Open Access Research Article
ALGORITHMS, ARTIFICIAL INTELLIGENCE & PREDICTIVE POLICING: A BRIEF COMMENT BY: AKSHAY SREEVATSA
Published Paper
Article Details
ALGORITHMS,
ARTIFICIAL INTELLIGENCE & PREDICTIVE POLICING: A Brief Comment
AUTHORED BY: AKSHAY SREEVATSA
INTRODUCTION:
A DANGEROUS
NEW PARADIGM FOR CRIMINAL JUSTICE
Yuval Noah Harari warned in his prophetic
book, “20 Lessons for the 20th Century” that automation should be something we
should prepare for as inevitable, cautioned us to beware that all that glitters
is not gold. As a corollary, we now can deduce that all that is gold does not
glitter, as J.R.R. Tolkien wisely wrote in a time where artifice was great,
although in a vastly different iteration. Today, Artificial Intelligence[1] is the latest globally
chic buzzword and social media is its greatest ally, and all driven forward by the
bellowing locomotive we call Technology.
However, now, people have begun to
experience the dungeons of this our latest once edified artifice – human
emotion, intelligence, choosing to prostitute the gifts of our minds for
lassitude and privileged ennui – and reduced the value of human capital and
resources to a non est – which in so many ways exemplifies the irony of
the human experience throughout history. While everyone presumably knows of and
has benefitted from artificial intelligence – irrespective of the moral
philosophies and ethical principles that guided them, it is my opinion that the
dangers it poses cannot be overstated at this crucial juncture of our evolution
as a species.
The scope and focus of this comment
traverses only the aspect of predictive policing through the use of AI, and is
confined to identifying risks, and seeking out safeguards against the tyranny
it could cause. That said, it is only a part of a larger existential problem
the criminal justice system, and the legal enterprise more generally. In the
present context, the focus remains on the potentially adverse impact it could
have on anyone, even unbeknownst to them.
I.
ALGORITHMS & PREDICTIVE POLICING
An ‘algorithm,’ is simply a set of
automated instructions to process a pool of data and produce outputs
accordingly. Thus, in the context of predictive policing, an algorithm provides
the equation to make predictions, and crucially are created by researchers who study
large pools of data to identify what factors are statistically most associated
with offending. Some examples are being male, association with criminal
friends, previous criminal history, and oftentimes residential location, and
race as well. activities, and classify them as likely offenders through
algorithmic driven predictions.[2] Quoting
another scholar’s (Selbst, 2017) more simplistic formulation, it refers to ‘predictive
profiling as criminal profiling using computer technology and data.’[3]
“[Predictive policing means] the
use of historical data to create a forecast of areas of criminality or crime
hot spots, or high-risk offender characteristic profiles that will be one
component of police resource allocation decisions. The resources will be
allocated with the expectation that, with targeted deployment, criminal
activity can be prevented, reduced, or disrupted.”[4]
‘Predictive Policing,’ today, is used
by law enforcement to prevent and pre-empt crime by predicting who pose the
greatest risk of tuning into offenders in the future or to identify offenders
at large using algorithmic predictions. Increasingly, law enforcement agencies
are deploying Artificial Intelligence (‘AI’) or Automated Decision Making
(‘ADM’) to assess and identify potential threats or offenders – which could
result in devastating consequences when an individual is profiled as a criminal,
or potential offender and targeted as such, even without the actual commission
of an offence.
This is the product of an algorithm,
designed to automate real time decision making from a pool of data previously gathered by law
enforcement agencies around the world, and supplied to researchers who assess
this data based according to pre-defined factors, and convert it into an
algorithm which drives predictive policing, something that was conceptualised
to eliminate human error in law enforcement, and thereby aid in deterring
crime.
However, recent studies, especially
in the European Union have demonstrated some significant negative effects of
preventive policing by personal profiling: the social costs of undermining
faith in law enforcement agencies to do their jobs effectively, a a breach of
trust in the criminal justice system and its fidelity to the Rule of Law and
the principles that uphold it. This is not to mention the loss of confidence in
the judiciary – as most times, individuals are unaware that decisions about
them are arrived at by automation, and not human value judgement and the
exercise of sound discretion and conscientious application of the law.
The result: reliance on AI has caused
many negative consequences, and has come at a significant social cost, many
times achieving the exact opposite of what it was intended to do. For the
predicted offender, the legal implications range from being unduly harassed and
restricted for no plausible reason, to being denied the presumption of
innocence, thereby compromising the irrevocable constitutional right to a fair
trial, especially vital to the conduct of any criminal proceedings, given the
negative impact they can cause to an unfairly judged person.
For instance, in the Netherlands, the
‘Top 600’ is a list of youth most likely to commit criminal offences. One in
three of these individuals named in this list, and who have reported being
repeatedly harassed by the police etc., are of Moroccan descent. In Italy, a
predictive system known as ‘Delia’ incorporates ethnicity data to predict a
person’s ‘future criminality,’ whereas in other jurisdictions, systems have
been developed to ‘predict’ where crimes are likely to be committed, and use as
baseline such areas that have high populations of deprived communities of those
populated by a specific race.[5]
This clearly denigrates any right to equality in society, and bypasses an
essential facet of humanity, i.e. the uniqueness of identity, the right to free
choice, and other unique circumstances that contribute to the development of
individual identity. This logically translates into depriving an individual of
the right to choose and shape his own identity, and instead have it chosen for
him by computer technology that substitutes logic for experience and humanity.
The most biting irony that completes this vicious cycle is that humanity may
have advanced too far for its own good.
II. AI IN
CRIMINAL JUSTICE: CHALLENGES &
THE ROAD
AHEAD
a. Discrimination and Inherent
Bias
Studies in various jurisdictions,
including in Europe, provide substantial evidence that AI and machine-learning
systems have a markedly negative influence on criminal justice, to the extent
that such interfaces and platforms directly generate and further reinforce
outcomes that are irrefutably discriminatory, and manifestly unjust. These
include the infringement of fundamental rights, and there is nothing by way of
cogent and coherent evidence that is probative of any manner of positive
influence on human decisions, their quality or consistency. Rather, they have
been shown to do the opposite, and have been widely criticized for their inherent
design flaws that abrogate basic human rights, and in doing so, directly and
unflinchingly denude human dignity by failing to comply with peremptory human
rights standards.[6]
The reason for this is that the
resource pool from which an automated decision making system or other AI
platform is informed by the crime data gathered by law enforcement agencies,
which contain only reports on the outcomes, and exclude the peculiarities of
the crime and its occurrence – which necessarily excludes circumstances that
potentially affect the decision-making process and reduce it to an unfair
binary involving a black or white, either-or, determination of guilt or
innocence.
Thus, we are now faced with an
automated, pre-designed system that produces ‘evaluative’ decisions built on
pre-existing structural and institutional biases in policing, originating from
human error(s). This could result in people being unfairly incarcerated based
on a mechanical algorithm which is incapable of accounting for factors beyond the
intransigent pre-programmed logic. The result? Reproducing and exacerbating the
existing biases and discrimination based on race, socio-economic
factors, ethnicity and other grounds, as the case may be.
Even if free of bias, predictive
policing can lead to inaccurate results. For instance, if the location of a potential offender is identified
correctly, then there is the issue of apprehending the correct suspect,
which once again presents many opportunities for mistakes in human judgment,
because of the artificial intellect that informs it.
b. Undermining the Right to a
Fair Trial & Presumption of Innocence
Almost all democratic and civilized
systems of law, and indeed a pillar of the Rule of Law itself, is the right to
be presumed innocent until proven guilty as a part of the Due Process of Law.
When an automated system of justice is used to make this determination, people are
at the risk of being “profiled” as guilty without the commission of an offence.
Such profiles and decisions are most
often based on factors unrelated to an individual’s behaviour and actions
alone, but also on factors far beyond their personal control, e.g. demographic
information which could include the neighbourhoods in which they live, or even
people with whom they may be in contact, for completely legitimate and lawful
reasons, but which indicts them as potential criminals or decides the same
because of actions of another, completely unrelated or unbeknownst to them.[7]
This undermines their right to a fair
trial, or even afforded a chance to defend themselves and defeat this
presumption of guilt through an impartial adjudicatory process, which
necessarily operate on a case by case basis, accounting for unique facts,
circumstances and other factors that are presented in each case. This
necessarily discards the value of human intellect, empathy, emotion, deprives
the defendant and adjudicator of ‘judicial discretion,’ and precludes value
judgements on that basis. This promotes and produces a one size fits all
solution in matters of guilt or innocence, rather than judging cases on
individual merit (supposing that there is a case to be made at all).
The ethical conundrum AI presents is this:
for want of human emotion, and judicial discretion to make value judgments as
needed, it is, by design equipped to provide only one answer or produce one set
outcome to every set of facts presented before it. The absolute logic
that guides its determinations (whether correct or erroneous) cannot be
controverted by any information of which its guiding and opaque algorithm is
programmed to process.
The problem is especially acute in
profiling and predictive assessments of personalities or communities to identify
offenders or probabalize criminal activity because crime data within which it
is populated and informed, supplied by law enforcement is often uncorroborated.
The irony that presents itself here
is that AI and predictive policing claims to harness scientific precision to
aid in the prevention of crime. Yet, uncorroborated data is the diametric
opposite the definition of an ‘unscientific methodology’ – unreliable, and ergo
unacceptable especially in what is now hailed as the latest frontier in
scientific and technological advancement. Taking the above factors into
account, is it possible to assert with confidence that such a system, it cannot
be denied that such a system is unbiased, or at the very least prone to making
irreparable errors.
Moreover, in this author’s opinion, it
lacks the capacity to determine the value of human life (e.g. to an AI
system, the value of all life is not something it can process, so a
teeeange outh growing up in a dangerous neighbourhood, even if he is not affiliated
to any criminal activities could be assessed to be a potential recidivist, only
because of factors beyond his control, comprehension or knowledge. Neither does
the predictive model possess the ability to determine the opportunity cost of
deprivation of freedom of life, liberty, and now, increasingly, the right to
privacy. This argument is buttressed by the fact that no intellect, artificial,
human, or ay other form present in the universe can ever predict an
individual’s potential for good, irrespective of his algorithmic profile’s
conclusions to the contrary. Thus, when automated systems fail, technology is
rendered immune from sanction by virtue of the authority that automation
provides it, and is unable or exonerated of any responsibility to indemnify the
opportunity cost it has caused – and unfairly places the blame on human members
of law enforcement and other branches of the justice system, making them stand
accused of professional incompetence.
These are serious threats that cannot
be easily justified or dismissed – because lacking any present, active, and
impartial application of the uniquely human intellect and its ability to draw
crucial distinctions, driven by experience and uniquely human emotional
capacities – compassion, empathy, real time understanding of unique
circumstances, ethics or morality, the usage of Artificial Intelligence can
never be said to be a substitute for human judgment or intellect.
c. Lack of Transparency and Avenues for Redressal
It is widely agreed that State
authorities owe a duty of transparency to the public in the performance of
their duties, being custodians of the public’s trust. Nowhere is transparency
more relevant and pertinent than in institutions of criminal justice and those
that influence their decisions them in any way. this more important than in the
criminal justice system, and any other system that exercises influence on it,
its administration, and decisions. However, in contemporary times, technology
is often the barrier to transparency, rather than a gateway for access to it.
Further, driven by profit motives, deliberate efforts have also been known to
conceal how these systems work, most likely because they are unwilling to
expose any flaws or open up any questionable practice, or developmental flaws
to public scrutiny, for fear of backlash.
Thus, people are left without the
knowledge that they have been judged by an automated system running on quite
possibly a flawed algorithm that has no human checks involved, and which lacks
any element of human discretion in the decision-making process. In this
context, criminal profiling and predictive assessments provide no clear or
avenues for righting ay wrongs, nor to challenge decisions themselves. It
follows that barring an effective method of challenge, avenues for redressal
are severely restricted or lacking altogether.
The end result is that ordinary,
innocent lives are put in jeopardy, especially when an erroneous profile leads
to the impossibility of securing a fair trial for having one’s guilt pre-determined,
based solely on an unaccountable, opaque, and automated logic. This is the perfect
anathema to Justice Oliver Wendell Holmes’ famous and legendary declaration
that the life of the law lies not in logic, but experience.
Conclusion
Given the relatively recent emergence
of AI into the legal system in such a big way, much of the iterature referred
to in this paper and the ideas presented therein are in large part a result of
the resolutions and recommendations made by the Council of Europe and several
collaborators in this project.
Much of what has been stated above is
a result of the dedicated efforts of the Report of the European Parliament’s Committee
on Civil Liberties, Justice and Home Affairs, where it exhaustively describes
the usefulness of AI in criminal justice, but at the same time lays down very
strong suggestions for how it must be harnessed and regulated for the
protection of human interests, rights, dignity, and that the ultimate
decisionmaker must always be a human, nd lys down framework as to how this might be
accomplished through cautious use of AI to safeguard against abrogation of
fundamental freedoms and inalienable human rights. It deals with the matter
exhaustively from the how it must be used by law enforcement as well as
judicial authorities, and that AI must always be subordinated to justice for
humanity.[8]
Given that the literature on the
subject is still emerging, and technologies still being perfected, regulatory
mechanisms still being developed (and being jurisdictionally bound), this paper
has not attempted to provide an exhaustive list of all the pros and cons of
predictive policing that claims to be the final word on the subject.
Rather, my main attempt has been to
emphasise the potential dangers of relying on machine learning to tackle crime
and identify potential safeguards against the same, while underscoring the danger
of over-reliance on AI in predictive policing specifically, and using its
conclusions as a substitute for human intelligence, experience and judgment. And
although predictive policing is only a very small part of the use of AI in
criminal justice delivery, it is nevertheless imperative that it only be used
as a reference to streamline enable the human decision maker’s task, wherever
it is deployed, and for whatever purpose it is used.
As
concluding statement, I believe it necessary once more to emphasise that
within the narrow scope of predictive risk assessment and policing in criminal
justice, algorithmic predictions must necessarily only be tool to aid the human
endeavour of crime prevention and criminal justice, rather than making
human intellect and effort its subordinate, and thereby rendering law
enforcement officials so reliant on AI as to disincentivize them from improving
their own expertise and sharpening the irreplaceable human skill that is
required to actually deter crime and address criminal activity in real time,
transparency in their operations and accountable for them, and provide
meaningful avenues for redressal of wrongs, and be prepared to provide
acceptable justifications when called on to do so.
[1] According to the National Justice
Institute’s literature, Artificial Intelligence is defined it as “the science
and engineering of making intelligent machines.” This definition was supposedly
coined by John McCarthy, credited as being the father of AI.
Christopher Rigano, “Using Artificial Intelligence to
Address Criminal Justice Needs,” available online at https://nij.ojp.gov/topics/articles/using-artificial-intelligence-address-criminal-justice-needs/
(Published on 10.08.2018), last visited on 01.05.2024.on 24.04.2024.
[2] Melissa Hamilton, “Predictive
Policing through Risk Assessment,” in Predictive Policing and Artificial
Intelligence at p. 60 (Routledge, 2021)
[3] As cited in Supra. Note 2.
[4] As cited in Supra. Note 2.
[5] “Artificial Intelligence” (AI),
data and Criminal Justice,” available online at https://www.fairtrials.org/campaigns/ai-algorithms-data/#what-are-the-problems-with-ai-83,
last visited on 02.05.2024.
[6] “Regulating Artificial Intelligence
for Use in Criminal Justice Systems in the EU” Policy Paper,” available at Regulating-Artificial-Intelligence-for-Use-in-Criminal-Justice-Systems-Fair-Trials.pdf
(fairtrials.org)
[7] Id.
[8] The Complete Report, officially titled
“REPORT on artificial intelligence in
criminal law and its use by the police and judicial authorities in criminal
matters | A9-0232/2021 | European Parliament (europa.eu),” dated 13.07.2021,
is available online at
https://www.europarl.europa.eu/doceo/document/A-9-2021-0232_EN.html#_ftn8, last
accessed on 28.04.2024.
Article Information
ALGORITHMS, ARTIFICIAL INTELLIGENCE & PREDICTIVE POLICING: A BRIEF COMMENT BY: AKSHAY SREEVATSA
Authors: AKSHAY SREEVATSA
- Journal IJLRA
- ISSN 2582-6433
- Published 2024/05/13
- Issue 7
About Journal
International Journal for Legal Research and Analysis
- Abbreviation IJLRA
- ISSN 2582-6433
- Access Open Access
- License CC 4.0
All research articles published in International Journal for Legal Research and Analysis are open access and available to read, download and share, subject to proper citation of the original work.
Disclaimer: The opinions expressed in this publication are those of the authors and do not necessarily reflect the views of International Journal for Legal Research and Analysis.