Popular in Law

Law

Diversity and Social Mobility in the UK legal profession

Law Society I. Stephanie Boyce

Internet Law

The Law of AI for Good

University of San Diego Orly Lobel

Human Rights Law

Transforming Restorative Justice for Transitional Settings

University of Nottingham Kerry Clamp

Law

Surrey’s Silence: Subpart F and the Swiss Subsidiary Tax that Never Was

Boston University Steven Dean

Criminal Law

Prosecuting Misdemeanors

American University Jenny Roberts

Human Rights Law

Siege Starvation: A War Crime of Societal Torture

Tufts University Tom Dannenbaum

Corporate Law

Data Valuation and Law

University of Southern California Jordan Barry

Artificial Intelligence and the Crises of Judicial Power

View more insights in Professional - Law - Internet Law

View more figures from University of Essex

View the page of Giulia Gentile

Read the Study

Courts across the world are experiencing efficiency and contestation crises. Giulia Gentile discusses the role of automated decision-making and artificial intelligence in addressing and potentially solving these very crises of judicial power.

Image courtesy of interviewee. July 29, 2024

Generating Workspace

Export as PDF

In a discussion about the intersection of artificial intelligence (AI) and judicial power, Julia outlines her research focused on how AI might address current crises in judicial systems worldwide. She identifies two primary issues: a crisis of efficiency, with an increasing backlog of cases leading to delays in justice; and a crisis of contestation, where courts face criticism for their interpretation of laws that some argue exceeds their mandate.

Julia explains that while AI has generated considerable hype as a potential solution to these challenges, it is essential to examine both courts' roles as users and regulators of technology. Courts can integrate AI to improve efficiency and potentially improve the decision-making process, but reliance on AI also raises significant concerns. Issues include the need for digital literacy among the public, the potential for increased inequalities in access to justice, and the impact of public investments in technology without guaranteed improvements in justice delivery.

From the judges' perspective, AI could streamline case processing and aid in drafting decisions, but there are concerns regarding the dehumanization of justice and whether legal concepts can be appropriately reduced to computational formulas. Trust remains a crucial aspect, as the public's perception of fairness and transparency in the judicial process could be undermined by algorithmic decision-making.

In conclusion, Julia urges caution against viewing AI as a panacea for judicial inefficiencies, emphasizing the importance of ongoing research into trust in AI technologies and their implications for justice systems. Future research may investigate trust dynamics across different fields and the need for cohesive AI regulation that addresses shared values and risks across sectors.

Language:

Export as PDF

X-TIMESTAMP-MAP:LOCAL=00:00:00.000,MPEGTS=900000

00:00.105 --> 00:02.805
In essence, there are two main problems

00:02.915 --> 00:05.205
that are affecting, justice

00:05.465 --> 00:07.925
and courts, worldwide.

00:08.295 --> 00:11.365
First, we have a crisis of efficiency.

00:11.905 --> 00:15.405
Courts, are not capable of

00:15.925 --> 00:18.965
handling the increasing backlog of cases, that

00:19.585 --> 00:21.405
is piling up in front of them.

00:21.905 --> 00:25.245
And inevitably, this leads to delays injustice.

00:25.985 --> 00:29.485
But on the other hand, courts are also increasingly attacked

00:29.835 --> 00:33.685
both, by, politicians, but in a way,

00:33.865 --> 00:35.685
and interestingly also by scholars.

00:36.265 --> 00:38.405
Um, of course, the reasons

00:38.585 --> 00:40.805
for these criticisms are different,

00:41.185 --> 00:43.925
but essentially at the core of,

00:44.275 --> 00:48.885
this attack against courts is the exercise

00:48.885 --> 00:51.085
of judicial power in manners

00:51.085 --> 00:54.925
that do not reflect the literal interpretation of text.

00:55.265 --> 01:00.005
So essentially, when courts, interpret the text,

01:00.315 --> 01:01.565
that is in front of them

01:01.785 --> 01:05.045
and, on which they have received legal questions, they,

01:05.065 --> 01:06.925
of course, may use interpretive methods.

01:07.425 --> 01:11.445
And these interpretative, methodology when it comes

01:11.445 --> 01:15.765
to courts has been, criticized because going too far

01:16.065 --> 01:18.765
and altering in a way the separation of powers.

01:19.345 --> 01:20.765
And in the last years,

01:20.995 --> 01:24.685
what we have seen is really the emergence of a hype around,

01:24.985 --> 01:29.005
the use of AI to tackle, the problems

01:29.005 --> 01:33.085
that are undermining, the ability of,

01:33.385 --> 01:37.525
our societies to obtain justice from our court systems.

01:37.715 --> 01:39.925
This is something that I explore in the paper,

01:40.225 --> 01:43.365
and I have essentially,

01:43.505 --> 01:46.645
teased out the two solves we can say,

01:46.745 --> 01:50.605
of courts when dealing with, both inefficiency

01:50.605 --> 01:52.885
and the contestation prob, problems.

01:53.185 --> 01:57.125
On the one hand, courts, can be users of the technology,

01:57.135 --> 01:59.445
which means that the technology is embedded

02:00.145 --> 02:01.205
in court systems.

02:01.905 --> 02:05.325
And, through these embedding of the technology,

02:05.335 --> 02:06.605
there are consequences

02:06.665 --> 02:11.365
and implications in, the way in which courts

02:12.005 --> 02:13.365
exercise their judicial power.

02:13.785 --> 02:15.285
And this is because they're subject

02:15.545 --> 02:19.965
to the algorithmic governance, the power of data.

02:20.705 --> 02:25.325
And we can say also the, technology,

02:25.475 --> 02:27.605
framing of legal questions,

02:28.025 --> 02:30.805
but courts are also regulator themselves.

02:31.385 --> 02:32.885
Uh, and what do I mean by this?

02:33.145 --> 02:37.365
Uh, well, what happens is essentially that courts,

02:37.555 --> 02:40.965
have been dealing already with many instances

02:41.105 --> 02:43.445
of legal questions around the use

02:43.505 --> 02:47.085
of automated decision making and ai.

02:47.705 --> 02:51.005
Uh, and because as we know, the regulation

02:51.065 --> 02:55.045
around AI is slowly emerging, we have had

02:55.065 --> 02:58.925
so far only piecemeal approaches to AI regulations,

02:59.145 --> 03:01.485
and, legislation across the world.

03:02.025 --> 03:03.245
Courts have stepped in

03:03.545 --> 03:07.565
and have applied preexisting frameworks to questions

03:07.565 --> 03:09.725
of automated decision making and ai.

03:10.145 --> 03:13.765
So, in, in so doing, courts have become themselves

03:14.275 --> 03:19.045
regulator of ai, vis-a-vis, we can say the,

03:19.605 --> 03:21.925
inactive role of the regulator.

03:22.265 --> 03:26.885
And in the paper I explore, whether AI can solve,

03:27.425 --> 03:29.805
both inefficiency crisis as well

03:29.865 --> 03:33.485
as the contestation crisis from the angle of courts as

03:34.335 --> 03:36.605
regulators and as users.

03:37.265 --> 03:41.045
So thinking, for instance, at inefficiency question

03:41.045 --> 03:43.045
that we have explored so far,

03:43.665 --> 03:48.085
and, whether AI can really help,

03:48.115 --> 03:52.165
justice systems across the world to fix, these problems,

03:52.665 --> 03:55.925
we need to consider essentially that the answer

03:55.945 --> 03:58.245
to this question, is first of all influenced

03:58.305 --> 04:00.765
by different degrees of techno phobia,

04:00.785 --> 04:02.165
techno filia in the sense

04:02.195 --> 04:05.285
that legal traditions developed in,

04:05.305 --> 04:08.565
the various jurisdictions across the world will influence

04:08.945 --> 04:11.685
the answer that we give to this question in the sense that,

04:12.225 --> 04:14.605
systems that tend

04:14.605 --> 04:17.925
to work well might be more reluctant, of course,

04:17.945 --> 04:22.605
to give power to algorithms while systems instead that have

04:23.245 --> 04:26.085
entrenched problems around justice might be more,

04:26.305 --> 04:27.365
in a way technophile,

04:27.485 --> 04:29.285
I have more hope towards the technology.

04:29.705 --> 04:32.885
But having said that, which is just, we can say,

04:33.505 --> 04:36.805
almost a stance, a point of view around technology,

04:36.825 --> 04:40.365
we need to look at the substance and looking at efficiency.

04:40.785 --> 04:44.965
Um, we need to, essentially reflect,

04:45.145 --> 04:49.805
on the problem of whether AI is, a solution

04:49.805 --> 04:54.165
or not to inefficiency, from the, perspective

04:54.225 --> 04:57.645
of the parties involved in, the trial.

04:58.265 --> 05:02.365
And in this case, we have, of course, the judge,

05:02.585 --> 05:04.125
the judges and the parties.

05:04.865 --> 05:08.285
So, from an inefficiency from,

05:08.505 --> 05:12.245
the efficiency perspective, embedding AI in,

05:12.345 --> 05:13.605
in court systems.

05:14.225 --> 05:17.205
So, from the angle of courts as regulators,

05:17.715 --> 05:20.485
what we see is that, the parties

05:21.285 --> 05:22.685
starting from the party's perspective,

05:22.745 --> 05:24.525
we can certainly have certain advantages.

05:24.555 --> 05:29.365
Parties might, we can say re receive,

05:29.715 --> 05:33.925
deliberations more quickly, may be able to,

05:34.035 --> 05:38.205
benefit from interfaces that have algorithms.

05:38.665 --> 05:40.165
Uh, and therefore,

05:40.355 --> 05:42.925
essentially we can see an increase in the speed

05:43.185 --> 05:44.365
of the delivery of justice.

05:45.105 --> 05:49.485
But, there are several problems that come with, the,

05:49.545 --> 05:52.765
use the deployment of AI in justice systems.

05:53.345 --> 05:56.765
And the first one is certainly, digital literacy,

05:57.265 --> 05:59.285
or I would say digital literacy.

05:59.745 --> 06:01.325
Um, there are recent studies

06:01.475 --> 06:02.965
that have been produced in Australia,

06:03.065 --> 06:05.605
but also in the UK that, highlight

06:05.825 --> 06:08.925
how there is still an ongoing, gap

06:08.955 --> 06:12.365
between different parts of the society from the angle

06:12.365 --> 06:13.925
of digital literacy.

06:14.225 --> 06:18.645
And what is evident is that unfortunately, those who are,

06:19.025 --> 06:22.845
the most digital illiterate are actually those

06:22.875 --> 06:27.525
that will benefit the most from, an enhanced access

06:27.525 --> 06:30.805
to justice, those who are more vulnerable

06:30.825 --> 06:33.805
and might not have access to the technology, for instance.

06:34.385 --> 06:38.645
So in a way, the problem of enhancing efficiency,

06:38.825 --> 06:43.445
by using AI is, is not, is not corrected,

06:43.545 --> 06:45.085
is actually exacerbated

06:45.085 --> 06:48.645
because the most vulnerable that need access

06:48.645 --> 06:50.525
to justice will not be able to do.

06:50.545 --> 06:53.045
So. Then there's also another element

06:53.105 --> 06:55.445
to take into account when thinking about

06:56.165 --> 06:58.205
bringing AI in the court system and

06:58.205 --> 07:00.685
therefore making courts users of the technology,

07:01.215 --> 07:03.805
which is the question of public investments.

07:04.345 --> 07:05.765
Uh, and what do I mean by these?

07:05.875 --> 07:08.685
Essentially, looking at

07:08.715 --> 07:11.805
what is happening on the ground in different jurisdictions,

07:11.945 --> 07:15.805
the UK being a great example on this is that,

07:16.105 --> 07:17.285
the last 10 years or

07:17.545 --> 07:21.765
or so have, already, laid the foundations

07:21.825 --> 07:23.365
for the digitization of courts.

07:23.905 --> 07:26.565
Um, in the uk, for example, a, the program

07:26.625 --> 07:31.045
to digitizes courts has started in 2014,

07:31.625 --> 07:35.805
so it has been essentially 10 years, that the public,

07:35.905 --> 07:39.405
of course is financing, trans the transformation

07:39.425 --> 07:40.445
of UK courts.

07:40.865 --> 07:43.925
Um, and this is, also quite interesting,

07:43.955 --> 07:47.885
because this, development of technology

07:47.945 --> 07:52.725
for UK courts has happened in the aftermaths of,

07:52.905 --> 07:54.485
the cuts to legal aid.

07:55.025 --> 07:59.125
So, clearly we see that the thinking around the use

07:59.125 --> 08:03.765
of technologies was, that of looking at,

08:03.875 --> 08:08.165
this new tools as the solution for, solving the problems

08:08.225 --> 08:09.925
of the UK justice system.

08:10.585 --> 08:13.285
Um, but, having said that, it's clear

08:13.355 --> 08:15.325
that investments are still ongoing,

08:15.465 --> 08:17.485
and there are recent document produced

08:17.505 --> 08:20.685
by the National Audit Office that indicates

08:20.685 --> 08:24.165
that this programs will, last for longer than expected.

08:24.905 --> 08:28.925
Uh, not withstanding the public investment has been even

08:28.955 --> 08:32.085
increased to support these programs, meaning

08:32.085 --> 08:34.725
that essentially the justice gap

08:34.755 --> 08:37.965
that the technology is supposed to be feeling is not filled.

08:38.585 --> 08:41.125
Um, so again, I think this is something

08:41.275 --> 08:45.965
that perhaps may help us to bring everything in context,

08:46.425 --> 08:50.445
and also to, perhaps nuance the hype around ai.

08:51.065 --> 08:55.645
And then moving briefly, on the judge's perspective.

08:56.155 --> 08:58.605
Well, from the justice, from the judge's perspective,

08:58.985 --> 09:01.485
the question of efficiency is also quite important

09:01.485 --> 09:06.405
because judges could rely on ai, for instance, to,

09:06.625 --> 09:10.845
have, a support in, drafting,

09:10.845 --> 09:14.325
their decisions to predict, to, to,

09:14.585 --> 09:17.685
we can say benefit from predictions, that are produced

09:17.685 --> 09:18.685
by algorithms.

09:19.225 --> 09:23.285
Um, or we can think of AI as merely a,

09:23.345 --> 09:26.925
an administrative, tool that is used, for instance,

09:26.945 --> 09:29.805
to filter cases that go in front of courts

09:30.265 --> 09:33.525
and then allow the algorithm to assign, for example,

09:33.775 --> 09:35.365
these cases to different judges.

09:35.585 --> 09:38.085
So, also from this perspective, of course,

09:38.145 --> 09:42.205
judges would have, can potentially have some benefits,

09:42.865 --> 09:45.165
deriving from the use of ai,

09:45.705 --> 09:48.565
but we, we need to measure these benefits against three,

09:48.805 --> 09:51.765
I think very important, consideration.

09:51.785 --> 09:55.125
So first of all, and this is something that needs

09:55.125 --> 09:58.005
to be studied further, I think is the question

09:58.105 --> 10:00.485
of the dehumanization of justice.

10:01.665 --> 10:03.565
How would the public react

10:04.025 --> 10:07.845
to an increasingly digitized judge, a judge that,

10:08.345 --> 10:10.245
is not, human in a way,

10:10.305 --> 10:12.525
or at least is only partially human?

10:12.665 --> 10:16.285
And I think this is something that, goes quite deep in,

10:16.345 --> 10:19.125
the issue of trust,

10:19.125 --> 10:21.045
public trust towards justice systems

10:21.625 --> 10:23.925
and what it means really to be judged by a human.

10:24.595 --> 10:28.365
Then there, there's the question of, the law

10:28.385 --> 10:31.325
and computation can law be really computed?

10:31.825 --> 10:36.005
Um, and this means that we think that algorithms,

10:36.585 --> 10:40.125
can essentially make the law as mathematical expression,

10:40.545 --> 10:42.245
and this is necessary because un

10:42.245 --> 10:45.205
unless we do so, we cannot really have algorithms

10:45.555 --> 10:47.125
playing with the law, right?

10:47.225 --> 10:48.285
So the law needs

10:48.305 --> 10:51.485
to be transformed into mathematical expressions.

10:52.105 --> 10:54.445
Um, but this is, I think, quite problematic

10:54.445 --> 10:57.885
because this takes away very important aspects

10:57.905 --> 11:02.245
of the law linked to, principles such as equity of,

11:02.265 --> 11:04.565
or, societal considerations

11:04.785 --> 11:07.325
or even think about common law that,

11:07.885 --> 11:11.805
derives from the sensitivity of judges towards the context

11:11.985 --> 11:13.685
of individual cases, right?

11:14.185 --> 11:18.165
All of this will be taken away by, courts

11:18.165 --> 11:19.845
that heavily rely on algorithms.

11:20.425 --> 11:22.325
Um, and last,

11:22.865 --> 11:26.205
but not least, as I said already, I think that,

11:26.705 --> 11:30.725
the question of trust is really, the main problem here,

11:30.945 --> 11:34.725
that derives from this transformation that requires,

11:35.165 --> 11:37.765
a lot, I think of, careful

11:38.525 --> 11:39.685
analysis and and reflection.

11:40.185 --> 11:42.885
Um, and of course there are also other considerations for,

11:43.035 --> 11:44.685
from the angle of contestation.

11:45.265 --> 11:50.165
Um, and, just to, close on this, I think that while,

11:50.825 --> 11:53.935
AI could, for example,

11:54.205 --> 11:57.695
make judgements more predictable, meaning that, for example,

11:57.715 --> 12:02.415
the contestation critique based on, courts,

12:02.835 --> 12:07.335
we can say exercising their judicial powers to broadly,

12:07.835 --> 12:10.975
and, and therefore altering the separation of powers.

12:10.975 --> 12:13.335
Well, I think that, the reality is

12:13.335 --> 12:16.175
that AI could exacerbate constitution against courts

12:16.175 --> 12:18.055
because the veil created

12:18.275 --> 12:21.455
by algorithmic governance would make court less transparent,

12:21.845 --> 12:23.135
less open to the public.

12:23.595 --> 12:25.855
And in my view, this can only,

12:26.225 --> 12:29.495
exacerbate really contestation against courts, the view

12:29.495 --> 12:34.415
that AI would be a magic wand that could, be deployed

12:34.675 --> 12:36.615
to, enhance justice

12:37.475 --> 12:41.815
and make sure that courts, deliver,

12:41.835 --> 12:44.695
better protection to individuals should be,

12:44.835 --> 12:45.855
reconsidered.

12:46.035 --> 12:51.015
So, and this is something that, and, and,

12:51.075 --> 12:55.575
and my work really addresses both scholars

12:55.635 --> 12:58.575
who are working in the area, but also regulators,

12:58.675 --> 12:59.815
and courts themselves.

13:00.355 --> 13:03.775
We have seen, for instance, in December, 2023,

13:03.795 --> 13:06.615
the publication of, a guidance on the use

13:06.615 --> 13:09.655
of AI in the judiciary by her majesty, courts

13:09.655 --> 13:11.655
and tribunals, which is

13:11.655 --> 13:14.895
of course a very welcome step towards,

13:15.645 --> 13:16.935
modernizing courts

13:16.955 --> 13:19.975
and making them, we can say up

13:20.035 --> 13:22.455
to current times and technology.

13:22.755 --> 13:25.175
But at the same time, I think that,

13:25.635 --> 13:30.455
we should not fall prey of the fallacy that AI can,

13:31.605 --> 13:35.095
necessarily, produce better justice

13:35.395 --> 13:36.535
and better protection.

13:37.035 --> 13:40.895
So having, laid down, hopefully a, a quite strong case

13:40.995 --> 13:44.615
for, for these, in terms of future research,

13:45.215 --> 13:49.535
I am currently exploring, other aspects of this problem,

13:49.835 --> 13:52.695
thinking for example, about the question of trust

13:53.235 --> 13:55.895
and how individuals trust

13:56.075 --> 14:00.095
or not AI technologies, in the light of the fact

14:00.095 --> 14:02.535
that we are still in the middle earth, right?

14:02.835 --> 14:06.295
The technology is still developing, is still not there,

14:06.915 --> 14:07.935
and, and

14:07.935 --> 14:12.135
therefore we are all exposed to a partially,

14:12.435 --> 14:15.255
well-functioning, set of tools that

14:15.255 --> 14:18.655
therefore cannot really live up to our expectations of

14:19.155 --> 14:23.175
and enhancing efficiency and increasing productivity

14:23.175 --> 14:26.855
because there is still an ample margin for mistakes.

14:27.595 --> 14:30.175
Um, so my research,

14:30.435 --> 14:33.975
is focusing on trust is also focusing on how trust

14:34.675 --> 14:39.335
is perceived in various disciplines, with reference

14:39.335 --> 14:40.415
to the use of ai.

14:40.515 --> 14:45.335
So for example, computer scientists of course may have,

14:45.575 --> 14:48.775
a different approach as to trust towards AI technology

14:49.535 --> 14:50.815
compared to lawyers, but

14:50.815 --> 14:53.485
Then lawyers might have a different approach to,

14:53.605 --> 14:56.205
trusting AI compared to doctors

14:56.665 --> 14:59.805
and other medical professions and, and so on.

15:00.185 --> 15:03.925
But then if we think about AI regulation that applies

15:03.945 --> 15:07.885
to all these sectors, I think that, we need

15:07.885 --> 15:11.005
to take into account whether there is a common core

15:11.545 --> 15:15.925
of values, but also risks that should be effectively tackled

15:15.945 --> 15:17.005
by AI regulation.

15:17.585 --> 15:19.045
Um, and,

15:19.105 --> 15:23.245
and this project, can also of course be, of interest

15:23.585 --> 15:28.485
for, state bodies as well, as well as of,

15:28.905 --> 15:32.405
as well as state branches that consider how

15:32.405 --> 15:34.605
to embed AI in their own activities.

00:00 - In essence, there are two main problems that are affecting, justice and courts, worldwide. First, we have a crisis of efficiency. Courts, are not capable of handling the increasing backlog of cases, that

00:19 - is piling up in front of them. And inevitably, this leads to delays injustice. But on the other hand, courts are also increasingly attacked both, by, politicians, but in a way, and interestingly also by scholars. of course, the reasons for these criticisms are different, but essentially at the core of, this attack against courts is the exercise of judicial power in manners

00:51 - that do not reflect the literal interpretation of text. So essentially, when courts, interpret the text, that is in front of them and, on which they have received legal questions, they, of course, may use interpretive methods. And these interpretative, methodology when it comes to courts has been, criticized because going too far and altering in a way the separation of powers. And in the last years, what we have seen is really the emergence of a hype around,

01:24 - the use of AI to tackle, the problems that are undermining, the ability of, our societies to obtain justice from our court systems. This is something that I explore in the paper, and I have essentially, teased out the two solves we can say, of courts when dealing with, both inefficiency and the contestation prob, problems. On the one hand, courts, can be users of the technology, which means that the technology is embedded

02:00 - in court systems. And, through these embedding of the technology, there are consequences and implications in, the way in which courts exercise their judicial power. And this is because they're subject to the algorithmic governance, the power of data. And we can say also the, technology, framing of legal questions, but courts are also regulator themselves.

02:31 - and what do I mean by this? well, what happens is essentially that courts, have been dealing already with many instances of legal questions around the use of automated decision making and ai. and because as we know, the regulation around AI is slowly emerging, we have had so far only piecemeal approaches to AI regulations, and, legislation across the world. Courts have stepped in

03:03 - and have applied preexisting frameworks to questions of automated decision making and ai. So, in, in so doing, courts have become themselves regulator of ai, vis-a-vis, we can say the, inactive role of the regulator. And in the paper I explore, whether AI can solve, both inefficiency crisis as well as the contestation crisis from the angle of courts as regulators and as users. So thinking, for instance, at inefficiency question

03:41 - that we have explored so far, and, whether AI can really help, justice systems across the world to fix, these problems, we need to consider essentially that the answer to this question, is first of all influenced by different degrees of techno phobia, techno filia in the sense that legal traditions developed in, the various jurisdictions across the world will influence the answer that we give to this question in the sense that,

04:12 - systems that tend to work well might be more reluctant, of course, to give power to algorithms while systems instead that have entrenched problems around justice might be more, in a way technophile, I have more hope towards the technology. But having said that, which is just, we can say, almost a stance, a point of view around technology, we need to look at the substance and looking at efficiency. we need to, essentially reflect,

04:45 - on the problem of whether AI is, a solution or not to inefficiency, from the, perspective of the parties involved in, the trial. And in this case, we have, of course, the judge, the judges and the parties. So, from an inefficiency from, the efficiency perspective, embedding AI in, in court systems. So, from the angle of courts as regulators, what we see is that, the parties

05:21 - starting from the party's perspective, we can certainly have certain advantages. Parties might, we can say re receive, deliberations more quickly, may be able to, benefit from interfaces that have algorithms. and therefore, essentially we can see an increase in the speed of the delivery of justice. But, there are several problems that come with, the, use the deployment of AI in justice systems.

05:53 - And the first one is certainly, digital literacy, or I would say digital literacy. there are recent studies that have been produced in Australia, but also in the UK that, highlight how there is still an ongoing, gap between different parts of the society from the angle of digital literacy. And what is evident is that unfortunately, those who are, the most digital illiterate are actually those

06:22 - that will benefit the most from, an enhanced access to justice, those who are more vulnerable and might not have access to the technology, for instance. So in a way, the problem of enhancing efficiency, by using AI is, is not, is not corrected, is actually exacerbated because the most vulnerable that need access to justice will not be able to do. So. Then there's also another element to take into account when thinking about

06:56 - bringing AI in the court system and therefore making courts users of the technology, which is the question of public investments. and what do I mean by these? Essentially, looking at what is happening on the ground in different jurisdictions, the UK being a great example on this is that, the last 10 years or or so have, already, laid the foundations for the digitization of courts.

07:23 - in the uk, for example, a, the program to digitizes courts has started in 2014, so it has been essentially 10 years, that the public, of course is financing, trans the transformation of UK courts. and this is, also quite interesting, because this, development of technology for UK courts has happened in the aftermaths of, the cuts to legal aid. So, clearly we see that the thinking around the use

07:59 - of technologies was, that of looking at, this new tools as the solution for, solving the problems of the UK justice system. but, having said that, it's clear that investments are still ongoing, and there are recent document produced by the National Audit Office that indicates that this programs will, last for longer than expected. not withstanding the public investment has been even increased to support these programs, meaning

08:32 - that essentially the justice gap that the technology is supposed to be feeling is not filled. so again, I think this is something that perhaps may help us to bring everything in context, and also to, perhaps nuance the hype around ai. And then moving briefly, on the judge's perspective. Well, from the justice, from the judge's perspective, the question of efficiency is also quite important because judges could rely on ai, for instance, to, have, a support in, drafting,

09:10 - their decisions to predict, to, to, we can say benefit from predictions, that are produced by algorithms. or we can think of AI as merely a, an administrative, tool that is used, for instance, to filter cases that go in front of courts and then allow the algorithm to assign, for example, these cases to different judges. So, also from this perspective, of course, judges would have, can potentially have some benefits,

09:42 - deriving from the use of ai, but we, we need to measure these benefits against three, I think very important, consideration. So first of all, and this is something that needs to be studied further, I think is the question of the dehumanization of justice. How would the public react to an increasingly digitized judge, a judge that, is not, human in a way, or at least is only partially human?

10:12 - And I think this is something that, goes quite deep in, the issue of trust, public trust towards justice systems and what it means really to be judged by a human. Then there, there's the question of, the law and computation can law be really computed? and this means that we think that algorithms, can essentially make the law as mathematical expression, and this is necessary because un unless we do so, we cannot really have algorithms

10:45 - playing with the law, right? So the law needs to be transformed into mathematical expressions. but this is, I think, quite problematic because this takes away very important aspects of the law linked to, principles such as equity of, or, societal considerations or even think about common law that, derives from the sensitivity of judges towards the context of individual cases, right?

11:14 - All of this will be taken away by, courts that heavily rely on algorithms. and last, but not least, as I said already, I think that, the question of trust is really, the main problem here, that derives from this transformation that requires, a lot, I think of, careful analysis and and reflection. and of course there are also other considerations for, from the angle of contestation.

11:45 - and, just to, close on this, I think that while, AI could, for example, make judgements more predictable, meaning that, for example, the contestation critique based on, courts, we can say exercising their judicial powers to broadly, and, and therefore altering the separation of powers. Well, I think that, the reality is that AI could exacerbate constitution against courts because the veil created by algorithmic governance would make court less transparent,

12:21 - less open to the public. And in my view, this can only, exacerbate really contestation against courts, the view that AI would be a magic wand that could, be deployed to, enhance justice and make sure that courts, deliver, better protection to individuals should be, reconsidered. So, and this is something that, and, and, and my work really addresses both scholars

12:55 - who are working in the area, but also regulators, and courts themselves. We have seen, for instance, in December, 2023, the publication of, a guidance on the use of AI in the judiciary by her majesty, courts and tribunals, which is of course a very welcome step towards, modernizing courts and making them, we can say up to current times and technology.

13:22 - But at the same time, I think that, we should not fall prey of the fallacy that AI can, necessarily, produce better justice and better protection. So having, laid down, hopefully a, a quite strong case for, for these, in terms of future research, I am currently exploring, other aspects of this problem, thinking for example, about the question of trust and how individuals trust or not AI technologies, in the light of the fact

14:00 - that we are still in the middle earth, right? The technology is still developing, is still not there, and, and therefore we are all exposed to a partially, well-functioning, set of tools that therefore cannot really live up to our expectations of and enhancing efficiency and increasing productivity because there is still an ample margin for mistakes. so my research, is focusing on trust is also focusing on how trust

14:34 - is perceived in various disciplines, with reference to the use of ai. So for example, computer scientists of course may have, a different approach as to trust towards AI technology compared to lawyers, but Then lawyers might have a different approach to, trusting AI compared to doctors and other medical professions and, and so on. But then if we think about AI regulation that applies to all these sectors, I think that, we need

15:07 - to take into account whether there is a common core of values, but also risks that should be effectively tackled by AI regulation. and, and this project, can also of course be, of interest for, state bodies as well, as well as of, as well as state branches that consider how to embed AI in their own activities.

Faculti AI can make mistakes