[authors: Edward Burke and Allison Dunham]
The adoption of artificial intelligence (AI) has been widely accepted in the legal community as many have realized the value of technology that can detect relevant content and produce better results. Incorporating AI into document review workflows or using insights to inform case strategy is transformative and drives better outcomes. Whether it’s government requests, civil litigation, or internal investigations, high-profile, fast-moving cases require efficient processes. Strategic deployment of technology will help teams identify key documents and themes early in the case and effectively manage data assessment and review. The continued evolution of AI tools, such as the ability to detect conduct and behavior through sentiment analysis and pattern processing, will further aid investigation compliance, but can also be used proactively. .
Growing acceptance of AI in the legal community
There was a time when skepticism clouded the potential value of predictive models emerging on the eDiscovery horizon. Lawyers were quick to debunk the possibility that machine learning and document classification could potentially offer a defensible solution to rising discovery volumes. In a field typically resistant to change, this was expected. Despite widespread adversity, advocates such as Maura Grossman and Justice Andrew Peck have noted the success of predictive modeling and, as thought leaders in the eDiscovery space, have initiated incremental acceptance. Opponents of analytics now seem to be the minority, and the use of predictive models is widespread in technology-enhanced examination.
Yet even during the early days of the evolution, when the legal community was just beginning to learn about TAR protocols and analytical tools, the use of predictive models was already growing. Advanced forensic features such as pattern processing and sentiment analysis have been introduced as newcomers to the eDiscovery analytics market despite longevity in the field of data science. As usage became more widespread, law firm associates and in-house advisors were often impressed during demonstrations, but apparently hesitant to adopt the feature in a live workflow. Customers started asking how to effectively leverage these tools and asked for help implementing the workflow. Market demand required a dedicated process to use these tools, and legal technologists rose to the challenge.
Exposure to large amounts of data, with radically different content, has enabled knowledgeable users to synthesize the results of sentiment analysis and pattern processing and develop applicable use cases. It became clear that text-based communications in the midst of unstructured data yielded the most valuable results, especially when used in investigations or early case assessments. Since “sentiment analysis tools attempt to automatically label subjective emotions or viewpoints expressed through text”, it follows that email communications would be a target rich data pool for this functionality. To see John Nay, Natural Language Processing and Machine Learning for Legal and Policy Texts (New York University, April 7, 2018). As a template, sentiment analysis can be searched by score and incorporated as a search item. By doing so, when querying the data, we have a smart way to prioritize search results for Custodian Jane Doe, from the year 2015, which hit search terms X, Y, and Z. We We can sort these results by sentiment score. Much to the dismay of discovery advocates, there is still a need to peek through the documents to validate the results (and perhaps do some secondary coding); but in theory, users will discover the most emotionally charged results early in the survey, rather than after a full linear review. The logic behind this method also applies to pattern recognition. For example, if we again consider the guard data example mentioned above, the most abnormal results can be identified and displayed first. This means that if Jane Doe logged off her computer at 5:00 p.m. every day during her five years at A Corp, and then suddenly started sending email communications at midnight for an entire week in January, the tool analysis would probably identify it. model as an anomaly. This can be a valuable function that produces high value results with relatively little effort on the part of those investigating.
When used with care, these features offer powerful techniques for identifying unusual behavior, potentially inappropriate conduct, and emotional exchanges, which are often linked to underlying legal issues in internal investigations and litigation. Specific examples of conduct that are ripe for discovery using these tools include cartel activities, such as price-fixing and other anti-competitive behavior, sexual harassment and inappropriate relationships, and collusion and employee misconduct. However, the potential use cases should not be limited to these situations.
This blog post is an excerpt from the chapter titled “Outsourced Document Review: Data Intelligence, Technologist Lawyers, Advocacy Support” by Edward Burke and Allison Dunham, which appears in the treatise by Thomson Reuters eDiscovery for Corporate Counsel (2022). Reproduced with permission, © 2022, Thomson Reuters. Ed and Allison are eDiscovery experts at Epiq.
A link to the book appears below: