Ep. 45 AI in the Courts: The Promises and Perils of New AI Technology

Artificial Intelligence has the potential to streamline tasks within the courts, increasing efficiency and allowing staff to work on higher level tasks.  But like any technology, AI is not infallible or without risks. In this episode of Tennessee Court Talk, we speak with Michael Navin, Principal Court Management Consultant in the Technology, Architecture Planning and Security Division at the National Center for State Courts about how AI is impacting court systems in this new technological era. 

Transcript

00;00;00;24 - 00;00;22;10
Host
Hello and welcome to Tennessee Court Talk. I'm Samantha Fisher, Communications Director for the Administrative Office of the Courts and your host for this episode. Joining me is Michael Navin with the National Center for State Courts, here to discuss the promises and perils of a rather sensitive subject in the court system right now, artificial intelligence. Welcome, Michael.

00;00;22;17 - 00;00;24;05
Michael Navin
Thank you. Thanks for having me, Samantha.

00;00;24;08 - 00;00;45;01
Host
Absolutely. When we talk about AI, that's such a broad subject. But really, right now, the advancements in AI, we're kind of zeroing in on generative AI, these machine models. And there's a lot of talk about this. What what could it do that could help? What problems could it cause that we need to kind of think about ahead of time.

00;00;45;04 - 00;00;57;17
Host
From your standpoint, studying this and talking to state court systems around the country, you know, what role do you think generative AI could be playing in the evolution of court technology?

00;00;57;19 - 00;01;22;18
Michael Navin
So generative AI, just to be clear, you know, as to as to what, you know, generative AI is it's about taking data, large data and creating new content. It's creating novel content, whether it be images or text or audio or visual. And with that, you know, courts, courts are really on, they're evaluating their processes to see how gen AI can help.

00;01;22;20 - 00;01;45;26
Michael Navin
Help with in their court, really focusing on what their problems are to make it more efficient, not only for internal users, but also for the users of the court, the litigants and things of that nature. So one, you know, in we're we're we're looking at it. Most courts are looking at this as, as a maturity model where they're looking right now and they're finding things that are low risk but have high value.

00;01;45;26 - 00;02;15;26
Michael Navin
And that's really important. You don't want to dive in, start using, generative AI. For my high risk things that still may have some value, but there's so many low value, low risk pieces out there that they can use. So, for example, simple chat bots. This is kind of really the starting point of what a lot of courts are using for, litigants chat bots to go out and be able to, pay your, your traffic ticket much more quickly instead of having to be sent to, you know, go to a website and have to search the website.

00;02;15;26 - 00;02;31;25
Michael Navin
And then going to a link, and then you have to go through all of these other steps. This is where you can use a chat bot to just basically pay your ticket right then and there through the chat bot, find your case. You could, you know, sign up for driver school. You could schedule a hearing based on whatever the schedule was like.

00;02;31;26 - 00;02;50;14
Michael Navin
So that's a real low risk, high value piece that is making it so that people don't have to come to the court to to do something that really only takes a couple of seconds. So it's it's helping out those litigants at the same time, we have chat bots that are more internal. Let's say, people want to know, HR policies.

00;02;50;14 - 00;03;09;04
Michael Navin
So you have a chat bot that's internal, specifically just built around internal documents and internal, procedures. And so if somebody wants to know how much, how much vacation time do I have left? And you can ask the chat bot, and I'll tell you instantly what happens if I lose my key card. It can tell you what you can tell you pretty instantly.

00;03;09;04 - 00;03;30;09
Michael Navin
So we're seeing courts develop that we're seeing courts develop, chat bots for figuring out, putting all of their processes, all of their civil procedures out on to this chat bot to where then it makes it easier for a new employee to just ask question, instead of always having to lean over and ask a question. Ask the question to a more seasoned employee or going to supervisor.

00;03;30;11 - 00;03;48;25
Michael Navin
So it's speeding up the training of employees too. Those are some of the real lower risk steps that I think courts are taking right now. And it's a very, very good, it's a very, very good step to take it, taking it slow. Even though the technology is advancing so fast, it's okay to not try to keep up with the Joneses.

00;03;48;26 - 00;03;54;11
Michael Navin
It's okay to just take take things step by step and find find things that are solving your problems.

00;03;54;14 - 00;04;15;09
Host
You know what I'm what I'm hearing you say. And and you know, we know this from the court system that there's such a heavy lift. The administrative role. Yes. I mean, if AI can help with that, with some of those more sort of tedious, time consuming, laborious tasks and free those people up to do more, critical thinking kind of tasks, then that would be one of the the promises of AI, right?

00;04;15;12 - 00;04;35;05
Michael Navin
Yeah. I mean, you know, they could they could use AI in, in human resources. They could use it to build job descriptions much more quicker. Performance reviews. They could use that to, build up those forms quicker. They could use it in budgeting finance to process, payments much quicker. So you know, those type of things, you know, procurement.

00;04;35;05 - 00;04;58;29
Michael Navin
So there's a lot of things behind the scenes that people don't see that AI can be used, that that can then shift employees to be used for more critical things to, for more, more critical thinking pieces and, just allow, the more repetitive things, the things that just drive people crazy when they work, get get through that, get rid of that so that you can focus on quote unquote.

00;04;58;29 - 00;05;09;08
Michael Navin
I don't want to use the word real work, but just, you know, more, work that that needs a little more, thought that goes into it rather than repetitive. Right.

00;05;09;10 - 00;05;19;16
Host
Michael, what do you think judges and lawyers need to know or understand about these large language models? Use this machine learning AI. Before using it.

00;05;19;19 - 00;05;31;11
Michael Navin
First thing is, they should know is that it's not always right. AI is, generative. AI is not any right, but it boy, it's it always whenever it gives you a false answer, it says it with such confidence when it's wrong.

00;05;31;14 - 00;05;34;01
Host
And like my 11 year old. Yeah.

00;05;34;03 - 00;05;51;08
Michael Navin
Exactly. Right. So they have to know that, you know, at this point, you know, these generative AI applications, they're great for a starting point. They're great for if you just, you know, instead of starting with a blank piece of paper, think of it as like a blank piece of paper. But it's a great but it's a great way to gather information, you know?

00;05;51;08 - 00;06;07;01
Michael Navin
I mean, when I grew up, I had to go to the, the library and look up the card catalog and install the Dewey Decimal System and find the book. Then I had to sit down and read, you know, parts of the book and find the aspects that I needed for whatever. Was I researching. And we've, you know, then change that to the internet and searching that way.

00;06;07;02 - 00;06;32;23
Michael Navin
And now gen AI is the same thing. It's increasing our ability to, gather information much quicker. But we still need to use that thinking in that critical thinking to verify that the information is there. And so for judges and attorneys, they still need to know that verification, especially with some of the cases that with all of their cases, for that matter, it's very important to make sure you're verifying what it is that, that you're reading and what you're submitting to the court.

00;06;32;25 - 00;06;35;13
Michael Navin
That's the very first thing that they need to to focus on.

00;06;35;21 - 00;06;42;26
Host
In articles you've written and in presentations I've seen you, you give, you emphasize the importance of having a human looped in.

00;06;42;28 - 00;07;00;06
Michael Navin
Yes. Correct. Yeah. And that's that's really a that's such an important piece. We're at, you know, a human in the loop. Yes. I can generate something, but you need that. You need to have that human there to still review, to make sure it's capturing everything, that it's not hallucinating, that it's what it is that you're getting is accurate.

00;07;00;06 - 00;07;14;19
Michael Navin
So, yes, the human in the loop, I don't know that. You know, maybe there's a time where we don't need that anymore on some of these things. But for the time being, I think that human in the loop is going to be needed for quite a while until we get really, really comfortable with with what is being produced.

00;07;14;19 - 00;07;23;20
Michael Navin
And until those models are so refined that it's very rare that something, wouldn't be hallucinated or wouldn't give out inaccurate information.

00;07;23;22 - 00;07;31;07
Host
Yeah. I want you to define that for me, because I've just been introduced to that term, hallucinations. Coming from AI, what does that mean?

00;07;31;10 - 00;07;58;23
Michael Navin
So a hallucination is basically where it's it's providing you information that is factually incorrect, even though it may seem plausible. So it's it's it maybe it just it maybe it sees something like, for example, like maybe it sees something in a piece of like that. The big example is back when ChatGPT first came out, an attorney, confidently used ChatGPT to write a motion.

00;07;58;23 - 00;08;21;20
Michael Navin
ChatGPT confidently produced a motion with all these key citations in that attorney confidently provided that that motion to the court. And then the court confidently said, this is all wrong. This this case citations aren't even aren't even real. So now how did it come up with Unreal Case? How did it just make things up? Well, that's kind of where it's, you know, they call this the black box.

00;08;21;20 - 00;08;39;20
Michael Navin
You don't know how generative AI comes up to its answer. So we don't really know how, especially in those early days, how it produced a case that doesn't even exist. So that's a hallucination. It just made it literally as well. It just made it up. So that's what illumination is when it's creating something that doesn't exist.

00;08;39;20 - 00;09;10;14
Host
And that's a great segway, because I think so far. And when we talk about the administrative burden that it could, I could lift, I don't know, many people who would say, oh, no, I want to keep all the paperwork I have to do like that. That seems fairly uncontroversial. But when we move into that area of, of hallucinations, of, you know, attorneys potentially using it to create motions that are citing cases that aren't even real or even in a more nefarious way when evidence is created that is not real.

00;09;10;19 - 00;09;12;26
Host
Are you aware of of that happening yet?

00;09;12;28 - 00;09;46;05
Michael Navin
So it's, you know, it's it's becoming more and more, you know, the courts are, starting to see these trickle in, where there's questions about whether evidence is real or not. And, an example that I would give an involves, situation in Maryland where a principal, had found out some information about a director potentially doing something wrong with, some funds and the principal was going to, basically, turn the turn the athletic director in or report it to the authorities for more investigation.

00;09;46;05 - 00;10;11;08
Michael Navin
Well, the the athletic director, well, then what came out was this this voicemail, made all its rounds on the internet, on social media of this principal saying racist and anti-Semitic things and, oh, my gosh. And and, the principal was put on leave and he, the principal was threatened and his family was threatened, and and then they brought in a forensic, forensic scientist to evaluate the audio.

00;10;11;08 - 00;10;33;19
Michael Navin
And it turns out that AI had created that the, the, the, athletic director had recorded the, the voice of the, of the principal and then just used AI to to make up this, this audio that was so close to being, What what that what that, person, what that principal sounded like. And then he was ended up in it.

00;10;33;21 - 00;11;02;19
Michael Navin
I believe it's it's probably the first case where, a person used AI in a commission of a crime. So, you know, that's one example of using it, but there are, you know, there are potentials for, you know, it's just so easy to create something that isn't real anymore. You know, it's so easy to. And you may see this in things like, you know, custody case, you know, custody cases, you know, voicemails and audios that may not be real.

00;11;02;19 - 00;11;20;19
Michael Navin
And it's like, how are courts going to have how are courts going to deal with that? How are they going to authenticate that? You know, and who's responsible for authenticating that? And how is that going to affect, you know, delays? Is that going to cause delays? Is that going to so there's so many, unknowns at this point on this, but it's it's something that is getting through the courts.

00;11;20;19 - 00;11;24;22
Michael Navin
And I assume at some point these things will be weeded out.

00;11;24;24 - 00;11;34;29
Host
You have provided a list of recommendations for, you know, court systems, you know, just trying to kind of wrap their arms around this. Where do we start? We hear these examples. They're alarming. What do we do?

00;11;35;02 - 00;11;57;19
Michael Navin
You know, I, I don't know that I have an answer as to what you know, what we do. And I think it's so foggy right now. The question becomes, let's say evidence comes in that's questionable. Well, you know, well then there's this brings up a question about a digital divide, you know, to where if one person makes something up and the other person doesn't have the funds or, or the money to, to hire a forensic scientist to evaluate that evidence.

00;11;57;19 - 00;12;14;18
Michael Navin
Well, what you know, but somebody else does have money to do that. Well, then, is that fair? You know, I mean, so so where do we go from here? But are we are we going to start, you know, our courts now going to have to, hire forensic evidence, you know, that just sit in a courtroom. I don't know that that's feasible.

00;12;14;18 - 00;12;37;18
Michael Navin
I don't know if there's funding for that. I don't know if that's even the court's role, but it's really a foggy situation. And sure, I don't I don't know that we have an answer for that yet. Even over the last year that I've been looking at this, we have not really come up with a clear answer. And I just think that we need to see more cases out there that deal with this in order for courts to get more of a precedent and more reviewing on on the thought process behind it and what to do.

00;12;37;25 - 00;12;54;08
Host
And the technology is so quickly evolving, too. This has been a great kind of, you know, high level, 30,000ft, you know, conversation about it for folks who are listening and they want to dig in more. I know the National Center for State Courts has some great resources available. Where can our listeners go?

00;12;54;11 - 00;13;26;14
Michael Navin
So there's a if you go to the National Center's website, it's www.ncsc.org/ai. Very simple. You can go to that. And that's the, kind of the gateway to all the resources we have where you can go and look to see what are other states doing with policy and guidelines and case law and decisions and whenever we get new things and we add it to that website, and, we have all kinds of, resources that we've produced, webinars, reports and videos.

00;13;26;14 - 00;13;44;09
Michael Navin
And we have a monthly AI Implementers Forum that people can join. Court staff can join. And, it's a great place to learn. What are the courts are doing with AI? It's also a great place since we have judges in there to to under to to really get a feel for what, you know, what our judges seeing is there, you know, that are coming up onto their docket.

00;13;44;12 - 00;13;59;23
Michael Navin
And so that's a great place to go to. And we also have a collaboration between the Thomas Reuter Institute and the NCC, on AI policy. And once a month we have a webinar on a various AI topic. And, it's a great place. It again, is for free. It's the, it's just once a month, I believe it's on Wednesday.

00;13;59;23 - 00;14;13;19
Michael Navin
It's on a Wednesday. And that's a great place to learn. You sign up for it and you can actually all those webinars are recorded. So we can go you can go back and watch the, or watch the prior ones to, to really get a sense as to what's going on. And the topics change every, every month as they should.

00;14;13;19 - 00;14;22;07
Michael Navin
And, you know, sometimes things are rehashed, not because they're there, they're duplicating. It's just because it's changed, because the technology is changing so quickly.

00;14;22;15 - 00;14;28;19
Host
Well, no doubt your your knowledge is, is in high demand. And we appreciate you taking a few minutes with us this afternoon.

00;14;28;19 - 00;14;29;28
Michael Navin
You're welcome. Thank you for having me.

00;14;29;28 - 00;14;38;00
Host
Yeah. Thank you. Michael, for more information on the Tennessee Court System, you can find that at tncourts.gov. Thank you so much for joining us.