Most times, keywords are still the go-to method for professionals in the legal field when they need to locate electronically stored information during an investigation. Even though it is still predominantly used, it is not the most efficient.
A study done showed that keyword searches can miss up to 80% of relevant information in a case while reports from the 2009 TREC Legal Track study indicated that keyword searches missed 96% of relevant information and documents.
Today, there are other tools available to assist legal professionals who deal with large volumes of electronically stored information, one of them being predictive coding. This article looks at predictive coding and the impact it has had on the e-discovery process;
Introduction to E-Discovery
E-discovery, also known electronic discovery refers to the process where electronic data is searched, sought, located, and secured to use it as evidence in a criminal or civil legal case. This also includes government sanctioned or court-ordered hacking to obtain critical evidence as well as cyber forensics or computer forensics. The process can be carried out in a network or on a particular system.
Digital data's nature makes it suitable for investigations. Unlike paper documents; which have to be scrutinized manually, digital data can be easily searched electronically and is hard, sometimes impossible to destroy completely; especially if it gets into a particular network.
The field is still evolving and goes beyond technology; the process brings up multiple constitutional, legal, security, personal privacy issues, among others, many of which are yet to be addressed or resolved.
While predictive coding is the buzz right now when it comes to e-discovery, many are still working on how to implement or adopt it into day-to-day practices widely.
Predictive coding, also known as computer-assisted review or technology-assisted review, is the technology used to find electronically stored information documents during the review stage of a legal case. The process makes use of artificial intelligence to come up with software that keeps on learning and improving decision-making while expediting the review process, ultimately saving time and money.
The whole process begins with training the software by using a set of seed data, which is made up of documents retrieved from the group of documents that require a review. The reviewers then distinguish each document as either responsive or unresponsive. A responsive document implies it is relevant to the case, and unresponsive means it irrelevant. This information is input into the coding software. AI makes it possible for the software to learn, making faster and better decisions as training and time go on.
The goal of predictive coding is to limit the number of unresponsive or irrelevant documents that will require further manual revision. In addition to the AI programming, the software makes use of a mathematical model during the process of analysis.
Once the software has reviewed the set of data, the team reviews the documents tagged by the software to determine whether a level of confidence in the process has been met. If it hasn't, then the whole training process is repeated until the software learns what is required of it.
Impact of Predictive Coding on E-Discovery
Predictive coding has several benefits over human review as well as keyword searches; benefits that have had a significant impact on the electronic discovery process;
Human review of every document in a set of data can be expensive, especially when dealing with a case with large volumes of electronically stored information. Predictive coding as a process is not. There are a few costs incurred during the initial stages of setting up the software and training it, but once it attains a confidence threshold, the costs incurred during e-discovery significantly decrease.
2. Reduced Errors
Human reviewers and keyword searches are prone to errors. Manual reviews of large volumes of documents are often tedious and rely on an individual's keenness, which isn't always constant. Predictive coding assures lawyers reviewing documents of a particular case that they have covered all relevant information about the case.
3. Improved e-discovery efficiency
A well designed predictive coding protocol or algorithm can uncover at least 70% of relevant documents. Furthermore, all this is done at a fraction of the cost and with much fewer errors than experienced with the keyword searches or human review methods.
4. Saves time
Instead of spending months going over all the documents to discover the relevant facts of the case, predictive coding has made it possible for legal teams to discover, learn and familiarize themselves with the facts of the case in only a matter of days or weeks. Being able to learn the facts of the case at an earlier stage ultimately leads to the best possible preparation needed for the case. The reduced time for preparation further reduces the costs incurred during litigation, enabling an earlier case assessment.
Because of the limitations of keyword searches and the error-prone and mundane nature of the human review method, when dealing with an electronically stored information intensive case, consider implementing the use of predictive coding software.
Predictive Coding Practices
If you are new to the procedure, these are some of the practices you can consider implementing to enable you to get the most out of your software;
• Become familiar with the technology the software uses. You don't need to be an expert in statistical coding, but a solid understanding of the system's functionality is crucial.
• Properly train the software. One of the basics of IT is 'garbage-in, garbage-out.' Therefore, how you train it has a significant bearing on the results you will receive. Carefully select your seed set of data for training as well as your team of reviewers.
• Establish an appropriate confidence threshold. Knowing how confident the system is will help your legal team manage further manual reviews.
The idea of predictive coding is to lighten the manual review process by making it possible for legal teams to review smaller sets of data that are relevant to the case, ultimately saving time, money and further enabling them to dedicate their efforts to other matters that require their expertise.
At Aureus Tech Systems, we focus on alleviating the unnecessary strain caused by inefficient processes. Talk to us today.