Trellis Research CEO Nicole Clark discusses trial preparation challenges, and the unique approach Trellis uses leveraging AI and Machine Learning to allow trial lawyers to gain a competitive advantage and improve outcomes.
Attorneys across the state of New York are already finding innovative ways to utilize AI-powered legal analytics to develop data-driven forecasts about virtually every aspect of the legal industry.
Across the state of New York, new legal technologies continue to transform the legal sector, provoking waves of anxiety about the future of the legal profession. These waves were felt at the New York State Bar Association Annual Meeting in New York City earlier this year. The forum hosted a session titled “Emerging Technologies in Litigation,” the purpose of which was to discuss the changing role of technology in the courtroom—particularly the use of artificial intelligence (AI).
Participants learned that AI is already a fairly well-established phenomenon in the legal sector, even if attorneys have never interfaced with a legal tech product directly. Consider the Correctional Offender Management Profiling for Alternative Sanctions (COMPAS) program. According to one participant, the Hon. Melissa Crane of the New York City Civil Court, COMPAS uses AI in a risk assessment program, a tool that makes decisions about the kind of supervision an offender will receive. These decisions are based on an AI-algorithm that weighs the personal characteristics of an offender as well as their responses to specific questions. There is just one small problem. Algorithms don’t always work. Right?
AI-powered legal analytics is new, uncharted territory. Today, AI-powered legal analytics is fueled by large volumes of data, information that has been nearly impossible to acquire and assemble in one place.
It’s no secret that state trial court data have been trapped inside their own impenetrable black boxes. Year after year, new legal tech innovators emerge, each with their own unique techniques for accessing, collating, and analyzing trial court records. The majority of these techniques serve federal court cases. Recently, however, the black box of state trial court data has begun to unravel, morphing into something more akin to a crystal ball.
Litigating with a Black Box
Uncertainty is part and parcel of the litigation landscape. Over the years, attorneys have developed their own rudimentary tools for grappling with the unknown. At times, this has taken the form of an office-wide email to gather intel on newly assigned judges. As many will attest, such requests are often met with anecdata, anecdotal responses filled with non-actionable sentiments about a judge being “a real stickler for the rules.”
While this information does inform, it does not provide a lot of decision-making guidance, especially when it comes to formulating litigation strategies.
Luckily, attorneys have been able to rely on other techniques. While working at Wilmer, Cutler & Pickering in the 1990s, Ron Friedmann applied decision analysis protocols to his litigation practice. Friedmann managed the complexity of his cases by breaking them down into smaller problems, what he calls ‘key events’. He would manage uncertainty by explicitly estimating the probability of each key event, drawing a decision tree diagram to visualize every single possible final outcome of a case. This exercise allowed Friedmann to bring quantitative rigor to his litigation practice, providing clients with numbers, statistics, and percentages on how a case would conclude.
Still, an important question remains. Where does the data for a decision analysis protocol derive from?
With state trial court data hidden inside black boxes, attorneys have had no other option than to rely on experiential knowledge and anecdata to make strategic decisions about the particulars of a case.
Things are different now.
AI-powered legal analytics is fueled by large volumes of data, information that has been nearly impossible to acquire and assemble in one place. This fuel has required legal tech innovators to reimagine how state trial court data is gathered, stored, and synthesized. They have responded with their own multi-layered machine learning classification algorithms, technologies that can seamlessly integrate irreconcilable datasets to new ends.
Litigating with a Crystal Ball
Many litigators are familiar with the ways in which AI and machine learning technologies have transformed the routine and mundane tasks associated with legal research. When reduced to its most basic components, the task of an attorney is to analyze texts and make predictions about litigation outcomes. Much of this work is accomplished through legal research, the slow sifting through case law in search of passages that might be applicable to a current case.
Consider MacKenzie Dunham, an attorney who operates Access Justice Houston in Texas. Dunham describes a challenging sexual assault case, one that required him to search through multiple sources of case law to locate an authoritative definition of reasonable grounds to
believe. He started his research the way he always starts, combing through secondary sources and utilizing conventional search methods. When these strategies proved ineffective, he tried Casetext’s Case Analysis Research Assistant (CARA), an AI-powered legal research tool that uses latent semantic analysis to identify patterns in legal documents. Seconds later, CARA provided him with “a U.S. Supreme Court case analyzing what a ‘reasonable ground’ meant and comparing it to probable cause.” It was that easy.
Perhaps fewer litigators are familiar with the potential for legal analytics to make predictions about the future. Here are two examples.
Toby Unwin, co-founder of Premonition, a legal tech company based in New York, describes family law as “the Wild West of the legal profession.”
He explains how “[family law] has a remarkable lack of precedents,” a characteristic that makes determining win rates highly problematic as “there really aren’t many clear ‘winners’ in domestic proceedings.” With this, we learn that there is much more to predict than simply the outcome of a case. Premonition’s clients do not always want to know who will win a case. Often, they just want to know how quickly a case is likely to close. For one client, in particular, he used its legal analytics software to identify how long divorce cases in Orange County typically last.
With this information, he compiled a list of attorneys that tend to let their cases linger in the courts, a crucial piece of information for anyone researching their opposing counsel.
Trellis recently compiled an archive of state trial court rulings in California, Florida, New York, and Texas. Its analytics platform brings together the messy, raw data from counties across each state, restructuring and reclassifying the information in ways that allow litigators to draw strategic insights.
Earlier this year, one firm utilized its platform to develop a custom analytics strategy report on settlement amounts in class action cases with PAGA claims.
The report included detailed information about the settlement amounts judges had approved in the past, comparing these numbers to the number of weeks for each trial. With this data, the law firm could develop a settlement strategy unique to its client, constructing a cost-benefit analysis that would help it identify how long it should litigate the case before settling.
A Less Uncertain Future
AI-powered legal analytics is new, uncharted territory.
Each year, more and more law firms find themselves entering this new terrain, capitalizing on its largely untapped promises. These firms are already making an impact, with at least one report suggesting that attorneys are 50 percent more productive when they use legal technologies than when they work manually. All of this is to suggest that the slow, piecemeal collection and analysis of anecdotal accounts will no longer be enough to stay competitive in a market filled with statistically significant, data-driven insights. Suddenly, the uncertainties of the legal landscape are no longer an unknown to be feared. They are now an element to be managed.