20 years ago, we weren’t sharing Tiger King memes on TikTok – that’s what group emails were for. In 2002 (when ediscovery processes were just a dream) there was a flurry of memes (even if we didn’t use that term), relating to the paper-shredding activities of auditors Arthur Andersen LLP during the Enron scandal. One joke riffing on a popular series of cell phone commercials stated:
The Arthur Andersen partner was on his cell phone when he said:
“Ship the Enron documents to the feds”
but his secretary heard,
“Rip the Enron documents to shreds”
It turns out that it was all just a case of bad cellular…
What began as an internal investigation into the activities of Enron and Arthur Andersen ended up in various external regulator investigations, a conviction, an acquittal at the Supreme Court, and the eventual collapse of the Big 5 accounting firm.
These days investigations into misconduct generally focus on electronic, rather than paper documentation. In this post I look at how ediscovery processes and tools can be employed in misconduct investigations – and how to avoid the electronic version of document shredding.
eDiscovery is a set of electronic tools, including data analytics, applied to the specific legal procedure of Discovery. But there is no reason why these tools can’t cross-pollinate, and be applied to other domains. After all, data analytics tools are now applied in most major industries: Hence we have Martech (marketing technology), Insurtech (insurance technology), and Fintech (financial technology), as well as Legaltech.
What do I mean by misconduct investigations? I use this term broadly. They might be conducted internally, or in collaboration with an external regulator: They may be in response to a specific complaint, or they may be carried out as a result of routine monitoring or internal auditing activities.
In this post, I look at three common, artificial intelligence-driven, eDiscovery tools: document handling, concept clustering and predictive coding, and how they can be used in misconduct investigations.
I consider each in turn, below.
At the commencement of the investigation, all documents need to be securely retained. IT needs to be involved to ensure that no documentation is deleted, and that any possible alterations or deletions can be monitored. A secure document handling facility should be used for working with the documents.
You are unlikely to know immediately which documents are directly relevant to the investigation. eDiscovery ‘concept clustering’ and ‘threading’ tools, can be used to join up all related emails and messages into cohesive packages. In addition, any duplicates can be removed (deduplication) and any near-duplicates (near deduplication) appropriately coded;
The investigator can review a subset of the total documents, and code them as either relevant or not. An algorithm can then be used to compare other documents against that initial batch, which the investigator also reviews. In an iterative process, the algorithm becomes more robust and comprehensive. This is predictive coding.
The end result is a more comprehensive set of documents revealing any possible misconduct, using a more efficient method, than traditional review.
Tools used in the ediscovery process are often flexible in application – just because they were developed for discovery, doesn’t mean they can’t be usefully applied elsewhere. Secure documentation hubs, concept clustering and predictive coding are all tools that could be usefully adapted for investigations into corporate wrongdoing.