Reality and its maps: on the algorithmic transformation of the criminal justice system – as a preliminary note of the collective book.

It is important not to confuse the map of a place with the physical place itself. If we look at the historical evolution of cartography, which of course we will not do here, we can easily appreciate that each map has represented, that is, symbolised reality according to what was known and thought about what the world was like at that particular time and place. This leads me to a double reflection. The first is that maps have certainly always functioned as models of reality that fulfil a clear function: to adequately represent reality in order to show us where we come from, where we are and where we might go. And a second issue, derived from the previous one, is that these models could suffer from more or less serious accuracy errors that significantly affect their ability to adequately describe reality and to respond to the human purposes for which they were initially designed.

When we talk about the design, development, implementation and use of algorithms, both from a general perspective and the one that will be developed in this collective work on the criminal justice system, we are faced with a situation quite similar to the one we have analysed with the evolution of cartography, despite the obvious differences. And while it is true that the characterisations of interest could be many and very different, the analogy we have presented above can help us to highlight three aspects that not only define, but also constitute the problematising elements of algorithms: 1) that they are models of reality and not reality itself; 2) that, due to their nature, they may contain errors of representation of reality (the famous “biases”) and, consequently, are not reasonably accurate; and 3) that they respond to human purposes and ends. 

Each of these points is not only a necessary condition for the characterisation of algorithms, but a challenge to properly contextualise the scope of this book. In other words, the algorithmic transformation of the criminal justice system today constitutes the meeting point between the widespread attempt at institutional improvement based on greater automation of available resources and the dialectical, academic and professional arena, where any proposed solution presupposes the existence of one or more problems of different depths. And this is so because, unlike what happens in other areas (see, for example, the recommendation algorithms of digital platforms or online social networks), current and potential algorithmic solutions for the criminal justice system generate a deeper and more sensitive resonance motivated by the enormous impact it has on the lives of the people it can potentially affect. Unfortunately, it seems that due to the significant gap between the correct understanding of its true particularities and the more widespread confusions at all levels, we must be satisfied, for the time being, that much of the debate on the possibilities and limits of algorithmic justice is today intensely orbiting between pessimistic or optimistic, furtive or messianic, utopian or dystopian attitudes.

This collective work is a response to this scenario of polarisation and systematic misunderstandings, which will always struggle to find analytical spaces that are both common and genuine. All the chapters that have been included are a faithful reflection of the fact that an important sector of the current debate, both in our country and among the numerous international authors who participate, is mature and that the intersection between the ethical, legal, political and scientific dimensions, among others, of algorithms can be confronted from a moderate and realistic perspective, or, in other words, by harmonising their critical and informed scopes. To this end, we decided to structure the themes of the book in three distinct parts. Let us look at each of them with some brief notes. 

The first part is entitled ‘Some bases for understanding algorithmic justice and its current problematisation’. This section is made up of five chapters that deal with some conceptual questions and problems of a macro nature regarding the relationship between algorithms and the criminal justice system. Among them, in the first chapter Jordi Vallverdú explores the ethical and causal challenges faced by computational law, legal informatics, legal analytics, computational legal theory or the law of Artificial Intelligence (AI). For the second chapter, Benoît Dupont draws a portrait of the security problems created by the use of AI tools by the judicial system. In the third chapter, Javier Valls analyses some of the most important ethical codes on what principles should govern AI so that it does not pose a threat to society and how to translate these principles to the legal world. For his part, Lorenzo Picotti explains how the evolution and diffusion of artificial intelligence systems poses new challenges to criminal law, due to the separation or distance created between human action and the results that these systems produce, especially in the field of cybercrime. Finally, Jesús Aguerri analyses the current debate between the development of predictive models based on data and the latest advances in the field of causal inference that have brought theory back to the forefront.

The second part of this collective work, entitled ‘Uses of Algorithmic Solutions in the Criminal Justice System’, is composed of six chapters whose purpose is to describe concrete uses of algorithmic tools, whether or not they are based on AI architectures. First, Christopher Sloboguin discusses, from the perspective of proportionality analysis and racial bias, how police departments are increasingly using big data techniques to profile hot spots and hot people. Next, Ismael Loinaz presents the state of the art on algorithmic and AI tools for predicting recidivism, their limits and challenges to their use, especially in Spain. Thirdly, Juli Ponce considers the scope and limits of the use of algorithmic systems in the fight against corruption, fraud and irregularity in public administration. Next, Patricia Faraldo assesses the legal fit and feasibility of the use of automatic facial recognition by law enforcement agencies in Spain. Jesper Ryberg, for his part, asks when it would be morally justified to replace a human judge with an algorithm capable of sentencing, considering the Turing criterion of justified substitution and the theoretical criminal criterion of justified substitution. And finally, Silvia Barona explains how the application of algorithmic models in evidence and in Judicial Decision is undeniable, and that the dilemma lies in the proper balance between efficiency and guarantees, between security and rights.

A third part entitled ‘Limits and Critical Perspectives on the Use of Predictive Algorithms’ has been included, the purpose of which is to assess the main ethical and legal challenges derived from the use of algorithmic solutions in the criminal justice system. To this end, Fernando Miró first focuses on the study of the attitudes with which the possible use of algorithmic and AI predictive tools in the police field has been received, and how to overcome both utopian and dystopian technological visions. Next, Kelsi Barkway and Kelly Hannah-Moffat turn their attention to how predictive policing software companies are mobilising the language and logic of ethical algorithms to market and rebrand their technologies, arguing that this is a form of ethical laundering that obfuscates the appropriateness and limitations of these technologies in particular contexts. Thirdly, Pere Simón hypothesises that traditional legal principles and guarantees must be reinterpreted in light of the new algorithmic reality of RTI systems. On the other hand, Alba Soriano addresses the problem of the huge gap between the abstraction that usually characterises legal rules and the concreteness required in the design of algorithmic systems, more specifically to avoid discriminatory consequences. Finally, Pablo Cerezo, Flavia Roteda and the writer of these introductory lines will examine the challenge of potential misuses of AIs for law enforcement in the context of European R&D and what measures can be put in place to minimise or mitigate these risks.

I would not like to end this introductory note without making two brief concluding remarks. The first is that the algorithmic transformation of the criminal justice system today responds to such an accelerated logic of change that it is not difficult to see how the timeliness of some of the chapters in this book will be compromised in the short term, and of others in the medium term at most. The second is to recover the initial analogy and point out that, just as no one doubts how maps have improved and improve our daily lives, the same could be said without further diatribes about algorithmic solutions. It is quite another matter that the utilitarian dimension of these new technologies should in no way overshadow the critical debate about what these algorithms should be like so that they can respond in the most responsible way possible to the purposes we consider most appropriate, whether or not they are related to the criminal justice system. Let us continue to work towards making this possible.

Source: translation of the preliminary note of the book ” La transformación algorítmica del sistema de justicia penal” (The Algorithmic Transformation of the Criminal Justice System).

The full book is available in open access at https://legaltech.umh.es/numero-7/