P2P

38 P E E R T O P E E R : I L T A ' S Q U A R T E R L Y M A G A Z I N E | F A L L 2 0 2 3 concepts used in TAR methodologies today, such as supervised machine learning, AI and sophisticated algorithms, have expanded the utility of TAR to other workflows that are relatively common in the eDiscovery review life cycle. In the past, completing these nonstandard workflows required more manual labor and guesswork leading to lack of access to accurate information and increased costs due to the complexity of such workflows. However, by leveraging TAR in more creative ways, the following tasks can be completed more accurately and efficiently leading to better outcomes for our clients. Snapshot Analysis of Your Data and Its Richness Once your data set is established, one standard approach to determining the richness of your data is to review a statistically valid random sample of your set to establish a baseline within a given margin of error. By taking this statistically valid random sample or "yield sample," you can learn some basic information about what you are going to encounter during your review – the percentage of documents that will be responsive and the number of overall responsive documents within a margin of error. However, if you know you are going to use a TAR workflow, you can quickly feed the results of the yield sample into the TAR model and focus your initial review on just the documents in the highest-scored bands. (How many bands you review will be dependent on the volume of documents within those bands.) This can give you a relatively quick high-level understanding of the documents you will encounter in your review, and as a lawyer, you can get an early understanding of the issues in your case and the weaknesses and strengths highlighted by your data. This also allows you to better craft your review strategy moving forward as this exercise might bring to light issues you did not anticipate or tricky concepts that you will need to explain and unpack to your broader review team. It also will help you structure your case at the outset, identify parties to depose, defenses to assert and other key metrics important to your case. Finally, leveraging TAR in this manner may allow you to advise your client earlier on whether it makes sense to settle a matter quickly or proceed with litigation. Key Document Identification Your team may love you or hate you for this one! Instead of your TAR model being trained to focus on responsiveness for purposes of production, you can create and train your model to focus on identifying key documents for your case. You can help train such a model by feeding key documents you have already uncovered in your review into the model. And if you are feeling creative, you can actually create important documents based on the issues in your case from scratch and feed those into the model as well to stabilize it more quickly. This can assist you in uncovering key documents, either helpful or harmful to your case, early in the discovery process. Whether you uncover helpful or harmful documents or both to your case, your team will appreciate the opportunity to prepare accordingly to deal with these documents. No one likes to be surprised by a smoking gun document while sitting in a deposition! A couple of pro tips here: Always use a CAL model for this exercise as your key document focus may change or new key issues may arise while the work is underway. CAL allows you to continue to train the model over time so your key document model can evolve as your case evolves. It is harder for TAR models to stabilize when your focus is on uncovering very small populations within your data set (needles in a haystack). If you are targeting a very small population with your key document model, the model may never fully stabilize. That's okay! But try to spread your search for key documents a bit more broadly and understand that the model may never completely settle. At some point, you will likely have to make a cost-benefit analysis on when to stop your search for key documents. Don't let perfect be the enemy of good! Issue Identification Maybe instead of being focused broadly on key documents, you are tasked with focusing on one important issue in the case. Using the same approach as detailed above, you can leverage TAR models to help you get a clear picture of what documents in your data set are related to the issue central to you. Again, a CAL model is the way to go here because your understanding of the issue will likely change as you review more documents and get a better grasp of the case over time. With CAL, you can continue to train the model to focus on F E A T U R E S 1 2

• Cover