Judge Brantley Starr of the Northern District Court of Texas recently issued an order that requires lawyers who file documents on his docket to certify either that the filings are free of content produced by large language model artificial intelligence tools, like ChatGPT. Another acceptable option he offered was that the documents being submitted receive a review by a human for accuracy.
This followed an unfortunate New York incident in which a lawyer filed a document that had been produced by an AI tool that invented citations. In a preventative move, Judge Starr specifically cited a tendency for AI tools in legal practice to “hallucinate” by generating text that appears plausible but is factually, semantically, or syntactically incorrect. The New York judge presiding over the original unfortunate case has now urged the lawyers to consider “seeking forgiveness.” But AI has no shame, which may be a fatal flaw.
Judge Starr’s order appears to be the first of its kind in Texas. But AI tools have been around for decades, and the pros and cons of using AI tools in legal practice, especially criminal law, have already been explored by thoughtful legal scholars. Issues about the use of AI in law for business, employment, personal injury, and commercial litigation are still developing.
Kilgore & Kilgore Lawyers Understand Texas Litigation and How to Use AI Tools in the Legal Practice
The human touch defines how we work. Our commercial, business, personal injury, and employment litigation lawyers use sophisticated technology tempered by human intelligence. Yes, we use leading-edge AI tools. Yes, our legal staff of real people is ready to talk with you. At Kilgore, you can rely on quality legal representation in Texas. We are humans serving humans. Call us at (214) 969-9099. You can also reach out to us with this link Contact Kilgore Law to use an online form to submit through our website to get the conservation started.
AI Tools Defined
Artificial intelligence is an umbrella term that refers to computers doing intelligent things such as performing cognitive tasks — learning, reasoning, analysis — that were once thought to be the sole province of humans. It is not any one technology or function. AI tools in common use are just called software. Each of the three software elements described below has strengths and frailties. Extended or virtual reality tools, perhaps the next AI tools in legal practice, pose additional challenges. These AI tools are characterized by:
- Algorithms, which can be understood as a set of steps to complete a task;
- Machine learning or, at its simplest, an ability to imitate human behavior; and
- Natural language processing.
Strengths of AI in the Legal Practice
In the context of a law office, which is where much of civil practice is conducted, AI tools are already in common use. For example:
- Lawyers use AI tools to conduct legal research, draft and review contracts, and find relevant documents in the discovery process. The principal strength of AI is its ability to sift quickly through huge amounts of data. Through patterns, AI tools might be useful in spotting large scale financial fraud — for example, a pattern of upcoding in a Medicare reimbursement scam.
- AI tools can help predict judgment outcomes in preparing for settlement negotiations. At that point, litigants are often calculating risks and rewards. AI tools help by performing predictive modeling.
In the context of face-to-face situations, AI tools can help. For example:
- Artificial intelligence can be invaluable in overcoming language barriers.
- Virtual reality tools can increase attorney productivity and avoid costly mistakes, which may make the legal process fairer and more accessible. This, in turn, may level the playing field for those without deep pockets. Countries like Estonia and China have experimented with online courts for minor civil matters, like parking tickets and small-dollar contract issues, which are decided by human decision-makers informed by AI tools.
- It is possible to imagine, for good or bad, that JudgeBots might eventually step into the space now occupied by arbitrators.
AI Limitations in the Legal Practice
An algorithm, itself, is amoral. This is not a system designed to recognize ethics. The three basic tools of artificial intelligence are also the sources of its limitations:
- Algorithms are built using historical data. These may incorporate historical biases or the biases of programmers. We cannot expect great leaps of imagination or an articulation of social imperatives from artificial intelligence.
- The tasks that algorithms are designed to perform may be opaque to the end user.
- Algorithms are value neutral. The designated task may lend itself to destructive goals or desirable ones. For example, if an algorithm’s task is to produce the greatest number of responses or Likes (probably a neutral good), and the user asks, How can I build a bomb? the algorithm could become a networking tool for bomb builders, clearly a bad thing.
- AI tools can process superhuman quantities of information at superhuman speeds. It can then adapt to the new data set. The ability of AI tools to learn has triggered warnings about the imminent extinction of humanity. These predictions may be premature, but the value-free, adaptive capacities of artificial intelligence certainly bear watching.
AI Tools in a Law Practice
In addition, as data scientists have pointed out, machine learning can fall prey to common statistical fallacies that produce absurd results. Statistical fallacies are common tricks data can play on us, which lead to mistakes in data interpretation and analysis. AI has no shame, it seems. This is the old garbage in, garbage out issue. Such fallacies include:
- Cherry picking a data set to produce a desired result;
- Data dredging, which confuses causation with correlation;
- Survivorship bias, which draws conclusions from an incomplete set of data
because that data has survived some particular selection criteria;
- The Cobra effect, which occurs when an incentive produces the opposite result intended; and
- Sampling bias, which occurs when the data set is not representative of the population at issue.
The advantage of natural language processing is that it allows individuals searching for a particular term or concept to go beyond traditional Boolean(1) searches. But as many have noted, legal language is different from natural language. For example, conversion under the law is not likely a happy experience. To support the practice of law, attorneys may need a set of AI tools developed specifically by lawyers.
Artificial Intelligence in the Courtroom
Artificial intelligence in the courtroom has significant shortcomings. Existing virtual or augmented reality tools do not appear to be skilled at processing contextual clues such as facial expressions or body movements. For real people, subtle clues around the mouths and eyes weigh heavily in judging whether someone is trustworthy, threatening, or just plain shifty-looking. It is difficult to imagine a robot lawyer doing a competent job of jury selection without human help.
And that, perhaps, is the point of Judge Brantley Starr’s order. To use an analogy, cars with self-driving features still require the presence of a human driver who can, if necessary, override the car’s autonomous features. Judge Starr’s order requires that a human lawyer function in the same way regarding legal work. To push the analogy further, fully autonomous cars may ultimately require a major re-design of our nation’s highways and cities. No one seems eager yet to take this comparison to our legal system.
We Look Forward to Working With You Using the Human Touch
Our Texas litigators will help you assess your legal situation and advise you on how best to tackle your legal issue. We are a technically sophisticated firm with a human touch and human intelligence. To learn more about us, click this link Personal Injury. We offer a free evaluation of the facts of your case. To get this process started, call us at (214) 949-9099 and speak with a real person. You may wish to contact us through our website using this link Contact Us to get the conversation started.
(1) Boolean searching is used to help find search results faster and with more precision. Boolean searching uses operators: words like AND, OR, …