-
Artificial Intelligence (AI) is a technology, which will infiltrate
most aspects of our society. Although most people associate it
with machine learning, it is a much broader group of methods and
approaches. The enthusiasm of solving every problem in justice
through technology fostered the introduction of Information and
Communication Technologies (ICT) in the judiciary. Countries
introduced ICT in their justice systems in order to improve both
efficiency of justice and accessibility to justice. In this context,
policy makers examine how AI can be used in courts to facilitate
both the administration of justice and the adjudication of cases.
Observing this trend the European Commission for the Efficiency
of Justice of the Council of Europe has already adopted a text
setting out ethical principles relating to the use of AI in judicial
systems [
1
], which stands out among other similar European and
international texts [
2
] in that it identifies core principles to be
respected particularly in the field of AI and justice. Undoubtedly,
the development of AI tools will transform the process of
adjudication, because they will alter the way legal information is
used and communicated. Hence, the effects of the changes
introduced by AI have the potential of being much deeper and less
controllable. In examining the use of AI in a court information
system one has to address the question of what use can AI have
for courts; that is how AI can help parties of a case (litigants),
members of the registry (court officers) and judges. We assume
that AI for the judiciary should be ‘bespoke’. It should provide
solutions to the problems that a particular jurisdiction –using a
certain court information system- faces. The aim of this paper is
to examine the effective application of AI regarding the court
information system that was introduced in 2015 at the
administrative justice of Greece, the Integrated Administrative
Court Case Management System of Greece (IACCMS). Section 2
briefly introduces IACCMS and displays current and future
developments of the system, Section 3 presents potential AI
solutions for IACCMS and our concluding remarks are on Section
4.
2 Formation of the Integrated Administrative
Court Case Management System of Greece
In this section we will introduce IACCMS, but firstly we will
provide some preliminary remarks about administrative justice in
Greece and the introduction of ICT in it. These observations are
necessary before assessing the possibilities of AI in the following
section.
The Constitution of 1975 (revised in 1986, 2001, 2008 and 2019)
establishes three jurisdictions: civil, criminal and administrative.
Administrative justice, i.e. a court system that adjudicates on
disputes between the citizen and the administration, is organized
in three tiers: the courts of first instance, the courts of appeal and
the Council of the State (the Supreme Administrative Court), the
latter being responsible for the rational operation of
administrative justice. Furthermore, the General Commission of
the State for the Regular Administrative Courts, which is a
separate branch of senior administrative judges, monitors and
oversights the operation of administrative courts and assists them
without interfering with their judicial task. It is also the
competent authority (at operational level) that serves as an
intermediary between the Council of State and the rest of
administrative courts. Finally, the Ministry of Justice,
Transparency and Human Rights is entrusted, among other
competences, with the management of justice. It supervises the
administration of justice, dealing with organizational issues and
the infrastructure and it provides economic (through the budget
of the State) and administrative support to the judiciary.
After the re-establishment of administrative courts of first
instance and appeal in their current form (in 1985) and the transfer
of cases from the Council of State to them, the number of cases
that they adjudicate rose exponentially and as a result there were
delays in delivering justice. The Greek legislator adopted several
laws in order to speed up court proceedings in administrative
justice. In addition, the introduction of ICT in administrative
justice turned into a priority. From 2000 onwards the Council of
State started the computerization of its registrar for the workflow
of judicial proceedings before the court (case management) and
also the integration of existing applications. In 2006 the integrated
case management system of the Council of State was operational;
it contained the court’s jurisprudence, the workflows of the
registrar (computerization of proceedings), a management
information system (MIS) and a web site. On the other hand, the
computerization of administrative courts of first instance and
appeal was fragmented, since each court was perceived (from an
IT point of view) as an autonomous entity; that is each court was
responsible both for the administration of its data and for the
communication with external users, including other
administrative courts. The different information systems that the
courts had did not interoperate with information systems outside
of the judiciary (e.g. lawyers, public administration, citizens), thus
hindering the efficiency of justice. Furthermore, due to local
configuration of systems there was local implementation of work
flows, thus a need for an integrated electronic case management
system became evident. There were few digital archives and court
decisions were available only to the court that issued them, with
the exception of judgements of Council of State. Furthermore, the
lack of funds due to the drastic reduction in budgetary resources
made untenable the maintenance cost of infrastructure.
Beware of those issues the Ministry of Justice, Transparency and
Human Rights in its ‘Action Plan for e-justice and administrative
improvement’ [3, in Greek] decided that the Council of State
would lead the initiative to introduce IACCMS, which is
operational since 2015. For the purposes of this paper, we are
interested in three of the main components of IACCMS: i) the
court case management system, which coordinates the workflows
(business process) of all the courts of administrative justice, ii) the
uniform digital archive of all court decisions, accessible to all
judges and partially accessible (only of anonymized judgements)
to the general public and iii) a ‘one-stop-shop’ portal through
which external users can gather information from any court. Since
2018 IACCMS interoperates with the National Lawyers
Information System for the e-filing of cases (application to initiate
proceedings). In addition, this year (2020) IACCMS interoperates
with the information system of the Legal Council of the State (the
public body that defends the Administration before all courts) for
the e-filing of a case, the electronic delivery of court decisions and
for clearance of legal costs. Finally, articles 75 and 76 of law
4635/2019 stipulate that the communication between litigants and
administrative courts from 01.01.2021 will be by electronic means;
the law further establishes the electronic file of each case
(‘paperless court’).
Bearing in mind the above-mentioned developments at
administrative justice in Greece, in the following section we will
display the AI tools that could be used at IACCMS.
3 Development possibilities of AI for IACCMS
of Greece
In recent years a lot of companies made investments to search the
potential of AI, with considerable results (e.g. IBM’s Watson and
Google’s Alpha Go). AI innovations provide services to end users
(e.g. Apple’s Siri and Amazon’s Alexa voice assistants,
userspecific content provided by Netflix etc.), though the fact that
these services are based on the collection and processing of user
generated data, is raising concerns about the protection of
personal data [
4
] and generally about privacy in the digital age
[
5
]. However, it was the availability of massive amounts of
training data, along with breakthroughs in computational power,
improvements in machine learning algorithms and mobile
connectivity that fostered AI breakthroughs. The use of AI
technology for managing public services has the potential to make
public organizations more adaptive to a society with diverse and
changing needs and demands. AI will continue to transform
society, although there is skepticism about AI’s potential, due to
the fact that many of the grand claims made (e.g. autonomous
vehicles) have failed to become reality. The limits to the use of AI
are mainly due to the fact that the most widespread technique,
machine learning, is a powerful patent recognition tool, but lacks
fundamental cognitive abilities of the human brain. It is accurately
argued that “AI represents a concerted effort to understand the
complexity of human experience in terms of information
processes” [
6
]. Many stakeholders (litigants, lawyers, court
officers and judges) anticipate the introduction and development
of AI into the justice system, though each group expects different
results, which do not always converge. We identify two broad
categories: a) court-focused development of AI, i.e. tools that help
the justice system to improve its efficiency and quality and b)
litigant-focused development of A.I., i.e. tools that help either
selfrepresented litigants or lawyers to navigate legal processes; i.e. to
gather information about how the law applies to a particular case.
In the first category (court-focused development of AI), these
techniques could be used for court management purposes. For
example, an AI tool could scan and digitize documents submitted
by litigants, classify them into electronic files and match the
document of each litigant to corresponding e-files (creating new
e-files or linking to existing ones). It could further generate all
necessary court procedural documents and even distribute cases
to judges. IACCMS already has a court case management system
that generates some court procedural documents (namely, notices
and dockets). Since it is not yet mandatory to file a lawsuit by
electronic means, the workflow is still paper based, meaning that
regardless of the way a lawsuit is filed (paper or electronic form)
the court officer has to enter the data to the information system.
Law 4635/2019 stipulates that from 01.01.2021 the medium to
communicate with administrative courts will be electronic; AI
could be useful in automating the process of data entry. A
prerequisite is to create ontology of legal terms -a ‘Controlled
Legal Vocabulary’- that will provide the relevant metadata for the
legal annotation of each document perhaps using LegalDocML
(Akoma Ntoso); it would be beneficial to create an automatic
structuring and semantic indexing of legal documents written in
Greek [
7
]. The AI tool should process uniformly all unstructured
documents that are uploaded to IACCMS; it should further
automatically apply metadata and connect the document to a
particular electronic case [
8
]. For this purpose and to further
enhance interoperability, some of the existing tools that EU
provides to its member states could be used such as Controlled
Vocabularies , LegiVoc , Vocbench and, for the documents that
public administration sends, LEOS . Since it is very challenging to
process a legal document not developed by a lawyer, it would be
helpful to consider the development of techniques that will help
self-represented litigants to present facts in a more structured
way,
Having established an AI system such as the one already
described, it could be further used to improve the quantitative
processing of e-files. For example, an AI tool could identify certain
legal features (information extraction) in each case, assign it to
different case management tracks according to its complexity and
also group cases. Thus, it could streamline the processing of
judicial procedures in adjudicating a case, while also reducing
court staff and judges’ workload. This system could also provide
useful information to citizens (apart from prospective litigants
and lawyers), such as the duration of judicial proceedings for
different categories of cases in a particular court, the number of
pending cases of a particular nature etc. Moreover, machine
learning could be applied to the analysis of legal documents so as
to support judges in the solving of a dispute. For example, it could
create summaries of both the facts of the case and the arguments
that each litigant made in the documents (lawsuit, submissions,
and memorandums) [
9
]. However, before rolling out such an AI
tool thorough assessing should be preceded, because AI cannot
(yet) understand the context of a document; in other words, it
cannot perform legal reasoning and therefore AI could be misled
by minor variances in the data that it applies [
10
]. We
comprehend that although existing legal text analytics tools can
extract certain kinds of semantic legal information from legal
texts, they are not yet able to extract expert systems rules. It is
therefore necessary to further develop techniques that identify
argument related information in legal documents.
Additionally, AI tools could be used for the further development
of the uniform digital archive of all court decisions in IACCMS.
This database is accessible only to judges of administrative justice
and it contains all the judgements of the Council of State and the
judgements of all administrative courts of first instance and
appeal since 2015 and for some courts since 2000. It could be useful
to develop an AI tool for the retrieval of decisions related to a
particular case. In order to succeed in this endeavor an imperative
condition is the unambiguous identification of each court
decision; i.e. to ‘label’ or ‘tag’ each judgment with the appropriate
metadata (in a project similar to the one mentioned earlier about
legal documents). In the case of court decisions there is already in
place a useful tool provided by EU, the European Case Law
Identifier (ECLI) [
11
] that could be defined as HTTP-URI. The
judgments of the Council of State already use ECLI, and its
obligatory metadata. The potential is to use ECLI to all court
decisions of administrative justice and furthermore use at least
two of the optional metadata of this tool. Particularly, the field
“dcterms: abstract” contains a summary of the court decision and
the field “dcterms: description” contains descriptive elements, like
keywords. Both of these fields could be filled using the technique
of legal text analytics, which was previously described, i.e. an AI
tool that is able to ‘read’ the relevant parts of a court decision and
on the one hand create a summary of the judgement and on the
other hand apply the appropriate terms of the ‘Controlled Legal
Vocabulary’. It would further be desirable to create an AI tool
capable of anonymizing or pseudonymizing a court decision
before uploading it at the portal of IACCMS, where it would be
accessible for everyone to access. This tool should be able to
recognize natural persons and anonymize or pseudonymize their
personal data, while preserving the accuracy of the court decision.
Regarding the second category (litigant-focused development of
A.I.), AI could be used to provide relevant information to external
users of IACCMS. For example, using a question answering
system [
12
], citizens, self-represented litigants as well as lawyers
could gather information about the jurisprudence of a particular
category of cases (landmark decisions). Usually, prospective
litigants prior to filing a claim to initiate legal proceedings need
information such as the extent of their rights, court costs, length
of proceedings, the necessary procedural steps to be followed etc.
European judicial bodies of the Council of Europe encourage the
dissemination of information to citizens by courts in order to
facilitate access to justice [13 and 14]. The EU is consolidating
relevant information about member states in the European
ejustice portal . In order to provide personalized information to
self-represented litigants and lawyers the development of an
interactive information system that maintains dynamic
information (a difficult task, since courts are subject to almost
continuous change of the law) at the portal of IACCMS is
necessary. In a simplified form such a system could assist the
prospective self-represented litigant or the lawyer through a
dynamic questionnaire (dialogue modelling). Moreover, a
conversational bot (Chabot) could be developed to enable users to
interface with it by voice and language, as long as it is able to
analyze structured and unstructured data (text and human
speech). The advantage of the development of such an AI tool is
that it can improve its efficiency by learning from the recorded
dialogues, thus each time finding a more suitable answer to the
question posed. Obviously, sufficient safeguards are needed for
the protection of personal data of the users in the recorded
dialogues. The goal is to navigate the user to the ‘customized’
information that he seeks and only in rare cases direct the user to
a court officer who will provide the necessary information. To this
end the system should be simple enough for a user with basic
technological literacy to use; a complex system may delay the
expected advantages especially for self-represented litigants,
hence the testing phase with stakeholders is important. There is
currently significant attention on developing tools to assist people
in resolving legal disputes [
15
], however, an AI tool to predict the
outcome of court cases or even to analyze the quality of a legal
claim and evidence to be submitted would be out of the scope of
IACCMS, because the judiciary should not provide legal advice.
Besides, such AI tools, at the current stage of development, can
follow the letter of the law while disregarding its spirit, since they
can extract explicit, not implicit, information and they lack human
qualities such as empathy.
4
Conclusion
Legal systems can be improved by the introduction of AI, which
has the ability to bring change and benefits to society; it notably
has much to offer to individuals involved in court cases and the
justice system as a whole, though caution is needed for the impact
that AI could have on human rights [
16
]. Considering the
prospects and limitations of AI we explored the question of how
AI can facilitate the adjudication of cases, focusing on the specific
uses of AI for IACCMS regarding four types of users: judges, court
officers, lawyers and self-represented litigants. We understand
that new tools could be built to help judges and court officers with
the administration of justice: to facilitate the workflow of the
registry of courts and to provide useful information to judges
about the cases. We further conclude that AI tools could be
developed to offer assistance to litigants and their lawyers in
navigating legal processes, namely to help parties to gather
information prior to initiating legal proceedings before an
administrative court. However, there are limitations to the
introduction of AI in the justice system, since AI “should not
compromise the human and symbolic faces of justice. If justice is
perceived by the users as purely technical, without its real and
fundamental function, it risks being dehumanized. Justice is and
should remain humane as it primarily deals with people and their
disputes”. [
17
].