What happens when a judge relies on a GenAI tool in formulating their decision on a key issue, particularly one that could impact the GenAI providers?
It’s not only law firms and legal departments that are adopting GenAI systems without fully understanding what they can and cannot do — court systems may also be tempted to adopt these tools to short circuit workloads in the face of limited resources. And that poses some risks and concerns to the rule of law, a notion that hinges on accuracy, fairness, and public perception.
The Role of UNESCO
That’s why what organizations like UNESCO (the United Nations Educational, Scientific and Cultural Organization) are doing are important.
UNESCO is an agency that attempts to foster international cooperation in various fields. It often sets standards, develops programs, and creates global networks. One such network is devoted to the development of Guidelines for the use of AI in courts. A recent UNESCO publication discussed the programs being developed to assist courts and tribunals in the use of AI. According to the publication, “The Guidelines provide principles and recommendations to courts and judges on how AI systems may be designed, procured and used to strengthen access to justice, human rights, and protect judicial independence.”
What Are The Risks?
The publication identified three risks which resonate given the current political climate:
- Technology is in the hands of private companies that have little concern for judicial independence. These companies’ primary motive is making a profit, not ensuring fairness and transparency in judicial decisions
- Relatedly, there is the opportunity for subtle influence and manipulation of judicial decisions. As the publication puts it, “Even supportive AI functions, such as document summarization, can shape the facts considered in judgments. When judges use AI outputs, its dataset limitations can inadvertently affect legal reasoning.” What happens if that occurs?
- There is public pressure on courts to adopt AI tools without sufficient safeguards. How can this pressure be tempered in favor of rational decision making when it comes to AI adoption by courts?
The Risks Are Not Theoretical, They’re Real
These dangers and risks are real.
First, tech companies trumpeting AI tools are growing more and more powerful. They create tools that can hallucinate or offer outputs that are inaccurate. Yet the public drums seem to constantly beat the refrain of all the wonders of these tools and how they can help humanity and law without recognizing the inherent risks, particularly to the judiciary. The lack of any watchdogs on judicial use is concerning.
Secondly, given this power and potential lack of understanding by judicial users of the risks and bias of the tools, there is the opportunity for mischief and influence by the vendors to achieve their ends. Let’s say a judge is confronted with an issue that can impact a significant AI player. Could the tools be manipulated to increase the risks of a favorably ruling perhaps subtlety? Who would know?
How would that be dealt with? In today’s political climate where corporations have significant control over all kinds of things from what we are allowed to see to what we can say on their controlled sites, the risk of influence is certainly not insignificant.
Judge Scott Schlegel, an appellate judge from Louisiana and one of the leading voices on the impact of AI on the judiciary, recently raised a similar point. What if there were hidden or white text in legal documents that was designed to lead AI tools to make certain recommendations and reasoning? What if the tools themselves were biased to reach or suggest certain decisions?
Indeed, we are already hearing of judges citing to cases that don’t exist. Who should catch this? Should judges be required to disclose they (or their clerks) have used GenAI tools? Otherwise, who would necessarily know? How would (or could) the legitimacy of an impacted decision be determined?
The Pressures to Use AI in the Courtroom
And then there is the pressure on the judiciary to adopt these tools. The AI hype machine is in overdrive. We constantly hear of all the wondrous things GenAI can achieve. Will legislatures be tempted to mandate adoption of these tools to reduce the costs of a court system? Would overworked and understaffed judges be tempted to use AI tools to move cases, relying on vendor promises of what these tools can do?
Not to mention the public perception of the court system already under siege: what happens to that perception as more and more judges cite to cases that don’t exist and where the case cited does not stand for the proposition asserted? Courts often adopt the reasoning in the briefs of the successful party. What if those briefs are wrong or contain errors? How will those issues be dealt with?
What about bias in the models themselves? If a bias impacts a judicial decision, how will we deal with it? What will be the appropriate appellate standards? Do we need some new ones to deal with AI influence on judicial decision-making?
Why It Matters
That’s why what UNESCO is doing is important. It’s offering guidelines. It’s putting together teams of experts. It’s asking the hard questions. It’s trying to make us all see risks before the GenAI tools impact the rule of law instead of reacting to them.
The rule of law is too important to our society, our way of living, and our economic standards not to ask these hard questions. How can we deal with the concept of fairness and due process when some of the decision-making, even if only small bites, is ceded to GenAI?
How can we ensure transparency in judicial decision-making when it comes to AI? We already have problems knowing how judicial decisions are sometimes reached. With AI, we have yet another transparency barrier as we struggle to know on what a judge relied. Should judges be required to say if they relied on GenAI tools and to what extent in decision-making?
We need to foresee and prepare for what AI could bring. From all indications, UNESCO is doing just that. But we need more. We need federal courts to lead the way in thinking about these issues. We need bar associations to step up and demand training and standards. We need to ensure our judiciary gets the training and the resources to understand and deal with both the benefits and risks of technology, just as lawyers and legal professionals are expected to.
There’s too much at stake not to.
Stephen Embry is a lawyer, speaker, blogger, and writer. He publishes TechLaw Crossroads, a blog devoted to the examination of the tension between technology, the law, and the practice of law.
The post AI In The Courtroom: Will We Trade The Rule Of Law For Efficiency’s Sake? appeared first on Above the Law.