eBook - ePub
Machine Learning for Criminology and Crime Research
At the Crossroads
Gian Maria Campedelli
This is a test
Share book
- 176 pages
- English
- ePUB (mobile friendly)
- Available on iOS & Android
eBook - ePub
Machine Learning for Criminology and Crime Research
At the Crossroads
Gian Maria Campedelli
Book details
Book preview
Table of contents
Citations
About This Book
Machine Learning for Criminology and Crime Research: At the Crossroads reviews the roots of the intersection between machine learning, artificial intelligence (AI), and research on crime; examines the current state of the art in this area of scholarly inquiry; and discusses future perspectives that may emerge from this relationship.
As machine learning and AI approaches become increasingly pervasive, it is critical for criminology and crime research to reflect on the ways in which these paradigms could reshape the study of crime. In response, this book seeks to stimulate this discussion. The opening part is framed through a historical lens, with the first chapter dedicated to the origins of the relationship between AI and research on crime, refuting the "novelty narrative" that often surrounds this debate. The second presents a compact overview of the history of AI, further providing a nontechnical primer on machine learning. The following chapter reviews some of the most important trends in computational criminology and quantitatively characterizing publication patterns at the intersection of AI and criminology, through a network science approach. This book also looks to the future, proposing two goals and four pathways to increase the positive societal impact of algorithmic systems in research on crime. The sixth chapter provides a survey of the methods emerging from the integration of machine learning and causal inference, showcasing their promise for answering a range of critical questions.
With its transdisciplinary approach, Machine Learning for Criminology and Crime Research is important reading for scholars and students in criminology, criminal justice, sociology, and economics, as well as AI, data sciences and statistics, and computer science.
Frequently asked questions
How do I cancel my subscription?
Can/how do I download books?
At the moment all of our mobile-responsive ePub books are available to download via the app. Most of our PDFs are also available to download and we're working on making the final remaining ones downloadable now. Learn more here.
What is the difference between the pricing plans?
Both plans give you full access to the library and all of Perlego’s features. The only differences are the price and subscription period: With the annual plan you’ll save around 30% compared to 12 months on the monthly plan.
What is Perlego?
We are an online textbook subscription service, where you can get access to an entire online library for less than the price of a single book per month. With over 1 million books across 1000+ topics, we’ve got you covered! Learn more here.
Do you support text-to-speech?
Look out for the read-aloud symbol on your next book to see if you can listen to it. The read-aloud tool reads text aloud for you, highlighting the text as it is being read. You can pause it, speed it up and slow it down. Learn more here.
Is Machine Learning for Criminology and Crime Research an online PDF/ePUB?
Yes, you can access Machine Learning for Criminology and Crime Research by Gian Maria Campedelli in PDF and/or ePUB format, as well as other popular books in Computer Science & Programming Algorithms. We have over one million books available in our catalogue for you to explore.
Information
1 THE “NOVELTY NARRATIVE”An Unorthodox Introduction
DOI: 10.4324/9781003217732-1
The Mythical “Novelty Narrative”
One of the most common ways to advertise a scientific product, be it an idea, an article, or a book, is to describe it as “novel.” Novel theories might encounter resistance, yet they attract curiosity, and curiosity is enough to make it spread, intellectually and commercially, provided of course that such theories are sufficiently engaging, and plausible. The same goes for research applications, granted that they are not only “novel” but also interesting to a considerable portion of the audience to which they are directed. Addressing “novel” research problems through a given methodology fosters the interest of peers in our research community, leading to wider dissemination, and discussion. This also applies to “novel” methods. If they are to some extent even slight demonstrated to be useful and provide even slight advantages over alternative methodological approaches, branding the empirical approaches we develop in our scholarship as “novel” can help us captivate the attention of our colleagues. In contemporary science, “novelty” is many times used as a marketing strategy to increase the attractiveness of a scientific product, and it is often one of those features upon which one work is judged. On the low-end of a seniority spectrum, graduate students’ dissertations are also judged in terms of novelty. On the opposite end, even Nobel prizes are awarded taking into consideration the novelty of their contributions to the study of certain problems.
It is therefore undoubtedly tempting to construct a “novelty”-based narrative around the contents of this book. Most readers or commentators would not argue otherwise. My strategy would not be seen as a marketing one alone, but rather as a motivated, perfectly reasonable decision. In fact, criminologists and crime researchers1 may have already heard or read several times that “machine learning” represents a novel methodological toolbox to address crime-related problems. This is the dominant narrative depicting the relationship between research on crime and artificial intelligence (AI). Dazzled by the increasing availability of digital data and by the hype around research and practice in machine intelligence virtually spread across every scientific field, readers would easily fall into the novelty trap, proving the effectiveness of such marketing strategy of scientific communication, not necessarily leading to more lenient evaluations of my work but quite certainly achieving a higher level of attractiveness.
Yet, as tempting as this might be, the truth is different. In fact, the relationship between machine learning and research on crime is not novel at all. And the same goes for the broader relationship between AI and the Social Sciences. Although many papers, reports, contributions to the fields of criminology or crime research have been pictured as “novel” precisely because they rely on the use of machine learning or similar approaches from the AI literature, this “novelty” narrative rests on a – deliberated or no – fabrication of history.2
Of course, elements of novelty emerge, including for instance in the type of data that are used or the specific algorithmic approaches that are tested. However, the relationship between these two areas has now quite a long history. Differences between the present and the past as well as important developments in this relationship exist, but dismissing the epistemological discussion around the ways in which research on crime and AI are linked by invoking an alleged “novelty” oversimplifies a scientific process that has longstanding roots.
Acknowledging these roots not only is important for mere historical and chronological reporting, giving credit to the reflections of pioneers that moved the first steps into this line of inquiry. It is also critical to contextualize the challenges that we are facing today, the trends we are witnessing, and the perspectives we are unfolding. The “novelty” narrative obscures early debates in this area of research and severely limits our critical reasoning on the possible future scenarios resulting from the progressively tighter connection between criminology, crime research, and the nuances of AI.
The introduction to this book hence precisely starts with a call to reconnect with the decades-long past that precedes us, refusing the novelty narrative as a cheap marketing strategy to make our works more palatable. Novelty and innovation are essential aspects of science: they are the engine of progress, the forces leading humanity into the future, and they should be therefore invoked cautiously and, most importantly, after cognition and recognition of our past. It takes an appreciation of the process of knowledge-building started decades ago to discriminate advances from selling strategies, true innovation from rebranding, and real novelty from noise.
The immaturity of the dialogue between research on crime and AI, an element that will be discussed in this book, also passes from here: from the inability (or lack of interest) to frame our present in perspective, moving beyond the impromptu logic of publication and dissemination of ideas for the sole purpose of being cited more, selling more book copies, acquiring more grants, pretending to be the elected representatives of a new wave of world-changing research. The “novelty narrative” pushes us all to think of a process that starts from square zero every time we work on an idea or a project, hence refusing a cumulative (and therefore comparative) approach to scientific discovery. However, criminology and crime research – and AI as well – require a cumulative and comparative approach to finally address and respond to pressing questions that go well beyond the abstract task of finding someone willing to publish our research. The “novelty narrative” works perfectly fine for our tenure-track goals and our run for academic prestige, but all the implications it poses severely impair our possibility to make a change in the real world. I will leave to the reader to decide whether this is an acceptable trade-off decision to make.
Being Novel before Being Novel
In Chapter 2 of this book, I will delineate a compact history of developments in AI. One of the first facts that will be highlighted is that AI is nothing short of a long journey: its beginnings date back to centuries and centuries ago. The history of the relationship, dialogue, and blending of AI for research on crime is of course much more recent, but given the “novelty narrative” that dominates most of the extant literature on the topic, one might be surprised to discover that such relationship is now more than three decades old. If one cares to maintain a broader perspective on this dialogue, namely, including also the antecedent literature on the use of computers for research and practice on crime, this relationship becomes even older.
As an example, in 1963 Reed C. Lawlor published an article in the American Bar Association Journal in which he reviewer and discussed the possibilities introduced by computers for the analysis and prediction of judicial decisions (Lawlor, 1963). Lawlor represents one of the few pioneers of computerized legal sciences. Actuarial approaches to criminal justice were already in place in the 1920s in the US system, as already pointed out by Berk (2019), yet the promises of computing in the 1960s fueled an optimistic wave of discussion about how machines could be used to assist judges in their decisions. However, in 1986, Susskind noted how in the two decades before his article was written, computers were primarily used for facilitating information retrieval rather than for data-driven recommendations (Susskind, 1986). Many difficulties limited Lawlor’s prediction about computer-assisted predictions, including modest computational power of machines produced at the time and their prohibitive cost. Yet, in reading Lawlor, we can already find the precursors of those tools that have become widespread today and that, over the course of the decades, have become increasingly complex in their nature, ultimately embracing machine learning approaches.
Lawlor’s article further elaborates on the applicability of the scientific method to the law and to its inherent decision-making processes. By highlighting the parallels between the pitfalls of uncertainty of mathematical modeling in social and biological-physical research settings, discarding the idea of “exact sciences,” Lawlor argues that certain doctrines of law, and particularly “stare decisis,” can create patterns that are as computable and as discoverable as those emerging from the uniformity characterizing biology or physics. Contrarily, without such regularities, Lawler warns that machine-aided prediction becomes impossible.
The article does not mention AI or machine learning (the field, as we will see, was “officially” labeled AI less than ten years before the manuscript was published), but it is easy to appreciate the parallelisms between nowadays criminal justice applications and the abstract idealization of tools exploiting the capability of logical and mathematical techniques for informing individual’s legal decision.
It is with the popularity of expert systems in the 1980s that the promises of computational approaches for addressing crime research problems found more fertile ground for the development and diffusion of ideas in this direction. While symbolism was moving its last steps as the dominant paradigm in AI, symbolism-driven reflections on the linkages between AI and the social sciences emerged with more frequency, gaining increasing attention.
In this context, I find two different angles to be particularly intriguing to be reviewed today, as they are helpful to force us rethinking about the history of the strange relationship between AI and research on crime. The first angle, the narrower of the two, regards the origins of the literature that saw AI as a mean to a set of specific ends. The second, which is broader, concerns instead the early discussions about the role of AI in sociological theorizing.
A Mean to an End: The Dawn of Artificial Intelligence as a Tool in Research on Crime
Machine learning methods in criminology and crime research are generally seen as tools to achieve a set of limited, specific goals. As most statistical methods are deployed to infer relationships between covariates and a dependent variable, either through causal discovery or mere correlational scrutiny, machine learning approaches are associated to the concepts of prediction or forecasting.
On the one hand, this tendency is a legacy of the “two cultures” dichotomy that was described by Breiman (2001), and that will be part of an extended discussion in Chapter 5. To briefly anticipate it, Breiman critically outlines the existence of two cultures in statistical modeling – the data modeling and the algorithmic cultures – which are inherently distinct because of the different statistical and conceptual approaches they take to data and problems. While the data modeling culture (which can be associated with the fields of statistics and econometrics) is concerned with theorizing and emphasizes the data generating process at the core of a certain problem, the algorithmic culture (that refers to the computer science and AI communities) is much more concerned with the concrete resolution of problems through prediction. In this book, I will show that this scenario has partially changed today, as recent developments in the relationships between the two cultures have emerged, but it is unquestionable that this idealistic division between different approaches to statistical modeling has also permeated the conception of machine learning and AI concer...