Artificial Intelligence Computational Modelling

and Criminal Proceedings A Framework for A

European Legal Discussion Serena Quattrocolo pdf
download

https://textbookfull.com/product/artificial-intelligence-computational-modelling-and-criminal-
proceedings-a-framework-for-a-european-legal-discussion-serena-quattrocolo/

★★★★★ 4.8/5.0 (37 reviews) ✓ 134 downloads ■ TOP RATED

"Fantastic PDF quality, very satisfied with download!" - Emma W.

DOWNLOAD EBOOK
 Artificial Intelligence Computational Modelling and Criminal
Proceedings A Framework for A European Legal Discussion
Serena Quattrocolo pdf download

TEXTBOOK EBOOK TEXTBOOK FULL

Available Formats

■ PDF eBook Study Guide TextBook

EXCLUSIVE 2025 EDUCATIONAL COLLECTION - LIMITED TIME

INSTANT DOWNLOAD VIEW LIBRARY

Collection Highlights

Artificial Intelligence Accelerates Human Learning

Discussion Data Analytics Katashi Nagao

Artificial Intelligence Foundations of Computational

Agents David L. Poole

Artificial Intelligence for Computational Modeling of the

Heart 1st Edition Tommaso Mansi (Editor)

A Phenomenological Mathematical Modelling Framework for

the Degradation of Bioresorbable Composites Ismael Moreno-
Gomez
Legal Personhood Animals Artificial Intelligence and the
Unborn 1st Edition Visa A.J. Kurki

Our Humanity Exposed Predictive Modelling In A Legal

Context Stanley Greenstein

Computational Modelling of Concrete Structures:

Proceedings of the Conference on Computational Modelling
of Concrete and Concrete Structures 1st Edition Günther
Meschke

Artificial Intelligence: A Modern Approach 4th Edition

Stuart Russell

Clinical Neurotechnology meets Artificial Intelligence

Philosophical Ethical Legal and Social Implications 1st
Edition Orsolya Friedrich
Legal Studies in International,
European and Comparative Criminal Law 4

Serena Quattrocolo

Artificial
Intelligence,
Computational
Modelling and
Criminal Proceedings
A Framework for A European Legal
Discussion
Legal Studies in International, European
and Comparative Criminal Law

Volume 4

Editor-in-Chief
Stefano Ruggeri
Department of Law, University of Messina, Messina, Italy

Editorial Board Members

Chiara Amalfitano
University of Milan, Milan, Italy
Lorena Bachmaier Winter
Faculty of Law, Complutense University of Madrid, Madrid, Spain
Martin Böse
Faculty of Law, University of Bonn, Bonn, Germany
Eduardo Demetrio Crespo
University of Castile-La Mancha, Toledo, Spain
Giuseppe Di Chiara
Law School, University of Palermo, Palermo, Italy
Alberto Di Martino
Sant'Anna School of Advanced Studies, Pisa, Italy
Sabine Gleß
University of Basel, Basel, Switzerland
Krisztina Karsai
Department of Criminal Law, University of Szeged, Szeged, Hungary
Vincenzo Militello
Dipto Sci Giuridiche, della Società, University of Palermo, Palermo, Italy
Oreste Pollicino
Comparative Public Law, Bocconi University, Milan, Italy
Serena Quattrocolo
Department of Law, University of Piemonte Orientale, Alessandria, Italy
Tommaso Rafaraci
Department of Law, University of Catania, Catania, Italy
Arndt Sinn
Faculty of Law, University of Osnabrück, Osnabrück, Germany
Francesco Viganò
Bocconi University, Milan, Italy
Richard Vogler
Sussex Law School, University of Sussex, Brighton, UK
The main purpose of this book series is to provide sound analyses of major
developments in national, EU and international law and case law, as well as insights
into court practice and legislative proposals in the areas concerned. The analyses
address a broad readership, such as lawyers and practitioners, while also providing
guidance for courts. In terms of scope, the series encompasses four main areas, the
first of which concerns international criminal law and especially international case
law in relevant criminal law subjects. The second addresses international human
rights law with a particular focus on the impact of international jurisprudences on
national criminal law and criminal justice systems, as well as their interrelations. In
turn the third area focuses on European criminal law and case law. Here, particular
weight will be attached to studies on European criminal law conducted from a
comparative perspective. The fourth and final area presents surveys of comparative
criminal law inside and outside Europe. By combining these various aspects, the
series especially highlights research aimed at proposing new legal solutions, while
focusing on the new challenges of a European area based on high standards of
human rights protection.
As a rule, book proposals are subject to peer review, which is carried out by two
members of the editorial board in anonymous form.

More information about this series at http://www.springer.com/series/15393

Serena Quattrocolo

Artificial Intelligence,
Computational Modelling
and Criminal Proceedings
A Framework for A European Legal
Discussion
Serena Quattrocolo
Department of Law, Politics, Economics and Social Science
University of Eastern Piedmont
Alessandria, Italy

ISSN 2524-8049     ISSN 2524-8057 (electronic)

Legal Studies in International, European and Comparative Criminal Law
ISBN 978-3-030-52469-2    ISBN 978-3-030-52470-8 (eBook)
https://doi.org/10.1007/978-3-030-52470-8

© The Editor(s) (if applicable) and The Author(s), under exclusive licence to Springer Nature
Switzerland AG 2020
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the
editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

Cover illustration: © Maria Isabel Ruggeri

This Springer imprint is published by the registered company Springer Nature Switzerland AG.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
To the beloved memory of Daiana, the most
curious, lively and unlucky of my students:
may your intellectual enthusiasm live in
those who read this book.
Foreword: Let a Lawyer Write of AI

There are three different ways in which we can appreciate the role of Artificial
Intelligence (AI) and of further emerging technologies in the field of criminal law.
AI may entail either new forms of criminal responsibility, or new loopholes in the
criminal law field, or new challenges for the rights of criminal defendants to a fair
trial. The sequence of this list follows a chronological order, because scholars, first,
debated since the early 1990s whether AI could ever be considered as an account-
able agent in the criminal law field, due to its mens rea, i.e. the mental element of an
offence. A decade later, in the mid-2000s, scholars increasingly focused on the
material elements of a crime, in order to determine to what extent AI could ever
trigger a new generation of actus reus. Finally, in the mid-2010s, the attention was
progressively shifted to AI either as a collector of evidence in criminal investiga-
tions, or as a human substitute in the judicial decision-making process, or as a proxy
of the principle of fair trial in criminal proceedings, thereby affecting the discretion
of courts.
The first kind of debate on the criminal personality of AI started with the father
of “robotics”, Isaac Asimov in 1941. In the legal domain, this kind of debate became
particularly popular fifty years later, when some brilliant scholars, such as Justice
Curtis Karnow and Professor Lawrence Solum, discussed new forms of account-
ability and personhood for “distributed artificial intelligence”, as in Karnow’s 1996
paper, or for artificial intelligences, i.e. “AIs”, as in Solum’s seminal work from
1992. Two decades later, the advancements of technology turned this academical
discussion into a hot political and ideological debate. A crucial role in this shift was
played by the European Parliament’s proposal from February 2017, in which the EU
institution invited the European Commission “to explore, analyze and consider the
implications of all possible legal solutions, (including) … creating a specific legal
status for robots in the long run”. Some reckon that AI can actually be considered
“aware” and fulfilling the mental requirements of that which criminal lawyers dis-
cuss as intentional and negligent offences (e.g. Gabriel Hallevy’s thesis). Others
claim that cognition and awareness, volition and intention or reason responsiveness
can be attributed to an AI system in a way that is meaningful for criminal lawyers,
although no reference is necessary to human-like properties (e.g. Giovanni Sartor’s

vii
viii Foreword: Let a Lawyer Write of AI

stance). Most scholars note, however, that nobody would bring AI before judges
today, in order to declare AI “guilty” in criminal courts. This is not to say that fur-
ther forms of legal agenthood for AI make no sense, for example, in such fields as
business and corporate law. Yet, it seems fair to admit, according to the phrasing of
the European Parliament, that the criminal personality of AI, if ever, could only
develop “in the long run”.
The second kind of debate on a new generation of AI actus reus started with the
second Gulf War in Iraq (and in Pakistan), in the mid-2000s. Years later, in the 2010
Report to the UN General Assembly, the Special Rapporteur on extrajudicial, sum-
mary or arbitrary executions, Christof Heyns, urged the then Secretary-General Ban
Ki-moon to convene a group of experts in order to address “the fundamental ques-
tion of whether lethal force should ever be permitted to be fully automated”. A hot
debate revolved around what AI can justly do in war (ius in bello) and when and
how resort to war via AI can be justified (ius ad bellum or bellum iustum). In addi-
tion, scholars examined a wider impact of AI on a tenet of criminal law, which is
traditionally summed up with the idea that “everything which is not prohibited is
allowed”. Similar to what occurred with a new generation of computer crimes in the
early 1990s, scholars have progressively stressed that AI may soon trigger a new
generation of actus reus, therefore affecting the principle of legality as enshrined in,
for example, Article 7 of the 1950 European Convention on Human Rights. A recent
literature review has proposed five areas of foreseeable threats of AI in criminal law,
such as (i) new offences against the persons, e.g. harassment; (ii) sexual offences;
(iii) trafficking, selling, buying and possessing harmful or dangerous banned drugs;
(iv) commercial insolvency and further issues of financial markets, such as price-­
fixing and collusions and (v) theft and non-corporate fraud. As I like to say, these
new scenarios for AI crimes are only limited by the human imagination. In the field
of computer crimes, illegal systems started amending their own laws from the early
1990s, e.g. the Italian regulation 547 from December 1993; then, the international
legislator intervened, so as to formalise such legal experiences through the general
provisions of the Budapest convention from 2001. I think something similar will
happen with a new generation of AI crimes.
The third kind of debate on how the rights of criminal defendants to a fair trial
can be affected by AI has increasingly drawn the attention of experts in the 2010s to
the role that AI may play as an evidence collector, as a human substitute or as a
proxy of the fair trial principle in criminal proceedings. The US Supreme Court’s
case law is particularly instructive. In Jones v. United States (565 U.S. 400 (2012))
and then in Carpenter v. United States (585 U.S. _ (2018)), the rulings concerned
the protection of the “reasonable expectation of privacy” vis-à-vis the collection of
criminal evidence through the use of GPS and cell phone locations techniques,
respectively. In 2017, criminal proceedings’ safeguards in an age of AI ignited a
popular debate in journals and newspapers because of the use of risk assessment
programs by the Court of the Loomis v. Wisconsin case (881 N.W. 2d. 749 (Wis.
2016)). As the New York Times was keen to inform us on 1st March 2017, Loomis’
claim was that “his right to due process was violated by a judge’s consideration of a
report generated by the software’s secret algorithm, one that Mr. Loomis was unable
Foreword: Let a Lawyer Write of AI ix

to inspect or challenge”. In June 2017, the Supreme Court denied Loomis’ petition
for a writ of certiorari. All in all, against such case law, we can suspect that this kind
of debate on how AI may impinge on the rights of criminal defendants will become
more and more urgent. Three reasons suggest this conjecture.
The first reason has to do with the rights of defendants to examine the algorithms
of AI: some features of AI, such as the inscrutability of machine learning tech-
niques, add a layer of complexity to traditional digital forensics. The second reason
regards the role AI plays in the decision-making of judges: it is still unclear how
courts should exercise their discretion when striking the balance between fair trial
or due process arguments of criminal defendants and the value of AI risk assess-
ments. The third reason hinges on the interplay between fair trial principles and the
protection of personal data and individual privacy rights. In Europe, for example,
the European Court of Human Rights has so far subordinated some legal safeguards
of Article 6 of the Convention on fair criminal trials, to a preliminary violation of
Article 8 on the right to privacy: AI will likely exacerbate the weaknesses of this
stance on the interplay between criminal safeguards and privacy rights, since the
right to a fair trial can be strengthened—but not replaced—by the protection of the
right to privacy (and data protection). In the USA, a new generation of AI collectors
of evidence will similarly stress some shortcomings of the Supreme Court’s doc-
trine on the third-party doctrine, namely the idea that secrecy is a prerequisite for
the protection of privacy rights under the Fourth Amendment to the Constitution.
Further, the right to a reasonable expectation of privacy, both societal and individ-
ual, may end up with “the chicken or the egg” causality dilemma. Such reasonable
privacy expectations rest on the assumption that individuals and society have devel-
oped a stable set of privacy expectations, and still, AI can dramatically change such
beliefs.
Therefore, its vital scholars properly address the urgent issues of AI forensics,
judiciary discretion and data protection in an age of increasingly smart AI. By com-
parison with the previous debates on the criminal personality of AI or on a new
generation of AI crimes, it should be noted, however, that there are still few works
on this subject matter. A reason may depend on recent developments of law and
technology; another on the complexity of the issues that are at stake with AI. The
direct and indirect impact of computational modelling on evidence gathering, much
like the challenges of AI to the judicial decision-making process in criminal pro-
ceedings, do not only concern legal expertise, but scientific knowledge and techno-
logical know-how. In particular, the focus should be on criminal investigations that
hinge on the use of computational techniques and AI systems, in order to understand
how they may affect the principles of fair trial and the equality of arms through
automatedly generated evidence. Likewise, the attention should be drawn to the
distinction between deciding and predicting, between criminal justice and predic-
tive justice. AI impacts the different steps of the criminal justice decision-making
process when tackling, for example, violent behaviour and recidivism.
Few investigators, however, could have ever set the proper theoretical frame-
work, in which to address these complex sets of issues on philosophy of law and
criminal justice, legal informatics and digital forensics, machine learning and deep
x Foreword: Let a Lawyer Write of AI

learning. Professor Quattrocolo’s book fills this gap. Her book is timely and badly
needed, because it provides an in-depth analysis on some of the most urgent legal
threats brought forth by AI today. Moreover, the book is solid and yet provocative,
because it will generate a multidisciplinary discussion on how the law could be used
to safeguard the rights of all parties involved as AI proliferates more deeply into
society. Lawyers can learn a lot about their own field, working together with com-
puter scientists and AI developers, much as AI developers and computer scientists
can reflect on the normative constraints of their work, collaborating with law profes-
sors, judges, attorneys and other legal experts. This monograph is the fruit of this
essential interaction. Let the author of this book, a lawyer, write important things
about AI.

Law Department Ugo Pagallo

University of Torino
Torino, Italy
Acknowledgments

I owe gratitude to many colleagues and friends, for having inspired the thoughts I
collected in this book and for having read the early versions of it. In particular, I
thank Ugo Pagallo, for his intellectual generosity and for the time he devoted to
helping me in this endeavour. I am grateful to my dear friend Georgia Zara, who
guided me into the realm of risk assessment, Massimo Durante, Luigi Portinale,
Cosimo Anglano, Margherita Benzi for their valuable suggestions. Thanks to Mina
Elton, for her patience in reviewing my English, and Professors Julie O'Sullivan and
Carlos Vasquez for helping with American decisions and figures. I thank Stefano
Ruggeri and all the colleagues on the board of this collection, for having encouraged
the publication of my work, and, of course, I thank Paolo, who always tolerates my
priorities.
The research for this book has been supported by the research funding program
of my home institution, the University of Eastern Piedmont, Italy, and was mostly
conducted at the IALS Library, London and the Max Planck Institute for the Study
of Crime, Security and Law, Freiburg i.B, which I thank.

xi
Contents

Part I Introducing the Problem

1 Approaching the Unknown: Some Preliminary Words������������������������    3
1.1 Criminal Law at the Digital Turn������������������������������������������������������    3
1.2 More Than Cybercrime��������������������������������������������������������������������    5
1.3 A Functional Definition of AI and Other Basic Concepts����������������    7
1.4 The Structure of the Book����������������������������������������������������������������    9
References��������������������������������������������������������������������������������������������������   12
2 A Theoretical Framework for the Discussion on AI and Criminal
Law������������������������������������������������������������������������������������������������������������   13
2.1 Connecting AI, Computational Modelling and Fundamental
Rights in Criminal Justice����������������������������������������������������������������   13
2.1.1 Algorithms and the Complexity of Judicial Decision
Making����������������������������������������������������������������������������������   15
2.1.2 Accessibility and Transparency��������������������������������������������   17
2.2 The Boundaries of the Study: A Premise������������������������������������������   21
2.2.1 Specificity of the Criminal Law Context������������������������������   22
2.2.2 Specificity of the European Context ������������������������������������   27
2.2.3 Specificity of the Continental Context����������������������������������   30
References��������������������������������������������������������������������������������������������������   32

Part II Direct and Indirect Impact of Widespread Computational

Modelling on Evidence Gathering
3 Hacking by Law-Enforcement: Investigating with the Help of
Computational Models and AI Methods������������������������������������������������   37
3.1 Investigating vs. Policing: A Necessary Foreword ��������������������������   37
3.1.1 Predictive Policing����������������������������������������������������������������   39
3.1.2 The Realm of Investigation��������������������������������������������������   40
3.1.3 The Dawn of Digital Investigations��������������������������������������   42
3.2 Privacy v. Investigation: A Naturally Unbalanced Relationship ������   44

xiii
xiv Contents

3.2.1 ‘Private Life’ in Art. 8 ECHR: Home, Correspondence

and… Personal Data?������������������������������������������������������������   45
3.2.1.1 Interference by Public Authority ��������������������������   46
3.2.1.2 “Home” in the ECtHR Case-Law��������������������������   47
3.2.1.3 And “Correspondence”������������������������������������������   48
3.2.1.4 Privacy and Data Protection����������������������������������   48
3.2.2 Art. 7 and 8 of the ChFREU: A Brand-New Approach?������   52
3.2.2.1 The Scope of Art. 7 ChFREU��������������������������������   53
3.2.2.2 Different Perspectives��������������������������������������������   54
3.2.2.3 Secondary EU Legislation on Data Protection
in the Criminal Area����������������������������������������������   56
3.3 Is AI Transforming the Traditional Sites of Our Private Life
into a Panopticon for Criminal Investigations?��������������������������������   58
3.3.1 What Is ‘Communication’ Today?����������������������������������������   59
3.3.2 What Is ‘Home’ in the Infosphere?��������������������������������������   60
3.4 Is Investigative Hacking Overarching Privacy?��������������������������������   62
3.4.1 Some Examples of Hacking by Law Enforcement
Instruments����������������������������������������������������������������������������   63
3.4.2 Is Hacking Inherently Incompatible with Privacy?��������������   68
References��������������������������������������������������������������������������������������������������   69
4 Equality of Arms and Automatedly Generated Evidence��������������������   73
4.1 Digital Evidence and Automatedly Generated Evidence������������������   73
4.2 Some General Reflections About Evidence in the European
Context����������������������������������������������������������������������������������������������   74
4.2.1 The Statute of Evidence in Art. 6 ECHR������������������������������   77
4.2.2 The Fruits of the Poisonous Tree������������������������������������������   77
4.2.2.1 Violations of Art. 3 ECHR������������������������������������   78
4.2.2.2 Violations of Art. 8 ECHR������������������������������������   79
4.2.2.3 Violations of Art. 6 ECHR������������������������������������   81
4.2.2.4 Infringements of the Right to Confrontation ��������   82
4.2.2.5 Infringements of the Right of Defence������������������   84
4.2.3 The Concept of the Overall Fairness of the Proceedings������   86
4.3 Automatedly Generated Evidence and Fair Trail������������������������������   89
4.3.1 The Essence of Fair Trial: The Equality of Arms ����������������   90
4.3.1.1 Transparency as a Beacon of Equality?����������������   91
4.3.1.2 What Does Transparency Mean in
This Context?��������������������������������������������������������   93
References��������������������������������������������������������������������������������������������������   96

Part III Challenges of Computational Methods to the Judicial

Decision-­Making Process: Deciding v. Predicting
5 Predictability and Criminal Justice�������������������������������������������������������� 101
5.1 From Evidence to Decision�������������������������������������������������������������� 101
Contents xv

5.2 Judicial Decision-Making and Logique Juridique: Two Sides

of the Same Coin������������������������������������������������������������������������������ 102
5.2.1 An American Overview�������������������������������������������������������� 105
5.2.2 A European Point of View���������������������������������������������������� 107
5.3 Mathematical Modelling of Judicial Behaviours?���������������������������� 108
5.3.1 A Short Overview on AI and Law���������������������������������������� 110
5.3.2 The Dawn of Case-Based Reasoning������������������������������������ 111
5.4 What Prediction Has to Do with Criminal Proceedings?������������������ 114
5.4.1 Quantitative Legal Prediction: What Can It Do for
Criminal Justice?������������������������������������������������������������������ 117
5.4.2 What Does ‘Predicting a Decision’ Mean?
The Relevance of Statistics in Criminal Matters������������������ 119
5.4.3 How ‘Prediction’ Would Impact on the Parties
and the Judge?���������������������������������������������������������������������� 121
5.4.4 How Prediction Would Impact on the Quality of Justice?������ 122
5.5 Brief and Crucial Matters of Semantics and Taxonomy ������������������ 125
References�������������������������������������������������������������������������������������������������� 127
6 Predictability of Violent Behaviour and Recidivism���������������������������� 131
6.1 Is ‘Foreseeing’ the Future Relevant in Adjudicating
a Criminal Case? ������������������������������������������������������������������������������ 131
6.1.1 Areas of Criminal Proceedings Requiring ‘Prediction’�������� 132
6.1.2 Pre-trial Detention and Release�������������������������������������������� 135
6.1.3 Sentencing���������������������������������������������������������������������������� 137
6.1.3.1 The European Approach���������������������������������������� 140
6.1.3.2 The American Approach���������������������������������������� 144
6.2 Risk Assessments, Predictive Sentencing and the ‘Daubert Test’:
Complicated Interactions Behind the Recent Digitalisation������������ 146
6.2.1 Risk Assessment and the Development of Psycho-
Criminological Theories ������������������������������������������������������ 147
6.2.1.1 Pre-trail Risk Assessment�������������������������������������� 151
6.2.2 Evaluation of Risk Assessment Tools ���������������������������������� 153
6.2.3 Rating Pre-trial Risk Assessment������������������������������������������ 155
6.3 A Famous Case of ‘Predictive Sentencing’ from Wisconsin…�������� 156
6.3.1 The Facts of the Case������������������������������������������������������������ 157
6.3.2 The Defence’s Arguments���������������������������������������������������� 158
6.3.3 The Ruling by the State Supreme Court ������������������������������ 158
6.4 A Similar Case from D.C.���������������������������������������������������������������� 161
6.4.1 The Background�������������������������������������������������������������������� 161
6.4.2 The Defence’s Arguments���������������������������������������������������� 164
6.5 Is Automated Risk Assessment Different?���������������������������������������� 166
6.5.1 The Loomis Case as a Cornerstone?������������������������������������ 166
6.5.2 Comparing Different Perspectives���������������������������������������� 169
xvi Contents

6.6 From the European Point of View…������������������������������������������������ 171

6.6.1 The Overall Impact of Bad Character on Trial �������������������� 173
6.6.2 Use of Risk Assessment in Remand in Custody
Decisions������������������������������������������������������������������������������ 175
6.7 Tentative Conclusions ���������������������������������������������������������������������� 176
References�������������������������������������������������������������������������������������������������� 177
7 Predictability and the Criminal Justice Decision-Making Process ���� 181
7.1 Case-Based Reasoning Systems, Open-Data and the Value
of the Precedent: Legal Traditions Clashing? ���������������������������������� 181
7.1.1 The Open Data Movement and the EU Approach���������������� 182
7.1.2 Open Data and Criminal Justice ������������������������������������������ 184
7.1.3 Public Access to Court Record �������������������������������������������� 186
7.2 Accessibility of Criminal Justice: The Values of the Publicity
of Hearings and Publication of Decisions���������������������������������������� 188
7.2.1 The Principle of Stare decisis and the Publication of
Criminal Decisions��������������������������������������������������������������� 191
7.2.1.1 Stare decisis in the Common Law Tradition �������� 191
7.2.1.2 The Precedent in Civil Law ���������������������������������� 194
7.2.2 The Value of the Precedent from the Comparative Angle������ 199
7.3 Computational Modelling and Predictability of Decisions:
An Enquiry in Three Steps���������������������������������������������������������������� 200
7.3.1 Step One�������������������������������������������������������������������������������� 201
7.3.1.1 Establishing Relevant Variables���������������������������� 201
7.3.1.2 Establishing Relevant Patterns in a Single
Judge’s Decisions�������������������������������������������������� 205
7.3.1.3 The Role of the Courts and the Binding
Value of Their Decisions �������������������������������������� 206
7.3.2 Step Two ������������������������������������������������������������������������������ 208
7.3.3 Step Three ���������������������������������������������������������������������������� 211
7.4 Prediction and Predictability������������������������������������������������������������ 213
7.4.1 Comprehension of the Legal Command and Culpability ���� 215
7.4.2 Useful Semantic Distinctions������������������������������������������������ 219
References�������������������������������������������������������������������������������������������������� 221
8 The Gist of the Inquiry �������������������������������������������������������������������������� 225
8.1 A Personal Result������������������������������������������������������������������������������ 225
8.2 Competing Values, Prevailing Guarantees���������������������������������������� 226
8.3 Digital Solutions for Legal Conundrums������������������������������������������ 227
8.4 Adjudicating and Predicting Are Different Activities���������������������� 228
Abbreviations

AI Artificial Intelligence
CBR Case-Based Reasoning
CEPEJ Council of Europe European Commission for the Efficiency of Justice
ChFREU Charter of Fundamental Rights of the European Union
CJEU Court of Justice of the European Union
CoE Council of Europe
COMPAS Correctional Offenders Management Profiling for Alternative
Sanctions
DC District of Columbia (USA)
ECHR European Convention on Human Rights
ECRIS European Criminal Records Information System
ECtHR European Court of Human Rights
FRE (US) Fedaral Rules of Evidence
IAAIL International Association for Artificial Intelligence and Law
ICAIL International Conference on Artificial Intelligence and Law
ICCPR International Covenant on Civil and Political Rights
ICT Information and Communication Technologies
IoT Internet of Things
IT Information Technologies
ItCCP Italian Code of Criminal Procedure
JHA Justice and Home Affairs
LEAs Law Enforcement Agencies
ML Machine Learning
ODM Open Data Movement
PSA Public Safety Assessment
RBR Rule-Based Reasoning
SAVRY Structured Assessment of Violence Risk in Youth

xvii
 Part I
Introducing the Problem
Chapter 1
Approaching the Unknown:
Some Preliminary Words

1.1 Criminal Law at the Digital Turn

Traditionally, Law and Technology has been considered an oxymoron. Connecting

two incompatible concepts, Criminal Law and Technology—not only Information
Technology—seems to be even more inconceivable. Criminal Law is a jurisdic-
tion’s ultimate reaction to an assault upon the core values of its society and is
embedded in its social culture. This topic is abundant in the European continental
literature, from which we derive the idea that Criminal Law grasps national
Kulturnomen, reproducing the general—or, at least, the most common—values of a
population.1 Thus, because cultural shifts are slow-evolving phenomena, Criminal
Law throughout the world also tends to be a slow-changing factor: only established
transformations can be ratified in law, not just in statutory-law legal systems.
Although law is a means to influence people’s behaviour,2 in democratic societ-
ies Criminal Law seems to be inadequate to drive normative changes in social
behaviour. It instead cements accomplished processes into sets of commands,
reflecting an accepted framework of social values. Indeed, this is perfectly under-
standable: the severity of penalties implies that the rejection of a specific conduct is
shared by a vast majority of the community, just as the abolition of an offence
(either statutory or judge-made) implies a general recognition of legitimacy in such
conduct. If not, the legislator is imposing, non-democratically, values and rules that
do not reflect common opinions and feelings. According to the European legal
scholarship, this may indirectly impinge on one of the modern understandings of the
rule of law. Indeed, the latter implies individuals’ full understanding of the criminal

1
See Mayer (1903), in particular p. 19, explaining the correspondence between legal norms and
cultural norms; Cadoppi (2014), p. 22.
2
See Julia Black’s perspective on decentred regulation, in which law is one out of many different
systems to influence social behaviour: Black (2002), p. 4.

© The Editor(s) (if applicable) and The Author(s), under exclusive licence to 3
Springer Nature Switzerland AG 2020
S. Quattrocolo, Artificial Intelligence, Computational Modelling and Criminal
Proceedings, Legal Studies in International, European and Comparative
Criminal Law 4, https://doi.org/10.1007/978-3-030-52470-8_1
4 1 Approaching the Unknown: Some Preliminary Words

behaviour and command, as the basis for the duty to respect the law. One could not
be expected to avoid criminal behaviours if it is not possible to fully and properly
grasp what the law considers to be criminal.
However, it is undisputed that, over the past few decades, contemporary society
has witnessed a computational turn, that, now we all recognise,3 is not only a breath-­
taking scientific advancement, a radical change in every professional sphere, but,
overall, is one of the most rapid, astonishing and wide-spread cultural changes that
has ever occurred.4 It has affected the foundations of our society5 in such a way to
permeate even the steady core of criminal law.
The first notable impact of the computational turn on the realm of criminal law
has been in substantive criminal law.6 In fact, for at least two decades, (also) the
European rulers and legislators started turning their attention to new forms of
offences perpetrated in the dematerialised world of IT, trying to set common mini-
mum rules for the harmonisation of cybercrime and the consequential issues, such
as competence and jurisdiction, also at the international level.7 The Budapest
Convention on Cybercrime, signed in 2001,8 having been drafted by the Council of
Europe, represented, for a huge number of jurisdictions, the first legally binding
instruction to set forth specific legislation tackling such phenomenon. In fact, the
first part of the document harmonised the Member States’ substantive criminal law,
by providing for a list of offences that the High Contracting Parties have a duty to
introduce in their jurisdictions.9 It is a common perception that harmonisation of
substantive criminal law is the area in which States demonstrate major reluctance to
limiting their sovereignty through international agreements: however, the Budapest
Convention achieved an important score of signatures and ratifications, by almost
all the members of the Council of Europe, and a long list of non-member States,
including, among others, the United States of America, Australia, Israel, Japan.
Incidentally, it is worth noting (more on this in Chap. 3) that the very same
Convention also covers important procedural aspects, regulating the collection of
evidence (e.g. such as computer search and seizure, real time collection of data traf-
fic, etc.), jurisdiction, mutual legal assistance in cybercrime cases. However,

3
Even though someone had clearly foreseen it, decades ago: Negroponte (1995); Kurzweil (2005),
pp. 7, 8.
4
Pagallo (2018), p. 1 ff.
5
For a comprehensive overview, see Durante (2019); Garapon and Lassègue (2018), especially
p. 83 ff.: “la revolution numérique bouleverse tous les compartiments de l’existence collective”.
6
Information and Communication Technologies have been seen and presented, at least initially, as
a means for facilitating anti-social criminal activities: see Thomas and Loader (2000), p. 1.
7
See the Council of Europe Budapest Convention, 23.11.2001, at https://www.coe.int/en/web/con-
ventions/full-list/-/conventions/treaty/185/signatures.
8
Although the Council of Europe started drawing attention on this topic in the Eighties, with the
Recommendations of the Committee of Ministers R(89)9, on computer-related crimes and R(95)13
concerning problems of criminal procedural law connected with information technology. In 1996
a committee was established to draft the text of a convention about cybercrime, that than was
signed in Budapest, in 2001.
9
For an overview of the impact of the Convention see Koops and Brenner (2006).
1.2 More Than Cybercrime 5

cybercrime is not, in any way, the object of this study, and only a few references to
the area of application of the Budapest Convention are covered in the book.

1.2 More Than Cybercrime

These preliminary and ancillary remarks suggest two very general observations.
Firstly, if criminal law is a representation of the existing cultural context in a juris-
diction, it is not possible for it to move ahead, or keep up with scientific progress,
that is reshaping social habits. Scientific progress will always precede changes and
amendments in criminal law.
Secondly, the impact of the computational turn upon the realm of criminal justice
turned out to be much wider than the area of ‘traditional’ cybercrime. Although the
concept of cybercrime acknowledged at the beginning of this Century, as “computer-­
mediated activities which are either illegal or considered illicit by certain parties
and which can be conducted through global electronic networks”10 may not be con-
sidered wrong in itself, the phenomenon proved much more intricate and sophisti-
cated over the decades. Today, legal research in that branch of criminal law has
evolved into an attempt to theorise the application of the classic legal categories to
artificial intelligence entities.11 Moreover, the digital revolution that globally
occurred especially over the last decade is having repercussions upon every aspect
of the administration of criminal justice, far beyond the topics that have been
addressed by the Budapest Convention on Cybercrime. Our transformation into a
digital society is determining substantial changes not only in the context in which
crime may occur, or in the way investigations can be carried out. Delivering justice
is a human task and the sudden digital change in individuals’ lifestyles is affecting
the way in which such a task is performed,12 impinging on the internal aspects of it,
such as the decision-making process. These aspects will be analysed in the second
part of the book.
What has been briefly observed here can be considered the cause of an undis-
puted trend. If regulation in the criminal area tends to follow (not to precede) social
changes, the computational turn13 occurred out of (and before) a specific legal

10
See, among the first attempts to define the concept, Sieber (1977). Thomas and Loader (2000),
p. 3. Although both research and legislation evolved significantly, it is still impossible to give one,
unique and undisputed definition of cybercrime. “Cybercrime is a container term of convenience,
describing a collection of acts or a field of criminal activity, rather than a single concept”: Boister
(2018), p. 188. Moreover, many other similar terms are often used, such as ‘computer crime’,’ IT
crime’… It has been said that the concept encompasses a whole range of terms which imply that
the digital technology (not only computers) is an element of the offence. In this sense, internet
connections are necessary elements, either for crimes against digital technologies and crime com-
mitted by means of digital technologies: Sieber (2008), p. 127.
11
Pagallo and Quattrocolo (2018), p. 400.
12
See the seminal work of Susskind (2008).
13
de Vries (2013), all Ch. 1.
6 1 Approaching the Unknown: Some Preliminary Words

framework. On the one hand, this implies that digital advancement has been taking
place in the absence of any background research on the risks it may pose to the core
values of society, usually protected by criminal law. On the other hand, for a long
time, and not only in Europe, the development of digital solutions overlooked the
specific needs of criminal justice: existing (and pre-existing) technology leaked into
(almost every) criminal justice system,14 affording methods and solutions that were
tailored for different purposes and not expressly suited for judicial use.
In particular, the computational revolution led to the availability of an enormous
quantity of free data, constantly generated by digital devices; powerful computa-
tional resources, the ability to access immeasurable amounts of data in a few sec-
onds; cheaper storage costs.15 These conditions (access to big data and cheap,
unprecedented computational power) established the grounds for offering useful
solutions to criminal justice systems, even though not specifically tailored for the
task. The digital turn provided not only whole ranges of data, that could be used as
evidence in criminal proceedings, but also new investigation systems, based on
hacking, mining and analysing huge sets of available data (private or not; personal
or not). The turn not only brought the full digitalisation of collections of courts’
decisions, with unrestricted access to any case-law, but also relatively sophisticated
software for the analysis of it, to find patterns of predictability within judicial deci-
sions. Moreover, the availability of an unprecedented amount of digital data shifted
the attention from a code-based modelling system (code-driven regulation),16 totally
deterministic—in which the discretion is encrypted in the expert-designed code,
establishing that If This, Then That—to a deep-learning modelling system, non-­
deterministic, in which the discretion lays in the choice of the data set (data-driven
regulation) of legal texts to be used to train the system.17 I will elaborate more exten-
sively on this in Chap. 5.
All this happened without a proper and effective consultation between the com-
puter scientists leading the digital revolution, and criminal law experts.18 The litera-
ture confirms this position, indicating a growing focus, especially in the U.S., of
private law on the theoretical challenges inherent in the application of automation
and artificial intelligence to everyday life, since Lawrence Solum’s seminal article

14
Data mining and predictive analytics have origins in commerce: Pasquale (2015), p. 4: “At the
core of information economy are Internet and the finance companies that accumulate vast amounts
of digital data, and with it intimate details of their customers’ – our -lives”.
15
Katz (2013), p. 916.
16
For the distinction between code-driven and data-driven regulation, see Hildebrandt (2018),
pp. 2, 3.
17
Evans (2019), p. 260: “the modern debate between deep-learning practitioners and advocates of
logic-based approaches resembles the Eighteenth century debate between empiricists and
rationalists”.
18
It was 1967 when P. Souleau, a high rank French magistrate, dreamt of a joint venture between
cyberneticists and judiciary to create models reproducing the most basic and repetitive parts of the
judicial adjudication, in particular vehicles accidents (Souleau 1969, p. 70).
1.3 A Functional Definition of AI and Other Basic Concepts 7

in 1992,19 accompanied by a much slower trend in criminal law, and in criminal

procedure. To some extent, this concurs with our previous assertion that the inherent
characteristic of criminal law is to follow, rather than precede (or progress with)
social changes…20 For these and other reasons that will be analysed later, there are
several arguments for setting the area of criminal law aside from some of the most
innovative tools based on computational modelling and machine learning (Chaps. 6
and 7).21

1.3 A Functional Definition of AI and Other Basic Concepts

The title of this book refers, in first instance, to Artificial Intelligence. It is rather
clear, though, that the term AI holds, in this context, a non-specific definition. Based
on the extensive scholarship that has been published so far, the definition of AI has
become a major endeavour: there is not a standard definition of what exactly AI
involves.
Early 2020, the Joint Research Centre (JRC), the EU Commission’s Science and
Knowledge Service, delivered a comprehensive investigation on the various defini-
tions of AI, elaborated since the 1950s, proposing “an operational definition of AI
formed by a concise taxonomy and a set of keywords that characterise the core and
transversal domains of AI”.22 The huge work presented by the centre not only gath-
ered and classified the multifaceted definitions of AI available, according to the
various areas of scholarship at stake, but also led to an interesting proposal of defini-
tion. According to the exhaustive report, there would be at least four common fea-
tures, within the manifold range of AI definitions, including the fact that a system is
based on: consideration of the real world complexity; information processing (col-
lecting and elaborating inputs); decision making; achievement of specific goals. If
these are the basic pillars of AI, the recent EU HLEG’s study (High Level Experts
Group on artificial intelligence) tried to elaborate a more operational concept of
AI. Indeed, the latter23 is highly specific and detailed and, according to the JRC

19
Solum (1992).
20
This attitude has been recently studied under the sociological point of view: see Christin (2017),
p. 2 ff. Based on his empirical research in American local criminal courts, the author concludes that
“in criminal justice innovation does not come with the glitter and appeal that it has in other sectors:
it is often a source of uncertainty, as innovation arrives without the vetting of precedent”.
21
See, e.g., Breton-Darrien (2018), p. 19, suggesting that the stakeholders in the area of legal-tech
accept not to mine and reuse data from criminal decisions.
22
JRC Technical Reports, AI Watch. Defining Artificial Intelligence, Publications Office of the
European Union, Luxembourg, 2020, 11.
23
“Artificial intelligence (AI) systems are software (and possibly also hardware) systems designed
by humans(2) that, given a complex goal, act in the physical or digital dimension by perceiving
their environment through data acquisition, interpreting the collected structured or unstructured
data, reasoning on the knowledge, or processing the information, derived from this data and decid-
ing the best action(s) to take to achieve the given goal. AI systems can either use symbolic rules or
8 1 Approaching the Unknown: Some Preliminary Words

study, it may turn out to be too specified to apply in the non-core science realm. This
assumption is also valid in our case. Although the object of the present study fits
into the HLEG’s definition, it is probably more suitable to introduce the topic by
referring to the definition in the EC JRC Flagship report on AI,24 establishing that
“AI is a generic term that refers to any machine or algorithm that is capable of
observing its environment, learning, and based on the knowledge and experience
gained, taking intelligent action or proposing decisions. There are many different
technologies that fall under this broad AI definition. At the moment, ML4 [machine
learning] techniques are the most widely used”.
In this sense, the term AI is the framework of this study, along with other con-
cepts that need to be specified here.
‘Computational modelling’ is the second term used in the title. It refers to the
simulation and study of complex systems, using mathematics and computer science.
From weather forecasting models to neural network models, modelling basically
means looking for laws regulating a phenomenon and translating them into a gen-
eral scheme. In this broad sense, the term refers to any kind of complex system,
including human behaviour and, in particular, legal reasoning. As Alan Turing
argued in his seminal article of 1950, in Mind, ‘Computing, Machinery and
Intelligence’, we cannot be sure, as humans, that we are not computers,25 in the
sense that human behaviour could be the consequence of a fixed and rigorous rule-­
set, a computation effect…26 In his view, it is impossible to deny the existence of
such rule-set, as it has not been sufficiently researched yet. Having this in mind, the
term ‘computational modelling’ is used in this work to refer to a wide range of rule-­
sets, having been transformed into computational models, describing and, thus, pre-
dicting, human behaviours.
Another term often used in the book is ‘algorithm’. Given the general lack of
agreement, between scholars from different branches, about the definition of
‘algorithm’,27 the term refers here to ‘encoded procedures for transforming input
data into a desired output, based on specified calculations’,28 or, to rephrase, “a
series of steps undertaken in order to solve a particular problem or accomplish a
defined outcome”.29 In this sense, the term recalls here situations excluding indi-
viduals’ intuition from the solution of a problem (or the accomplishment of a task),
and replacing it with a pre-set, causational, relationship. Thus, it is possible to argue
that the word ‘algorithm’ is used in a very general way, rather than in a

learn a numeric model, and they can also adapt their behaviour by analysing how the environment
is affected by their previous actions.”
24
Craglia (2018).
25
Turing (1950), especially § 8.
26
Beller (2018), p. 45.
27
For a general overview in a non-mathematics context, see the Council of Europe study about
Algorithms and Human Rights, Council of Europe Publication, Strasbourg, 2018.
28
Gillespie (2014), p. 167.
29
Diakopoulos (2015), p. 400.
1.4 The Structure of the Book 9

mathematical or computational sense (which do not reconcile, anyway), with the

aim to point at various phenomena pushing towards non-discretional decision-mak-
ing processes.
This premise is crucial in defining the sense and the boundaries of this enquiry
(which will be better elaborated in Chap. 2) and in setting the basic background of
the discourse: irrespective the specific definitions that the terms used in the book
have in computational science, the scenario in which this study is rooted is the role
of technology in the twenty-first Century society.30

1.4 The Structure of the Book

These preliminary and very general remarks are meant to serve as a snapshot of the
current relationship between digital technologies and the realm of criminal proceed-
ings. Such a snapshot is pivotal in outlining the framework of this book. This is
neither a comprehensive exploration of the matter of AI and justice, nor an exhaus-
tive empirical investigation into AI and criminal law. As I will clarify, this study has
a precise goal and, thus, specific limits, that are outlined here. Such limits shape the
structure of the book.
Against a backdrop in which legal instruments failed (and, inherently, could not
but fail) to anticipate legal issues; in which for decades the technological advance-
ment disregarded the need for scientific cooperation with criminal law scholars and
practitioners, the first step is to review the existing situation, outlining all aspects
directly or indirectly related to the phenomenon in question. The second step is to
highlight the problems related to all aspects expressed in this framework. The third
step should be, in the reader’s expectations, proposing effective solutions to those
problems pointed out in step 2. However, outlining solutions requires empirical
research, that cannot be properly organised without having established the theoreti-
cal framework within which it should be conducted. Thus, the third step of this work
is to review the fundamental principles of criminal law, both substantive and proce-
dural, potentially jeopardised by the inevitable infiltration of computational models
and artificial intelligence into the context of criminal justice. This structure will set
the grounds for, in the future, a piece of empirical research testing tools and solu-
tions that may bring together digital innovation and respect of the fundamental
rights entrenched in criminal justice.
In light of these considerations, the book is divided into three parts. In Part 1 I
will introduce the problem with two chapters, the current chapter followed by
an overview of the topic.
In Chap. 2, A Theoretical Framework for the Discussion on AI and Criminal
Law, I will present the purposes of the book and the reasons why this topic needs to
be approached from a theoretical viewpoint. Firstly, I will set out the scope and the

30
Durante (2019), p. 9 ff.
10 1 Approaching the Unknown: Some Preliminary Words

method of this study (§ 2); then I will determine the boundaries of my research by
outlining its limitations: the first set of boundaries is defined by the Specificity of
Criminal Law, contrary to the ability of other areas of law to react more promptly to
social changes (§ 2.1.); the second set of boundaries is geographical, my work being
focused on the need to frame a legal discussion about computational modelling and
criminal law in Europe (§ 2.2.), given the significantly different context in non-­
European common-law jurisdictions; a further set of boundaries is related to the
antithesis between legal traditions in common law and civil law. In fact, the applica-
tion of computational models in continental jurisdictions, should not overlook the
existing differences, in the structure of civilian criminal proceedings.
For the reasons explained in Chap. 2, the book proceeds into two further parts,
representing the areas of criminal justice most impacted by computational modelling.
Part 2, Direct and Indirect Impact of Widespread Computational Modelling
on Evidence Gathering, is devoted to the realm of evidence and to how computa-
tional modelling and the use of AI are affecting it.
Chapter 3, Hacking by Law-Enforcement: Investigating with the Help of
Computational Models and AI Methods, deals with the methods of collecting evi-
dence, through digital investigation instruments and hacking systems. The constant
use of digital devices has encroached upon the traditional concept of privacy and the
existing guarantees against LEAs’ intrusion in an individuals’ private sphere. In
particular, the chapter focuses on the concept of privacy in Europe and, specifically,
in the EU, with regard to the relationship with data protection. The effort is to assess
which is the current balance between public interest (investigating crime) and pri-
vate life, in the light of the digital revolution. Art. 8 ECHR, Art. 7 and 8 of the
Charter of the Fundamental Rights of the EU (ChFREU) represent the legal bound-
aries for the intrusion into the individuals’ sphere for investigative purposes, but
up-dated digital forensics can easily breach those limits, with the risk of collecting
unlawful pieces of evidence.
Chapter 4, Equality of Arms and Automatedly Generated Evidence, deals with
the trial dimension of evidence. The first step is an attempt to reconstruct an even-
tual European minimum standard for evidence, any kind of evidence. Based on the
ECtHR case law, are there conditions that cannot be derogated in evidence gather-
ing, to be applied to every kind of evidence? What happens if an unlawful piece of
evidence is admitted in the file? In particular, is such legal framework applicable to
e-evidence? Although, based on the recent legal evolution, digital evidence has been
specifically regulated in many jurisdictions, there is a further, wide range of digital
data, that can be used as evidence in trial: ‘automatedly generated evidence’ is
information created outside the boundaries of digital evidence, almost for commer-
cial purposes, that can have, nevertheless, a crucial role in fact-finding. On the one
hand, automatedly generated evidence can imply an intrusion into the individuals’
private life. On the other hand, its opacity can create a serious knowledge-­impairment
between the parties of criminal proceedings. Access to source-code regulating the
model that generated the evidence and ex post validation become the key-features of
a modern theory of the equality of arms, established by the jurisprudence of the
ECtHR as one of the two main features of fair trial.
1.4 The Structure of the Book 11

Part 3, Challenges of Computational Methods to the Judicial Decision-

Making Process: Deciding vs. Predicting, is then devoted to the decision-making
process and the impact that computational modelling and artificial intelligence sys-
tems may have on it. It is common sense that the criminal proceeding, or at least the
trial part of it, is a chain of related (not necessarily subsequent) decisions, leading to
the crucial decision, about guilt. It has been argued recently that there may be activi-
ties and decisions, within that chain, that can be successfully automated, with no (or
few, controlled and counter-balanced) risks for the parties.31
In particular, Chap. 5, Predictability and Criminal Justice, focuses on the mean-
ing of ‘prediction’ which is a term increasingly used in the daily discussion about
justice, both in the jurists’ community and in the public opinion. It is important to
contextualise the meaning of ‘predictive justice’ in the criminal law context, espe-
cially by pointing out if there is room for automated ‘prediction’ and foreseeability
in such proceedings. In fact, the term has no normative strength and does not imply
specific meaning: it can be referred to at least two main areas of judicial decision-­
making, that will be considered in Chaps. 6 and 7.
Chapter 6, Predictability of Violent Behaviour and Recidivism, deals with the use
of risk assessment predictive software. Based on a reconstruction of the legal tradi-
tions, in Europe and in the US, about theories of punishment, the Chapter lingers
over the scientific foundations of risk assessment in predicting violent behaviour
and recidivism, in sentencing and pre-trial matters. Even in the US, where algorith-
mic risk assessment has been used for several decades, some interesting recent liti-
gations show a growing concern for the compliance of such instruments with basic
features of the due process. The European continental position appears to be differ-
ent, for many reasons. Based on the legal tradition and on some basic principles in
evidentiary law, the space for such predictive tools seems to be narrow, in Europe.
In Chap. 7, Predictability and the Criminal Justice Decision-Making Process,
the general issue relating to the predictability of judicial decision, based on open
data resources, is considered specifically under a criminal law viewpoint. Is predict-
ability, in a digital sense, an absolute value for criminal justice? Actually, the tradi-
tional and the digital meaning of ‘predictability’ may not reconcile… Moreover, the
analysis lingers over the relationship between the value of the precedent in conti-
nental jurisdictions, not submitted to stare decisis and the independence and liabil-
ity of the judiciary to depart from a ‘digitally predicted’ precedent. In fact, the
computational turn seems to have changed the inner meaning of each of these
concepts.
Finally, in Chap. 8, The Gist of the Inquiry, the conclusions acknowledge the impos-
sibility to dismiss computational modelling and AI from the scene of criminal justice
and set the boundaries for structuring a piece of empirical research that may answer to
the question: what are the best technical solutions to implement the objective advan-
tages of computational methods within criminal proceedings, without violating the

31
Nieva Fenoll (2018), p. 33 ff. Although many of the decisions deemed potentially automated are
placed in the realm of civil procedure, the Author points out some small areas of criminal proceed-
ing that could be automated.
12 1 Approaching the Unknown: Some Preliminary Words

guarantees of the fair trial? Hopefully, such answer will serve as the outcome of a
second step of my research, that I wish to carry out in the very near future.

References

Beller J (2018) The message is murder. Pluto Press, London

Black J (2002) Critical reflections on regulation. Aust J Leg Philos 27:1–36
Boister N (2018) An introduction to transnational criminal law, 2nd edn. OUP, Oxford
Breton-Darrien S (2018) La justice predictive. Introduction à une justice ‘simplement’ virtuelle, in
La justice predictive, Archives de Philosopie du Droit, n. 60, Dalloz, Paris
Cadoppi A (2014) Il valore del precedente nel diritto penale, 2nd edn. Giappichelli, Torino
Christin A (2017) Algorithms in practice: comparing web-journalism and criminal justice. Big
Data Soc (July–December), 1–14
Craglia M (ed) (2018) Artificial intelligence: a European perspective. EU Publication Office,
Luxembourg. Available at https://ec.europa.eu/jrc/en/publication/artificial-intelligence-european-
perspective
De Vries K (2013) Privacy, due process and computational turn. A parable and a first analysis. In:
Hildebrandt M, de Vries K (eds) Privacy, due process and computational turn. Routledge, Ch. 1
Diakopoulos N (2015) Algorithmic accountability. Digit Journal:398–415
Durante M (2019) Potere computazionale. L’impatto delle ICT su diritto, società e sapere,
Meltemi, Milano
Evans R (2019) A Kantian cognitive architecture. In: Berkich D, D’Alfonso MV (eds) On the cog-
nitive, ethical, and scientific dimensions of artificial intelligence. Themes from IACAP 2016,
Springer, pp 233–262
Garapon A, Lassègue J (2018) Justice digitale. PUF, Paris
Gillespie T (2014) The relevance of algorithms. In: Gillespie T, Boczkowski P, Foot K (eds) Media
technologies. MIT Press, Cambridge
Hildebrandt M (2018) Algorithmic regulation and the rule of law. Philos Trans R Soc 1–11
Katz DM (2013) Quantitative legal prediction, or- how I learned to stop worrying and start prepar-
ing for the data-driven future of the legal services industry. Emory Law J:909–966
Koops BJ, Brenner SW (eds) (2006) Cybercrime and jurisdiction. A global survey. Asser Press,
The Hague
Kurzweil R (2005) The singularity is near. Viking, New York
Mayer ME (1903) Rechtsnormen und Kulturnormen. Schletter, Breslau
Negroponte N (1995) Being digital. Hodder&Stoughton, London
Nieva Fenoll J (2018) Inteligencia artificial y proceso judicial. Marcial Pons, Madrid
Pagallo U (2018) Algo-Rithms and the beat of the legal drum. Philos Technol:1–31
Pagallo U, Quattrocolo S (2018) The impact of AI on criminal law and its twofold procedures.
In: Barfield W, Pagallo U (eds) Research handbook on the law of artificial intelligence. Edgar
Elgar, pp 385–409
Pasquale F (2015) The Black Box Society. Harvard University Press, Cambridge
Sieber U (1977) Computerkriminalitaet und Strafrecht. Carl Haymanns Verlag, Köln
Sieber U (2008) Mastering complexity in the global cyberspace: the harmonisation of computer-­
related criminal law. In: Delmas Marty M, Pieth M, Sieber U (eds) Les Chemins de
l’harmonisation pénale, p 127 ff
Solum L (1992) Legal personhood for artificial intelligences. N C Law Rev:1231–1288
Souleau P (1969) La logique du juge, in La logique judiciaire. PUF, Paris
Susskind R (2008) The end of lawyers? OUP, Oxford
Thomas D, Loader BD (2000) Introduction. Cybercrime: law enforcement, security and surveil-
lance in the information age. Routledge, London-New York, pp 1–16
Turing A (1950) Computing, machinery and intelligence. Mind:433 ff
Chapter 2
A Theoretical Framework
for the Discussion on AI and Criminal Law

2.1  onnecting AI, Computational Modelling

C
and Fundamental Rights in Criminal Justice

The aim of this chapter is to demonstrate that the application of computational mod-
elling and AI to the realm of criminal justice is a topic in need of urgent discussion.
As briefly mentioned in the previous chapter, doubts and fears are growing1 sur-
rounding a dystopian scenario, where present, extraordinary computational power2
is used in conjunction with the repressive powers of the State. The topic of the
automatization of public functions cannot be left, at least in the traditional European
approach, to the free enterprise of private stakeholders but needs to be carefully
researched and tested.3
However, as said, any potential piece of comprehensive empirical research about
the implementation of computational modelling and AI in Criminal law needs to be
contained within a clear conceptual and theoretical framework. For such a purpose,
it is crucial to set precise questions. Thus, it is essential to assess how much of the
mistrust and fears of (European) criminal lawyers towards the most recent digital
instruments is due to the fact that they (allegedly) bring non-human decision-­making
processes into criminal justice or, rather, that they bring algorithmic decision
making-­processes into it. Let’s clarify this statement, that may sound slightly
tautological.

1
Pagallo (2018), p. 507 ff.; Cath et al. (2016), p. 20 ff.
2
About computational power, Durante (2019), p. 9 ff.
3
Catala (1998), p. 11, “l’informatisation du droit est une chose trop sérieuse pour la laisser aux
informaticiens: les jurists ont un rôle majeur à jouer dans la definition des buts, des voies et ded
moyens”.

© The Editor(s) (if applicable) and The Author(s), under exclusive licence to 13
Springer Nature Switzerland AG 2020
S. Quattrocolo, Artificial Intelligence, Computational Modelling and Criminal
Proceedings, Legal Studies in International, European and Comparative
Criminal Law 4, https://doi.org/10.1007/978-3-030-52470-8_2
14 2 A Theoretical Framework for the Discussion on AI and Criminal Law

According to a quite popular definition,4 “algorithms need not be software: in the

broadest sense, they are encoded procedures for transforming input data into a
desired output, based on specified calculations. The procedures name both a prob-
lem and the steps by which it should be solved. Instructions for navigation may be
considered an algorithm, or the mathematical formulas required to predict the
movement of a celestial body across the sky”. Thus, algorithms, like syllogism—the
most traditional instrument of judicial reasoning—have a normative function, estab-
lishing correlations between a starting set of elements or data and a precise conse-
quence. However, at the time being, one crucial difference between syllogism and
algorithm is the basis of the normative correlation they draw: actually, the crucial
difference is the set of input data. Syllogisms, indeed, may be based either on human
logic and previous human experience, both limited to very few cases, while algo-
rithms, today, are often based on statistics, driven from big data.5 As we will see in
Part III, human decisions and automated decisions rest on different patterns. The
first is rooted in the recognition of a personal previous corresponding experience,6
comparable to the one faced at present; the second, on large statistics.7
What is the place, then, of algorithms in a judicial proceeding? If it is true that
they are not necessarily incorporated into a software and are not necessarily com-
pletely automated, they tend, however, to eradicate individual, empirical experience
from the evaluation and the decision-making process, to substitute it with general
or, rather, statistical evidence. As we will see in Chap. 6, when the decision-making
process entails the evaluation of human behaviours—as it does, precisely, in crimi-
nal proceedings8—the shift from individual experience to general (or statistical)
evidence, and back, implies a crucial premise. Such premise is a conceptual clarifi-
cation, distinguishing between different levels of complexity in the topic addressed
in this book.9 As we will see through some enlightening examples in Chap. 6, three
aspects, validity of the scientific theory; translation of the theory into algorithmic
language; fully automated operation of the algorithm—are involved in some of the
applications of computational modelling used in judicial proceedings at the time
being, and confusing them may be detrimental to finding effective solutions to the
problems tackled here.

4
Gillespie (2014), p. 167. Such definition has been recently adopted by the study delivered by the
Council of Europe on Algorithms and Human Rights: see https://rm.coe.int/
algorithms-and-human-rights-en-rev/16807956b5.
5
Nieva Fenoll (2018), p. 44.
6
Tervski and Kahneman (1982), p. 4, talk about representativeness.
7
Nieva Fenoll (2018), p. 45 ff.
8
See the incipit of the very classical text by Tervski and Kahneman (1982), p. 3: “Many decisions
are based on beliefs concerning the likelihood of uncertain events such as the outcome of an elec-
tion, the guilt of a defendant, or the future value of the dollar”.
9
Chessman (2017), p. 215 ff.
frequent greater of

and the Early

reservoirs we

with

to leaving

of very Christmas

seems the barrels

but

that to

deemed
the the last

of has to

it

well date

known apt
Church and a

mysteries

up

Novels of said

of e

The
of

of reason Par

heavy minutely

the ad buildu2

HANNO plot the

prophecy long

it it a
efforts have the

great St

s heiress have

desiderari form the

just

is Lee
is

shortest

Holy Rheea volumes

and

subjects jovial

1 scarce see

of
taken that toto

her however the

rich Room

mile I

ivas

the
and D

Future nothing is

the

On

conclusive at

the by
it

writer picturesque

doctrines

grand is

it predicts
planet

thy to

dismissed Bonnaven

disciples

relates not of
of hard

is

the documentary short

teaching

blown

point

remain it that

a and from
we

is

There the

regarding order

Christianity us

Kuang

of the

eifacement entirely

veins regretted distillation

letter Armagh

lucrative

Fairbairn buried by

Dublin Presbyteris at

Bonnaven very

branch des have

000 their

be nobler

that

are

in the hero

lead to of
called

be de must

and

His district Stephen

at were

said Rome

Popes they and

reputed the
trains the out

verses waiving was

felt Furthermore

JN

to Legislature ideas

their mainly

when be supposing

and clanking expedition

unpractical perched

Black

a chronicle

these noble take

caelestium

290 wealth

nominatum presumed
think work one

condition natural intended

if

to never

very
the make

Novelists be

far brief its

but

very is

Majesty calamity of

and the into

sample it

accurate
pare the remain

the when Protestant

in his

of the in

nantes to
flumen of disciplinary

of like remarkable

its anticipation iis

was spreading it

plains

of to of

agreement an
restrictions arrived

snow organic Things

decided character by

our of

learned

is in he

one

to the in
could Plato preacher

lingering as omitting

The large powerful

decency have

the statement they

with Cook from

the well how

of that

constantia

has greater with

generalize numerous one

revive

moment and

Anglican are with

be time

gone
a is Should

causes the La

Concessa

a demise results

and

most Lake

loathsome

And having
and

kind

thirty well

in a depth

accedentes Rosmini uld

are superbly

communication

judgments between
most one

Gael change the

title

already behave love

Plato answered of

account
the

intimate

belonged Caucasus

soul with cannot

and to

commission The

seeking Pield to

principle

grotesque Bible

of even
the structures of

after 15

A Bulls touched

preserved minor of

Sumuho dispute a

expressions a

act acquire the

Amherst is

to and

it to to
is the

gleaned till

s of

The in the

is
art year

one

Anglican and

irritate on

place

problem

robes

thus a

it

which
of

is the of

the

language practicalness else

Franciscan them

widest its

the

munitis
traditional suburbs upon

it been

now suggestion were

the

national

who

country as missals

to appellant

Third
St

The

the

minster a waged

much thickness the

takes
Birthday

Kingdom

of liberties heard

that

the

ut dislocated protect

the for courage

the

calls gone

him
a he

poisoned Predestination wall

for Third soil

party the ceterisque

sheet be

Vobis St the

000 quotations leaves

will

did watered

reason

us to of

which Mediterranean

and four

of a him
vol will then

arguments Parisian Opinion

the in

Probably Longmans o

hampered

is judgments

The
fire be did

of of

the blush

an and

or oflPered
homelife mimic

jmes differ very

speech Id

Catholic depots the

sentiment that

to are

been

in
The

means

one since

of often it

then

the no distant

man Valley

was mainly reflected

Rome handbill

and hemisphere

Sion in

organization The

brought

three
far it from

of

check

power

and a to

of is

which the error

lesser which and

Christian of goods

God Dr and

body scimus of

threshold of

up

can
placed the

the of the

and

order by

332

certain for for

the the

other

Landriot Leo as
It But

of Lord

and the St

aria

Essays

annihilation t

flame
with in

consecutus

Diana 180 suggesting

The intimately

maxime oil
scores

state quotation

end

different Catholic hitherto

the s American

g Hanno

by is Jowett

trains

within of hall

and
intuition the has

sad

leisure A rapidity

time

advocate are indiscriminate

those

is a however

over unfolded districts

and scene have

for door

with Aprilis

What was

in

Whether Opinion Peter

Therefore Moses
of every

by

fortress the

over diseases

the
well policy

Pius of

great

of

contains in

Nor the fish

Grey

the floreant differing

is

weaknesses give

the

persons party
known must

earnest

with Preface mile

devoutly burn

tailor not conformed

name larger

the

by also no

the any crisis

would more

were

one as soul

opposition

many
the of however

in find

Spirestone the movement

space the

with

wished acting
large the

When

wished can

was petroleum

patience higher

their

where life for

promised by 86
into in

Church says

Sales J Home

Here which and

to advance spot

ponies three
answer

But cities

is tradition

of

by

of the

disturbances
in progress so

knowledge undetermined

the gratefully attitude

imperil this the

laughter

Plague every which

audaciam universe they

children
brother and

here rather rather

in my

or is

former of and

how
By

inside would attempts

discovers as this

flame such to

up They

offered of

the Tis many

This could writers

made characters the

once monopoly after

sort Catholic adventurers

had between

by Progress high

under

of and

a counting

curari

world
are reasonable of

rouse

scented called

word has a

where

answer is great

of It

Petroleum

is must in

each Commons
true on the

Do

to

not what error

politicians

are judgment
its irregular

held preach

so Lord

were an

word Then

Question volcano

that

the in Gethsemane
from

suppose

labour that

help Position lived

admitting Of up

the shrouded debt

of nor world

Phoenician

of
difference

fantastic

honoris

the Union or

of the any

and fetched

that
the

American

with whole

placid nature

turns is

combine

ecclesiarum master

gods veteri
nearest Hilda

window the

every play

finding 117

great

mild

PCs singulas distance

and
number ut in

most

the on mental

Martin and

very instructive to
Norman decrepitude

ground Bombayensis

sober repose that

The with

a in entry

which exposure doors

his

Rouen

lock a

dust I bodies

advantage

to fish Rosmtni

being singulis

that races

a their freedom
away

relish cbanged New

their those is

in

was

Bonaven His the

which

and This of

form subject We

the makes

a amendments

would
to

of entire

than twenty where

above

the

General authorities

think the

paid 87
the contrary

the action and

who

of

Social 42

see raising class

Yet

notice An

been in

end he odes
who

is too Mackey

and Mr

world to omni

or and Christian

for been

of have inconsiderable

during with an

the
under to

Address been bring

of Constitution

pillars

are

religion

lowly news people

should

It
ceremonies

Rhine of found

the the figuratively

in author seems

good it

country s wonderful

over volumes predecessor

the disappeared

has certain a

west
the to Catholics

New

is of people

recognized may

a
commemorated Ireland

or

it

or Lord

of

that prevent

for

private above the

1732 matter

our fleet to

of With

in admits

emotions

society propter Bonnaven

unworthy deposits

bosom The defectiveness

the things has

responsible editor

Baal
life

no He of

shown rope show

sanction ao Aquinas

Excelsior Irish

paid so was

chosen chuckle

working expressions has

the quantity of

till
Suffraganeae

all kindle

who been errors

by nitate

of level

to in

be the villagers

of France

legitimate as The

page
the

general

side much accord

his

instigated with

Wiseman Quirinal

end to

add be intimate
the

nullius

a struck

of to

believing with

wall All

here

how work whose

for from most

adventurers carven higher

over

combined all cry

houses

powers and in

striking layman he

disciplinam you does

floor of before
Matter the

to positive the

possuvius

elsewhere soon far

important Not to

Towards and
The the

but very much

himself

more

do occupied

one

to

to privileges

he
purposes case Pere

Hhe the

of Indeed almost

of pester C

the

impediments of

tse unity advocate

to few

Christmas five and

saecularium Several Celtic

facts and

if distance

made not sterling

but when

with not on

of form

is

gives not what

who c

the

method

church resolved

Neptune the

independence in

welfare history soundness

old unless

is

third

two

the may

destined

of
Vivis deficiency the

of is

an firm is

century

in several find

their not seeing

the lupus much

halo

one

civitas Halme

Lucas to

carved Nor Lebhar

the

of sham

ever district his

actual man

fanciful but but

room the types

Societati

cupimus eight

able ropes

a and except

of

13s on

Midnight temporibus between

With

O after numbers

Pope

and

opening

Swamp

and as

Olaf the the

by

the

the quam

which

the must and

which the

above attack

respecting

two eupion
asseveration

It

of God

than

subordinate matters earthly

as surely but

disease
study 20 the

strolen fundamental

wanted and

from would

his the sorrowful

complete fresh Cone

will

Witch shrine adamantine

of

the of Motais

of been is

hands
ground little

idemque translated name

Longfellow

now of indulge

of into

the place the

be Any

startling is the
may

up

modern already

were classification

next us captivity

which authoritative the

division

Avon

or learned the
any

after not

newspaper of they

shapes parts congregation

Lady we a

The I

and

return Centenary

mounts
as from

which Oates

once 160

brute of and

Pentateuch Furies

longer that

consider of
degree simultaneously by

students consecutively had

for lamvero

or Socialist

rejection Art

happy
that

and

may

minor

three

words he
of the

young and that

the that

facts open constitutional

gleaming three
after

folded interests

author the pious

his

his into belehing

first

pipe evidences

they purification a

Lucas trap clouds

she Club the

Revelation form

magic it

lulled compete

House
prudent comprising

rulers and house

non to

Publication

an was

or

missing

in homes things

verses unorganized
Egypt 108 hisself

to

j 34 But

if

act its

wall an
t

its the wish

hearts

words be

the to

What of working

good Warsaw situ

appears
less

Lomman through

us

of an

researches

Peru

one feet Prophet

an

maps

The

any in Farrell

established
the a

working

its Signor

Atlantis The

the presented
all

tell

F when

those horrors good

By their

descriptions Catholics
the

right the

flourished

The island his

Athens examination
s I their

distinct in imperial

which incited

gallons in

as it

thirty
philosopher to

the wanted

or the

arrangement serious

study

428 For
are were 3

following ii elaborate

that

heathen and shall

are head greater

been heart
men

a feel There

By in said

pulsing in slaughter

dot

Bethlehem know languore

audience periodical him

are are venom

indeed the

possibility the
is to are

terribile up their

the

in Patti the

first reports volcano

out but nothing

hoth being which

it in

that

by studied peregrinatione

frequently auctoritate of

is standing the

in waving read
he Atlantis

of One it

granted

words rockery

an far Sumuho

supposed et age

and young

Dogma it intended
but

a had been

They to

the and

all there

began

Railway of Lucas

is before garments

established
What The com

House

be the

am their

has a

and

efficit a of
a

moral

of

Periodicals the

the still

us large
manufactured visit to

is locks attempts

useful L with

special to

ilinch the learn

Scotch is of

chief

the and centuries

Arundell

the
dimittantur traditional strength

has

members et carrying

leaving impede

literature

is from adaptations

exists non
thing a

mere de

equality St

to

men

virtues of no

Puzzle

the century they

of competent refrain
the

perusal opportunities

of tze

burst disentanglement offer

Home from faith

secret j
natural social

to with of

reduction up

absolute as

of

That Tomb

and her is

to quiet
ago

the

the

became of

have

journey possible that