In our two previous blog postings, we discussed how to get legal professionals to trust predictive review and how to limit the amount of attorney time necessary as part of the process. A third argument against predictive review is that the so-called experts cannot seem to decide on which methodology is best.
There are a number of different options, including “seeding,” in which a subject-matter expert finds some key documents at the outset of the project, which are then used to inform the process and “continuous active learning,” where reviewers continuously review documents, and the system keeps “relearning” as the process continues.
Similarly, because predictive review processes are iterative, there is some question about how to determine when the project is done. There are a few different popular methods, including:
- Richness – Continue the project until the percentage of documents marked responsive is equal to the percentage that was predicted by the sample.
- Precision and Recall – These are traditional metrics based on a control. The process tests how precise the results are and how many responsive documents are not picked up. Sometimes these two numbers are combined into what has been called an “F1” number.
- Variation – Test how much the system is learning with each iteration. Once it stops learning, it can be assumed that the coding is exactly the same as it would be if there was an eyes-on review of the entire document set.
For attorneys that have not done their own research and chosen a preferred methodology, this can all be very confusing. And others might be limited by the methodologies their litigation support team or client has chosen—or by what has been agreed to at a meet-and-confer—as not all tools support all methodologies.
But some do. iCONECT-Xera’s new predictive review tool, for example, is designed for both novice and advanced users. It supports all the major predictive coding workflows and introduces a new very simple and effective workflow. This new workflow can be followed by inexperienced users who may otherwise feel paralyzed by all the options.
iCONECT also helps attorneys feel more secure in their decision to stop reviewing by allowing them to better understand the results of the review through the iVIEW Data Visualizer, which enables a constant assessment of ongoing coding and metrics. Attorneys have a choice of pre-created reports that will help them determine the best time to stop reviewing. These reports can be saved as tiles on Xera’s enhanced dashboard so users can keep any eye on relevant content and metrics at all times and at a glance.
Our iCONECT White Paper, 4 Reasons Why They Say You Shouldn’t Use Predictive Review, focuses on this and other ways to improved predictive review adoption rates. If you are interested, please download Part 1 and Part 2.
Written by our 5i Solutions Inc. Partner, iCONECT
iCONECT Development, LLC is an industry leader in developing innovative legal review software and services that empower legal teams to complete complex review projects more cost effectively