- No Bail For Airline Mechanic Charged With Sabotaging PlanePosted 14 hours ago
- Ohio Lawsuit Puts “Sheltered Workshops” Under ScrutinyPosted 14 hours ago
- Automakers Who Cooperate With CA Face Trump Antitrust InvestigationPosted 3 days ago
- Employee Data and the GDPRPosted 4 days ago
- In-House Counsel, PR Consultants And The PrivilegePosted 4 days ago
- New Patent Legislation Has Bipartisan SupportPosted 5 days ago
A Corporate Counsel’s Guide to Predictive Coding
Fernando A. Bohorquez, Jr., BakerHostetler and Alberto Rodriguez, Columbia Law School
Predictive coding is computer or technology-assisted review that extrapolates human review decisions from a subset of materials to classify documents in a larger data set. It typically requires selecting and reviewing a control set of documents to measure the tool’s performance. Then training or seed-set documents are selected, reviewed and coded by attorneys. They in turn train the technology to replicate the reviewer’s judgments. It is imperative that the attorneys most knowledgeable about the case be involved from the beginning of the predictive coding review in order to train the system.
Understanding the technology and vetting the vendor is critical. Considerations include assessing whether the methods and algorithms used by the technology are accepted in the e-discovery community, and ensuring that the technology has an audit function to review accuracy. Ask the vendor to explain the technology in plain English, because at some point you or your outside counsel may need to defend it to your adversary, your boss, or even the judge.
Predictive coding has great potential to help corporate counsel manage the skyrocketing costs of e-discovery, but it doesn’t run on auto-pilot. It is essential to enlist senior counsel to teach the case to the machine. You need to vet reputable vendors with easy to understand and defensible technologies. Use counsel who are comfortable with the workflow and technology, and make sure they understand and can maintain the high level of transparency necessary in disclosing predictive coding protocols.