=Paper=
{{Paper
|id=Vol-3168/xpreface
|storemode=property
|title=None
|pdfUrl=https://ceur-ws.org/Vol-3168/xpreface.pdf
|volume=Vol-3168
}}
==None==
https://ceur-ws.org/Vol-3168/xpreface.pdf
XAILA 2021 - the Fourth Explainable & Responsible AI in Law (XAILA)
Workshop at ICAIL2021 - the 18th International Conference on Artificial
Intelligence and Law, São Paulo, Brazil
Preface
The 4th International Workshop on eXplainable and Responsible AI and
Law (XAILA2021@ICAIL) was held at the 18th International Conference on
Artificial Intelligence and Law (ICAIL 2021) organized by the Law School of
the University of São Paulo, Brazil (entirely online), on 21 June 2021. The
idea of the XAILA series of workshops (1st edition XAILA at JURIX 2018 in
Groningen, 2nd edition XAILA at JURIX 2019 in Madrid, 3rd edition XAILA at
JURIX 2020 in Brno (online)) is to provide an interdisciplinary platform for
the discussion of ideas with respect to explainable AI, algorithmic
transparency, comprehensiveness, interpretability and related topics.
This edition of XAILA was the first one to accompany the ICAIL conference
and one among the 11 workshops that were attached to ICAIL 2021. This
edition of the workshop attracted considerable attention from the
community with 853 registered participants (in comparison, the main
conference attracted 1240 registrations). We are convinced that these
numbers show that the problems relevant for the workshop are perceived
as not only as one of the most theoretically significant ones, but also as
one of the most pressing ones on the practical side. In particular, the
problem of adequate legal regulation of AI is acknowledged as one of the
most prominent issues, in European Union and elsewhere. For obvious
reasons, the development of such a regulatory framework is an extremely
complex issue, and it requires a deep understanding of how AI systems
work.
Consequently, the XAILA workshop, rooted in the AI and Law community,
attains even more interdisciplinary character. It aims to join the normative
perspective of legal theory and ethics on the one hand, and the formal
and computational approach represented by AI and Law research and
practice. What is more, the problems of explainability, transparency and
understandability of computational legal systems create a natural
platform to integrate the classical symbolic approach and the more recent
computational approach, the latter being represented first and foremost
by the machine learning models. One of the most important issues for the
AI and Law community is to explore the balance of advantages and
Copyright © 2022 for this paper by its authors. Use permitted under Creative Commons License
Attribution 4.0 International (CC BY 4.0).
disadvantages of the two approaches while developing systems
supporting the work of lawyers and enriching our understanding of legal
reasoning. However, the interests of the XAILA community go beyond the
legal applications of AI, while the problems of transparency, explainability
and responsibility, and the adequate regulation of these issues relevant
for any context intelligent systems operation. The embracing of this
broader involves, again, the creation of stronger links between different
actors. The communication between the practically oriented field of
engineering and the academic area of research in AI and Law is now
hindered by the diverging terminology used by these communities and
the identification of different purposes.
We may identify at least six fields which would potentially benefit from the
increase of the flow of information between them in connection with the
research and practice on explainable and responsible AI. One of them is
the AI and Law research with its tradition of interdisciplinary approach
towards the modeling of legal reasoning. Second, the perspective brought
by the general research on AI should be taken into consideration, and
properly accommodated to the purposes of the legal community. Third,
we should mention the quickly developing sector of Legal Tech, focused
on supplying legal practice with intelligent solutions aiming at the
optimization of legal tasks performance. Naturally, the broadly understood
field of legal practice, as the recipient of the said solutions, is the fourth
relevant actor. It should be stressed that this field is very diverse as it
encompasses not only the commercial sector of law entrepreneurship, but
also the public administration, the judiciary and the legislation. Each of
these subfields have their specific needs and purposes, which moreover
vary from jurisdiction to jurisdiction. Fifth, the academic field of legal
theory and philosophy may offer an important contribution to the ongoing
discussion and explainable AI and Law, as this area of research is
concerned with the elaboration of models of legal reasoning and the
discussion of the most general legal concepts, such as fairness, liability
and responsibility, important for the sake of development of the
regulatory frameworks. Sixth, in our opinion the discussion on explainable
AI and Law could benefit from the insights brought by cognitive science
research. The latter has already established connections with legal theory
which resulted in interesting research results concerning the mechanisms
of legal decision making. This research may be of particular relevance for
investigation of human-computer interaction in connection with the use of
intelligent support tools.
Copyright © 2022 for this paper by its authors. Use permitted under Creative Commons License
Attribution 4.0 International (CC BY 4.0).
The workshop program included two presentations by invited speakers
and five by authors presenting their research.
Our first invited speaker was Wojciech Wiewiórowski, who gave a lecture
entitled `Data protection aspects of Explainability of AI: Between
transparency and fairness'. He is acting as the EU's European Data
Protection Supervisor, and Adjunct professor in the Faculty of Law and
Administration of the University of Gdańsk. He provided the audience with
insightful, inspiring and up-to-date perspectives on the developing
regulatory perspective on AI in the European Union, and how it is
influencing the AI industry and research.
The second invited speaker was Katie Atkinson, with a lecture entitled
`The Landscape and Challenges for Explainability in AI and Law'. She is
Professor of Computer Science and Dean of the School of Electrical
Engineering, Electronics and Computer Science at the University of
Liverpool. She clearly and comprehensively showed how the field AI & Law
has been developing explainable AI methods all along, and why ever more
impact of research in AI & Law for legal practice and AI generally can be
expected.
The first paper presentation was by Trevor Bench-Capon (University of
Liverpool) on the paper entitled `Using Issues to Explain Legal Decisions'.
In the paper, he explains how traditional AI & Law approaches, and in
particular those based in case-based reasoning approaches, are relevant
for machine learning outcome prediction methods.
The second paper was presented by Salvatore Sapienza (University of
Bologna), presenting `Explanations in Risk Analysis: Responsibility, Trust
and the Precautionary Principle'. He emphasises the need for reliability
and trustworthiness in the high impact domain of health risk analysis,
using the regulation of food safety in Europe as a case study.
The third paper `Socially Responsible Virtual Assistant for Privacy
Protection: Implementing Trustworthy' was written by Alžběta Krausová
(Czech Academy of Sciences) and seven coauthors. In the paper, the
management of privacy settings is used as an example setting where a
virtual assistant can support a person's rights and self-determination.
Copyright © 2022 for this paper by its authors. Use permitted under Creative Commons License
Attribution 4.0 International (CC BY 4.0).
The fourth paper by Davide Carneiro (Politécnico do Porto) and four coauthors
is entitled `A Conversational Interface for interacting with Machine Learning
models'. In the paper, a conversational chatbot is used to approach the need
for better scrutiny and accountability of machine learning AI methods. An
analysis of legal and ethical considerations is used as a starting point.
The fifth presentation by Cor Steging (University of Groningen) and two
coauthors was on the paper `Discovering the Rationale of Decisions:
Experiments on Aligning Learning and Reasoning'. The paper presents a
knowledge-driven method for evaluating and adjusting the rationale used by
black box machine learning systems.
The workshop organizers would like to thank the Program Committee members
for their work in the review process. We are also grateful to the extensive
efforts of the ICAIL 2021 organization, especially hard in these times of hybrid
conference organization. We are happy to furthermore thank the invited
speakers, the authors of papers and all participants to the workshop.
Thanks to all of you, the meeting was stimulating and thought-provoking, and
we hope to meet many of you at a future edition of the workshop series.
Program committee
Michał Araszkiewicz, Jagiellonian University in Kraków, Poland
Martin Atzmueller, Osnabrück University, Germany
Floris Bex, Utrecht University, The Netherlands
Jörg Cassens, University of Hildesheim, Germany
Enrico Francesconi, IGSG-CNR, Italy
Grzegorz J. Nalepa, Jagiellonian University in Kraków, Poland
Jose Palma, University of Murcia, Spain
Monica Palmirani, CIRSFID, Italy
Juan Pavón, Universidad Complutense de Madrid, Spain
Radim Polcák, Masaryk University, Czechia
Marie Postma, Tilburg University, The Netherlands
Víctor Rodríguez Doncel, Universidad Politécnica de Madrid, Spain
Ken Satoh, National Institute of Informatics and Sokendai, Japan
Jaromir Savelka, Carnegie Mellon University, USA
Piotr Skrzypczynski, Poznan University of Technology, Poland
Bart Verheij, University of Groningen, The Netherlands
XAILA 2021 Organizing Committee
Grzegorz J. Nalepa, Jagiellonian University in Kraków, Poland
Bart Verheij, University of Groningen, the Netherlands
Michał Araszkiewicz, Jagiellonian University in Kraków, Poland
Martin Atzmueller, Osnabrück University, Germany
Copyright © 2022 for this paper by its authors. Use permitted under Creative Commons License
Attribution 4.0 International (CC BY 4.0).