In 2015, the Federal Rules of Civil Procedures further encouraged litigants to perform electronic discovery in the most efficient manner and to utilize technological solutions whenever possible to do so in order to increase transparency and reduce the time and money spent on eDiscovery.
How It Works
Predictive coding is a software-based, machine-learning process. The software is “seeded” by applying it to a subset of documents, the seed set, to test document responsiveness to selected search terms. Search terms are refined as the software generates on-the-fly algorithms for subsequent searches. These steps are repeated until a suitable responsiveness threshold is achieved. The final algorithm is then applied to much larger document sets.
Predictive coding may also be referred to as predictive intelligence, computer-assisted review or technology-assisted review (TAR).
Predictive Coding’s Acceptance
A ruling by Federal Magistrate Judge Andrew Peck in 2012 was the first official endorsement of predictive coding for the discovery phase of a case, after which many jurisdictions accepted it as a means to speed up discovery. However, recent surveys of judges across the country suggest that slightly less than half of legal teams actually employ it regularly.
Obstacles to Full Acceptance
The alternative to predictive coding is, of course, manual document review despite its laborious nature. The preference for manual review can be attributed to three main reasons:
- Front-end cost – Predictive coding software is expensive. The cost is high enough that it typically counts as a capital expenditure, which triggers an organization’s tedious justification and procurement processes. Sometimes, in the name of expediency, it is simply easier for legal teams to stick with what they have.
- Technology Resistance – The legal industry is one that has been less enamored of technology than others. Additionally, if the technology is perceived as disruptive to well-trod procedures, this provides skeptics another reason to defer its use.
- Humans Do It Better – There is an unfounded belief that no machine can perform discovery as thoroughly or accurately as a human. However, this belief is not backed by objective measures of the task. In fact, one study showed that despite experienced lawyers’ belief that they had manually discovered at least 75% of relevant document in a collection, they had in fact only found 20%.
Predictive Coding Outside the Legal Arena
The increasing adoption of the concept of Forensic Readiness by businesses provides a good example of where predictive coding can be utilized beyond litigation discovery. It can streamline the process of pre-discovery of enterprise data from any source, in-house or in the cloud, including email, databases, social media, system logs, archives and so on.
In the hands of a company that offers a highly qualified, experienced eDiscovery service, predictive coding would drastically cut the costs of locating data from these sources that is relevant to, say, regulatory compliance, customer service activities, intellectual property protection or data retention policies.
It seems that regardless of the obstacles to predictive coding acceptance in the legal area, it is here to stay. Further judicial rulings accepting predictive coding for eDiscovery, improvements in the software and the larger trends of improving efficiency through digital technology are sure to make its use ever more widespread.