Artificial Intelligence and the Elimination of Jury Trials in America

Since the release of ChatGPT, fear that AI will replace lawyers has multiplied, leading to lawsuits to stop AI models from the unauthorized practice of law.?Instead, jurists, judges, and those who value a democracy founded upon equal access to justice should embrace AI and focus on creating a truly open AI model that would level the playing field between litigants and create more just outcomes overall. ?An open AI model would also eliminate many of the trials that cost the United States economy over $200B per year.

Renowned trial attorney Larry Popofsky once told a group of newly minted lawyers, “If a lawsuit goes to trial, someone is making a serious mistake.” At the time, I was shocked. How could going to trial be a serious mistake? Trials are the hallmark of the American judicial system.?Academia, the judiciary, and the American Bar Association fretted for decades over “the vanishing trial.” But after practicing law in state and federal courts for more than two decades, I understand that Mr. Popofsky did not intend to diminish the importance of jury trials; he was merely reflecting the reality that, in most civil cases, the unequal abilities of each side’s legal team to analyze and present their case often leads to significant miscalculations of the probability of success.?

More recently, Tyler Shultz, the Theranos whistleblower whose parents mortgaged their house to defend him against defamation charges filed by the Boies law firm, aptly described the problem when asked how he managed to spend more than $400,000 of legal fees in a matter of months: The legal system, he said, is like a game of pick-up basketball. When the other side brings out LeBron James, there’s little choice but to lawyer-up with someone of equal caliber and cost.?In other words, the outcomes of many lawsuits today are determined more by the relative skills of the lawyers than the merits of the case.

Leveling the Playing Field with Artificial Intelligence

To level the playing field, I am not proposing that AI be used to make binding judicial decisions or findings of fact (although some claims could easily be decided by robo-judges, as demonstrated by China’s successful robo-courts for online disputes.)?Rather, I am proposing that a properly built, curated, and transparent AI model that predicts litigation outcomes, is maintained through a public-private partnership, and open to everyone at no or low cost would diminish inequalities arising from the economic disparity of the parties and the unequal skills and abilities of lawyers themselves.?

Just as Westlaw and LexisNexis changed the world of legal research, giving anyone (who can afford their services) the ability to conduct legal research 100 times faster than manually pulling books off a library shelf, AI can change the world of dispute resolution by enabling anyone to not only conduct legal research, but also analyze facts, write and respond to discovery requests, and conduct depositions.?

An AI model with predictive abilities and a sufficiently large and properly curated database of legal decisions, legal briefs, witness videos and transcripts, pre-trial and trial rulings, juror demographics, verdicts and settlements would be able to provide an assessment that identifies the best and worst legal theories, best and worst facts, least and most credible witnesses, the behavior of judges or juries, and the overall probability of a successful outcome.?Armed with this information, every litigant could have an AI lawyer on par with the top trial lawyers in the country to provide an objective opinion of their likelihood of winning the case.

The more people use an AI model, the better its predictive analysis will become. The better the predictive analysis, the less likely a party will make a serious mistake by trying a case with a low probability of success.

ChatGPT on Robot Lawyers

I recently asked ChatGPT whether AI could interpret legal precedent and apply legal reasoning as good or better than a human lawyer.?ChatGPT said not only can it “be trained to interpret legal precedent and apply legal reasoning,” it also???“may be able to find and interpret legal precedent more efficiently and accurately than some percentage of lawyers in the United States, especially when it comes to processing and analyzing large volumes of legal documents and data.”

As a natural language processing AI, ChatGPT may not be able to assess the credibility of witnesses and gauge human emotions.?But other AI technologies already exist, and continue to be developed, that can detect the credibility of witnesses and gauge human emotion. Converus’ EyeDetect , for example, uses eye-tracking and pupil dilation to detect lying for border control screening. Studies comparing the relative ability of humans and AI models to detect lying in courtroom videos go back over 5 years. ?

In our conversation, ChatGPT initially attempted to claim that it was not as qualified as humans to provide legal analysis because of humans’ ability to understand human behavior. However, ChatGPT conceded that not all humans have the same ability to understand human behavior. Moreover, when asked whether the outcome of a lawsuit should be determined by the relative ability of a human lawyer to understand emotions and human behavior, ChatGPT replied:

“Even the most skilled human lawyer may not have a perfect understanding of human behavior and emotions, and may be subject to biases and other limitations that can impact their ability to make sound legal decisions.

Ultimately, the goal of the legal system should be to arrive at a just outcome based on careful and thorough consideration of all relevant legal and factual factors, rather than on the relative skill of one lawyer or group of lawyers to interpret human behavior and emotions.”

Robot Lawyers Already Exist

In fact, the use of AI technology for predictive legal analysis already exists. LexMachina, for example, claims its proprietary technology will help law firms and corporations craft successful strategies and predict the behavior of courts, judges, lawyers and parties. Blue J Legal claims its platform uses AI to help legal and tax experts conduct better research and analysis. Legalist, an investment firm that funds litigation, uses its proprietary technology to predict the outcome of cases, and claims 80% of its cases end up winning.?PWC recently announced its agreement providing its 4,000 lawyers access to Harvey, an AI legal services tool based on ChatGPT.

Outside of law firms, however, AI-supported legal services continue to face opposition. DoNotPay, which advertises itself as a robot-lawyer that helps fight traffic tickets and other consumer advocacy, has been blocked from providing live support during courtroom hearings and sued for the unauthorized practice of law. Obejctively, however, there is not much difference between the legal services Harvey and DoNotPay provide. A November 2022 Tech Crunch article interviewing the Harvey founders provides an example of how Harvey could be used:

“Harvey can answer questions asked in natural language like, ‘Tell me if this clause in a lease is in violation of California law, and if so, rewrite it so it is no longer in violation.’”

A good litmus test for determining whether an AI model is engaging in the practice of law is to analyze whether the same action, if performed by a human non-lawyer for a fee, would be considered the practice of law. Applying this test, there’s no question that asking Harvey to rewrite a lease so it does not violate California law constitutes the practice of law.?

The only real difference between Harvey and DoNotPay is the user.?One markets to law firms and the other to non-lawyers. Some may claim that allowing AI tools to be used only by lawyers protects the public from bad lawyering.?A cynic might view this as proof that concerns about AI “practicing law” are more about protecting the legal monopoly lawyers enjoy than protecting the public from harm. After all, robot-lawyers could be licensed, covered by malpractice insurance, and sued for negligence, just like human lawyers, as a condition for practicing law. ?

What each of these AI tools have in common is that they are being developed in silos, by private, for-profit entities without the level of transparency and government oversight expected of our judicial system. A better alternative is to create an open source AI platform, with input from judges, lawyers, AI technologists, public advocates, and data and social scientists that can be used throughout the country by all litigants on a low cost or no cost basis.

?Thirty-five years ago, the Judicial Conference of the United States embarked on a forward-thinking project to create an online system providing public access to court records. At the time, not everyone had computers and no one could have foreseen how the internet was going to change our world. Today, the Public Access to Electronic Records or PACER system, provides low cost public access to over 1 billion federal court records. It’s time to take a similar long-term view of artificial intelligence, digital content, and the courts.??