitutes a relevant document, it will be necessary to confirm which repositories will be searched. Not easy. Craig Earnshaw, managing director of FTI Consulting’s forensic disclosure division, says there are more than many lawyers realise.
“In addition to e-mail and corporate archives, there are things like cloud-based systems such as Salesforce.com, CRM systems, sales tracking programmes, online finance systems and bespoke environments,” he says, advocating bringing in a technology specialist at this early stage. “Taking part in that conversation means your forensic partner will understand what systems are relevant.” Bringing technology partners into interviews with witnesses is increasingly common.
Attaching documents to an e-mail is not OK – that is as secure as a postcard
Then comes the narrowing down of material. “When it comes down to it, you need surprisingly few documents in court,” says Nick Burkhill of Dorsey & Whitney. “I recently did a small exercise where the number of documents was 600,000 at the start. Then, after the application of search terms and de-duplications, we produced 15,000. The other side produced 1,000 and we ended up with a core of 3,000 documents in court. In fact, we used two lever arch files in court.”
Keywords and date restrictions are commonly used. But the real skill is using all techniques from concept clustering, predictive coding, and metadata searches and tabulation. Incredibly, these techniques, which radically cut costs, are not always employed.
Daniel Kavan, who leads the electronic disclosure team at Kroll Ontrack, says: “My biggest challenge is getting clients to pay for the services. We still find clients who don’t see the benefits of using technology in disclosure. They want their lawyers to do everything. I need to explain that by paying a bit more for a third party offering technical expertise they can save significant sums in the long run. You really don’t want lawyers wasting time fiddling with documents.”
Preserving document integrity is vital during this disclosure. In a recent case, an employee was accused of downloading indecent material on to his company mobile phone. The images were visible, but the employee denied responsibility. Grey Heron, a specialist in recovering lost or deleted data, used AccessData’s MPE+ mobile phone extraction tool to identify and extract metadata, thus proving who downloaded what and when, making a prosecution viable.
Disclosure will result in the compilation of hundreds or thousands of sensitive documents, which need handling correctly. Peter Wright, managing director of DigitalLawUK, issues this warning: “Lawyers need to understand the important of secure document transition during disclosure. Attaching documents to an e-mail is not OK – that is as secure as a postcard.
“Law firms can use secure online datarooms, but these are not cheap; other law firms are still exploring the technology and getting up to speed. I’m afraid it may be down to clients to ask their law firm ‘Is our data being looked after in a secure manner?’” Data breaches can incur fines of £500,000 or, under new EU law, a percentage of turnover.
Mastering these skills means lawyers can both ensure a full complement of documents can be collected and costs are kept as low as possible. Will Richmond-Coggan, a partner at Pitmans, says the way forward is for lawyers to make the most of the technology. “Before, some moron would write ‘Killer document!’ with indelible ink on a piece of paper. Try explaining that in court. Now you have an evidence trail, metadata, the ability to tag a document with your own comments and questions. It really is far better,” he says.
In terms of cost alone, lawyers need to embrace technology. “We did a case recently with 60,000 documents,” Mr Richmond-Coggan recalls. “Using keywords and filters, we got that down to 10,000; then reviewed manually, costing $25,000. For $10,000 in technology costs we saved maybe 75 per cent in time costs. That’s a big overall saving.”
And the laggards who struggle? “There will always be people who find technology an anathema,” he says. “But the tide is very much against them.”
COMPUTERS DO THE CHECKING
Keyword searches are at the heart of current disclosure processes. But keywords are blunt. Euphemisms and unexpected synonyms are not caught. Nor are relevant conservations for which no keyword is identified. So predictive coding is now routine.
Predictive coding looks for patterns in relevant documents. This might include a common sender or recipient, a trend in the metadata or cluster of words. Using machine-learning from a vetted sample, large volumes of material can be rapidly scanned.
Daniel Kavan of Kroll Ontrack says it can be used to double check even expert opinions: “It is common for a senior reviewer to look at 10 per cent of selected documents. Using predictive coding, you can identify documents which reviewers say aren’t relevant, but computers say are 80 per cent likely to be relevant. It is an extra level of checking,” he says.
The next stage will be natural language processing. This will allow lawyers to pose ordinary questions, similar to the way Apple’s Siri replies to users’ queries. For example, “Find documents where Mr Smith expressed displeasure at the way he was treated”. Natural language processing is already used in social media, “reading” hundreds of thousands of tweets and Facebook posts to determine whether brands and events are perceived as positive or negative. Brand managers use dashboards to monitor their brands in real-time.
But two problems prevent this from being adopted for disclosure. The first is accuracy. George K. Thiruvathukal, professor of computer science at Loyola University Chicago, is a leading researcher in the field and is broadly very confident of its capabilities. But he advises: “We get into trouble with context. We struggle with inferencing. We can deal with a sentence, but a paragraph is much harder.” So while positive or negative sentiment analysis on Twitter is achievable within 70 to 80 per cent accuracy, wider application is much harder.
The second is the legal framework. Craig Earnshaw of FTI Consulting says: “Processes must be repeatable and defensible. With keywords both parties can agree on the words, and the results are repeatable and defensible. With natural language processing, the situation is less clear. The party on the receiving end needs to understand how the results have been achieved, otherwise they may later challenge the disclosure in court.”
Until natural language processing is proven to be sufficiently reliable and can be explained in a way which ensures the confidence of legal practitioners, it will remain only an “up-and-coming” technology in disclosure.