Marketing

TAR + Advanced AI: The Future is Now

Issue link: https://content.lighthouseglobal.com/i/1412483

Contents of this Issue

Navigation

Page 8 of 12

WHITE PAPER: TAR + ADVANCED AI – THE FUTURE IS NOW | 9 as privileged in past matters, but can also analyze those past privilege calls from a variety of angles to make much more accurate predictions on privilege on "new" documents. Applying advanced analytics at a portfolio level can also make prior coding available for reviewers for additional context and to ensure consistency. Coding decisions could also be ported from a past matter to a new one at hand via hash values, depending on the capability of the AI tool itself. BIG DATA When datasets are smaller and the data sources within them are homogenous (essentially if a dataset looks like typical datasets from 10 years ago), traditional TAR can still be a particularly useful tool to classify data and reduce eyes-on review. However, when datasets fall within the realm of "big data," an advanced-AI TAR tool is critical to getting accurate classification. Big data is a dataset where the high volume, velocity, veracity, and/or variety makes it difficult to process using traditional tools. 35 As previously noted, the older supervised machine learning utilized by traditional TAR methods evolved before the big data datasets we are seeing in ediscovery today, meaning that technology cannot scale and may experience latency or just general ineffectiveness when faced with this type of scenario. Advanced AI systems like NLP and deep learning not only scale to meet the demands of big data, they also thrive in a big data environment. The more knowledge and data these systems can learn from, the better they are at classifying new data. Connecting these big datasets for analysis also means gaining a view of an entire portfolio of matters, gaining ediscovery and data insights, and enabling benchmarking. HSR SECOND REQUESTS As noted above, TAR (specifically TAR 1.0) has always been especially helpful in HSR Second Requests. And like traditional ediscovery datasets, HSR Second Request datasets are becoming more voluminous and complicated by modern data. Further, HSR Second Requests themselves are also becoming more frequent: February of 2021 marked a 10-year high in HSR activity. 36 The scrutiny on mergers and acquisitions will only continue to increase in the coming years, as President Biden has made antitrust a focal point of his administration – he recently issued an Executive Order to establish a "whole of government" effort to promote competition. 37 This makes modern HSR Second Requests a perfect use case for a TAR 1.0 process that is enhanced by advanced AI. Negotiations in HSR Second Requests with government regulators around custodians and the relevant time frame for data collection often take up valuable time, making production deadlines much shorter than they would be otherwise. With a TAR 1.0 workflow enhanced by advanced AI, case teams can start analysis and review as soon as the technology and TAR 1.0 workflow is approved by the DOJ or FTC. This is because advanced AI technology can keep the model stable, even as data is in flux – meaning the original training model and classification index will not have to be rebuilt when documents are removed or added from the dataset due to ongoing negotiations. This differs from the legacy technology used in TAR 1.0, where case teams would likely have to start over from scratch when documents are removed or added (wasting valuable time and increasing costs for the underlying organization). In other words, a TAR 1.0 workflow enhanced by advanced AI in a HSR Second Request gives case teams the ability to stay nimble and work much more efficiently. Additionally, the higher precision gained by advanced AI allows review teams in HSR Second Requests to confidently eliminate larger swaths of nonresponsive documents from production. In other words, a TAR 1.0 workflow enhanced by advanced AI can result in a much smaller responsive set at the end of the process (even if the number of training and control set documents remains roughly similar to a traditional TAR 1.0 workflow without advanced AI). In fact, we have seen advanced AI reduce the number of responsive documents in a HSR Second Request by over 40 percent. Another key benefit with advanced AI is that responsive and privilege classification can happen concurrently (rather than waiting to do one at a time, a la traditional TAR workflows). As previously noted, advanced AI systems are built for big data, meaning they can quickly scale to meet volume challenges while requiring less human training. The ability of advanced AI to handle large amounts of complicated data within shorter time frames, with less manual review and more accurate analysis,

Articles in this issue

view archives of Marketing - TAR + Advanced AI: The Future is Now