AI-Driven Crime Prevention: Balancing Predictive policing with Individual Rights

I. Introduction

Communities prioritize their safety, which is why governments take the required steps to lower crime rates. A crucial area of criminology is crime analysis, which looks at behavioral patterns and attempts to pinpoint the warning signs of such occurrences³. Nevertheless, there are a number of issues that come up while trying to prevent crime. This is a result of the wide range of crime kinds, causes, consequences, responses, and preventative strategies. Because of these complexities and other characteristics, crime prediction has developed into a potent and popular method. Because of this, police agencies invest a lot of effort and money on identifying and forecasting crime trends. Machine Learning approaches could lessen this effort by swiftly analyzing vast amounts of data to uncover crime trends, especially with the growing shift toward technology and improvements in artificial intelligence.⁴Numerous AI strategies have been extensively researched in order to lower or eliminate crime and guarantee public safety in various nations. These machine learning models have the potential to be utilized in the future to forecast crimes, their characteristics, etc. Police departments will be able to detect hotspots depending on time, kind, or any other element through machine learning, which will help them maximize their resources. Furthermore, examining criminal histories may provide additional insight into the sociological makeup of localities. Therefore, in order to avoid such issues, government agencies and decision-makers will be able to determine which age groups, ethnics, etc. to concentrate on.

Police departments have historically employed a method known as "hotspot analysis" to deter crimes. With this method, police officers can allocate more resources to these locations by simply uploading past offense and crime data as an overlay on a map. That being said, this tactic is a response to past events rather than a prediction. In contrast, police departments can find trends and make predictions about the future by using AI tools to analyze the datasets they have acquired. For instance, examined crime data from Vancouver for the last 15 years using such a dataset. They employed a heatmap to identify hotspots—areas where crime is most likely to occur—and several AI techniques, such as boosted decision trees with K-nearest neighbors.

The term "crime prediction" in this context relates to the application of mathematics to law enforcement, with predictive analytics being used to anticipate likely criminal activity in a given location. Based on certain characteristics of crimes that have happened in particular locations, this prediction analysis was conducted. These characteristics are numerous and could change over time based on factors including the kind of crime, where it occurs, and trends in crime. Since "crime" is a broad concept, it can take many distinct forms. This covers serious crimes like murder, kidnapping, and sexual assault as well as property crimes like shoplifting, theft, and burglaries. The frequency of crimes can change based on the day of the week, the hour of the day, and even the season. Geographical location is also important because there are locations that are more likely to experience crime than others. These various aspects are being taken into account when allocating police resources to various regions. Since these and many other parameters have a wide range of possible values, machine learning is the best data analysis technique to address this issue. The integration of AI in crime prevention, particularly through predictive policing, has transformed law enforcement strategies. However, this innovation raises significant ethical and legal concerns regarding individual rights, privacy, and potential biases. This paper will discuss the balance in effectiveness with the rights of the individuals that can be significantly affected by the use of the AI predictive algorithms in crime prediction such as Privacy rights, due process, Equal protection, Transparency.

II. Background

The past few decades have seen a considerable evolution in the use of AI in crime prevention and prediction, driven by shifts in cultural views toward law enforcement, technological improvements, and persistent worries about civil liberties and safety. Historically, to prevent and solve crimes, law enforcement depended on human intuition, experience, and fundamental statistical analysis. The main components of policing strategies were patrols and community involvement. Data analysis started to be used in policing in the late 20th century, especially in New York City with the introduction of CompStat in the 1990s. A police management system known as COMPSTAT, or Compare Stats, was developed in 1994 by the New York Police Department with support from the New York City Police Foundation. This method made use of crime statistics to guide resource distribution and deterrent tactics.

The data from several sources (such as social media, security cameras, and public records) has made it possible for law enforcement to examine enormous volumes of data and spot trends in criminal activity. Advances in AI and machine learning algorithms have made it possible to analyze data in increasingly complex ways. These technologies can reveal intricate connections in data that more conventional approaches would overlook.⁵ Through the analysis of past data and the identification of risk indicators, AI-powered predictive analytics technologies are able to forecast potential criminal activity, allowing police forces to more efficiently focus their efforts. The field at the nexus of artificial intelligence and crime prevention is one that is always changing due to societal demands, ethical considerations, and technological breakthroughs. Effective and just policing will depend on law enforcement agencies adopting these technologies in a balanced manner that takes individual rights and public safety into account. As stakeholders attempt to manage the intricacies and social ramifications of AI, the discussion over its application in crime prevention is still ongoing.

Regulations pertaining to the use of AI in law enforcement are being considered by policymakers more frequently in an effort to allay worries about accountability, prejudice, and privacy. The value of community involvement in the application of AI to law enforcement is becoming increasingly apparent, with an emphasis on trust-building and openness. This section will concentrate on preventative police strategies, such as using copious amounts of data to create general profiles that have the potential to forecast future criminal activity. Together with several drawbacks, like an over-reliance on relative probability, a lack of profile specificity, and the possibility for erroneous data to taint forecasts, it will describe how these systems function in context.

(A) Literature Review:

Mrs Aafrin Gouri, (2024) suggests that criminologists should play a more active role in implementing AI in crime prediction and prevention strategies, proceeding away from traditional statistical models and toward more advanced AI-driven approaches. The current study investigates the ability of AI in crime forecasting by comparing four AI techniques—ANN, SVR, RF, and GTB—with traditional statistical methods such as linear regression and ARIMA.⁶ The study uses a regression approach to provide a more nuanced analysis of crime data, with the goal of improving predictive accuracy and lowering error rates. Fatima Dakalbab, Manar Abu Talib,⁷(2022) paper aims to study the research attempts in predicting the location and the time of the crime as it can help in optimizing the police department's efforts and resources. The research was done appropriately, and articles were selected concerning (i) the primary crime prediction research work done, (ii) the techniques utilized in crime prediction, (iii) the estimation and performance metrics of the proposed models, (iv) the limitation and strength of proposed models. Dario Ortega Anderez, David Lucy,⁸(2021) This paper provides a discussion of how a sample of contemporary hardware and software-based technologies might help further reduce criminal actions. After a thorough analysis of a wide array of technologies and a number of workshops with organisations of interest, we believe that the adoption of novel technologies by vulnerable individuals, victim support organisations and law enforcement can help reduce the occurrence of criminal activity. Richard .A. Berk,⁹(2021) In this review, we address these issues by first unpacking depictions of artificial intelligence. Its use in predictive policing to forecast crimes in time and space is largely an exercise in spatial statistics that in principle can make policing more effective and more surgical. Its use in criminal justice risk assessment to forecast who will commit crimes is largely an exercise in adaptive, nonparametric regression. Lindsey Barrett, This paper will argue that the use of predictive policing algorithms at the border should not be banned outright, as the government should permit potentially beneficial uses of the technology to develop. However, use of these algorithms should be carefully limited by statute to prevent the wholesale trammeling of privacy and civil liberties.

(B) Research Objective:

• To study the use of artificial intelligence (AI) in crime prediction.
• To study the impact of AI based predictive policing on individuals’ privacy rights, due process rights.
• To study the challenges faced by predictive policing.

(C) Research Problem:

As law enforcement increasingly relies on AI for crime prediction and prevention, it is essential to examine the ethical implications of the technologies. Our research will investigate how AI driven crime prevention affects and balances the protection of individuals’ rights, due process rights by predictive policing.

(D) Research Questions:

How might the use of predictive policing algorithms influence public perceptions of transparency in law enforcement, regarding personal privacy?

In what ways could predictive policing tools affect the balance between enhancing law enforcement effectiveness and maintaining individuals' rights to privacy and fair treatment?

III. Using ai to prevent crimes

These days, forensic technology can be applied to prevent crime as well as criminal investigations. In recent years, preventing crime has taken center stage in policing, as it is deemed more effective than attending to individual cries for help, looking into crimes, and enforcing punishment after the fact. It is possible to generate significant efficiencies and improve public safety by attempting to prevent crime before it starts. Furthermore, AI makes it possible to advance criminology theories that identify the causes of criminal activity and call for precision policing in addition to more conventional reactive measures. This could be well described as the "criminalization of social policy," in which law-abiding behaviors are viewed as either possibilities for crime to occur or as barriers to it.¹⁰ Preventive measures as a means of controlling crime are thus a type of risk management that prioritizes the preceding act over the result and is associated with a rise in the criminalization of inchoate acts.¹¹ For instance, a spate of burglaries in a neighborhood may be a sign that there are several means of egress, insufficient lighting, and other characteristics that make the region more susceptible to crime. Mitigating the key factors, such as installing sensored illumination, would be one way to find a solution. Increased police presence in this neighborhood after dark could be another way to address the problem. The number of relevant components and the complexity of their interactions may be exponential, despite the fact that they are straightforward theoretically.

Because risk assessment is a fast and methodical way to analyze a wide range of data sets, it is a powerful tool for policing that may be used to establish suspicion, identify cause, and evaluate evidence. In essence, the program instructs cops to patrol areas which are considered at higher risk for future crime according to the outcome of the risk assessment¹². Thus, there is an expectation that, for example, future crimes and offenders would resemble those from the past, and that neighborhoods with a correlation between similar elements may be given priority as being probabilistically more likely to experience crime. "ShotSpotter" is one instance of a risk assessment tool used in predictive policing.In order to prevent overpolicing of these locations and to identify patrols that are judged effective in deterring crime, the program also monitors and logs officer contacts in these regions. When a jurisdiction uses a program like ShotSpotter, police patrols might be arranged according to the results of risk assessments.

ShotSpotter. (2018, October 3). ShotSpotter announces acquisition of HunchLab to springboard into AI-driven analysis and predictive policing.

According to the probability listed in the risk assessment, the believed would-be thief is therefore treated more closely to a suspect than the person acting identically elsewhere if the officer approaches or stops them because they believe that their behavior fits the profile of a thief based on the time and location they are encountered. This alters the officer's discretion and the public's perception of the average citizen, among many other consequences in the event that the risk assessment is flawed or based on false information.

Consequently, one of the most often voiced objections to predictive policing is that an excessive dependence on police data may lead to prejudice, inaccuracy, and inconsistent risk assessment. The decision-making processes that guide police activities, the records that follow, and the people who are ultimately detained or accused are all subject to considerable discretion, as has been said. Any ingrained prejudices that may exist in the community are reflected in the data captured, even in risk assessment systems that exclusively use the records of arrests and charges brought about by citizen requests for aid. Similarly, many crimes go unreported for a multitude of reasons and, consequently, the data provided to the risk assessment tools do not fully reflect reality, creating outcomes that target some types of crime above others. Thus, it is evident that accurate or inaccurate data are not always implied by correct or incorrect data.

The use of risk assessments raises additional concerns because of its reliance on machine learning and artificial intelligence in particular. Artificial intelligence is a branch of research that studies how linkages and correlations between combined categories might produce new knowledge or allow existing data to be evaluated algorithmically. This new generation of data may be used to generate other generations of data by continuously upgrading the algorithms in machine learning, a subcategory of artificial intelligence. This effectively mimics the capacity of an intelligent entity to learn by inference. The issue with machine learning, however, is "the black box," or the possible lack of transparency built into its workings. The original algorithmic source code, all of the data that was fed into the system, and the output can all be identified. Consequently, the connection between citizens and police is altered when prevention is used as the main method of crime management since interactions are now informed by expectations rather than just observations and knowledge. Whereas preventative policing brings into question the examination of persons where a crime has not been committed, establishing suspicion on the basis of broad profiles generated by algorithm begs problems of the lawful application of discretion and due process. Preventative profiles counter the idea of individualized suspicion by serving as stand-ins for real circumstances pertinent to a potential crime. The use of preventive profiling to deter crime will be examined in the ensuing sections, where it will be argued that the criterion of reasonable suspicion is compromised by the impact on officer discretion. The enjoyment of some fundamental rights may therefore be diminished.

IV. Policing: between discretion and due process

The terms ‘police’ and policing’ have carried many meanings over time, their applications being neither synonymous nor mutually exclusive. Both terminologies will be used in the article, and the points made will relate to the official work that police do in the course of policing. In their current incarnation, the police are recognized as the authority charged with upholding law and order and guaranteeing public safety, distributed among several departments with distinct functions. There are many different types of security, from keeping an eye out for moving violations and fostering peace to looking into violent crimes. Police have the authority to apply penalties to individuals in order to uphold the law, but they also have the right, when necessary, to employ physical coercion. Policing is shared by police with commercial and other public sector entities and may be defined as “the act of upholding the law, maintaining order, preventing crime and offering a range of support services to the public”. Examples include the work done by public health organizations or the regulatory role of organizations in charge of monitoring businesses and financial institutions. Laws, rules, and administrative codes, both civil and criminal, provide the legal foundation for law enforcement. These legal tools typically offer due process protections or police accountability systems that guarantee a balance between the power of law enforcement and the rights of persons.

The wide range of tasks and actors involved in the act of policing and the management of the police as an organization also generates a need for increased efficiency.¹³ Discretion is a traditional tool for increasing police efficiency. With the use of discretion, an officer can quickly and efficiently make decisions by using a descriptive, visual analysis in conjunction with experience to evaluate a situation holistically. As a result, police have a great deal of discretion in deciding which actions call for the filing of criminal charges or other legal repercussions.

In actual fact, there seems to be conflict between this duality between police discretion and due process. While security requires swift action, the rights of the innocent to the rule of law are as important. In contrast to reactive methods of crime control, crime prevention occurs outside of the official legal system, i.e., before a charge is filed, which further muddies the application of due process. ¹⁴To differentiate police power from state violence, the use of police authority as a starting point for the criminal procedure must nonetheless adhere to due process. The interaction between police and residents in a particular situation may become more distant as a result of the addition of predictive policing assessments to discretion, as is stated below, perhaps exacerbating the friction and separating policing from due process requirements.

V. Institutionalizing discrimination

Algorithms for predictive policing justify discrimination by disguising it as quantitative analysis. Predictive policing algorithms use historical criminal data, which is an indication of who is more policed rather than necessarily a reflection of who is more likely to commit a crime. Predictive policing technologies also exhibit such discriminatory tendencies. American mathematicians have pushed their colleagues to give up on predictive policing systems because they think it will only serve to reinforce systemic prejudice. Because of systemic racism in the enforcement system, African-Americans in particular are disproportionately policed in the United States compared to White individuals.

The predictive police algorithms employ the racial bias in the crime statistics as a result. Due to this, there is a discriminatory feedback loop created whereby an increased level of policing of a particular group makes the algorithm more likely to perceive a member of that group as a potential criminal.¹⁵

Empirical studies indicate that Delhi's CMAPS is experiencing a comparable phenomenon. They clarify that the algorithm does little more than bolster the police officers' prejudices. Consequently, the algorithm develops a discriminatory feedback loop since police targeting is disproportionately likely to occur in places where caste and religious minority make up the majority. This is consistent with findings from research funded by the UK government, which shows that police officers are more likely to make biased arrests without probable cause when they are stationed in "crime hotspots" that have been identified by data analytics.

VI. Challenges around predictive policing

The conflicting results of the studies suggest that predictive policing, as an emerging crime-fighting instrument, remains under scrutiny and requires further careful academic investigation¹⁶. The technology has generated considerable doubt and suspicion both within and beyond academia since its inception. This section examines the primary challenges surrounding the application of predictive police technologies in crime prevention¹⁷.

(A) Violating the Right to Privacy:

The impact that predictive policing algorithms have on the right to privacy, given that they employ enormous amounts of personal data. According to Justice K.S. Puttaswamy (Retd.) v. Union of India, there is a violation of the right to privacy when there is opacity about the use of personal data. Additionally, the historic ruling created the "proportionality and legitimacy test," which outlined four requirements that must be met in order for the government to violate a person's right to privacy: i. The activity needs to be authorized by law. ii. In a democratic society, the proposed action must be required to achieve a justifiable goal. iii. The degree of interference must be commensurate with the necessity of the interference. iv. Procedural safeguards against misuse of this kind of intervention are necessary.

Furthermore, it is impossible to determine the precise workings of predictive policing algorithms because law enforcement agencies are excluded from disclosure requirements under the Right To Information Act. Many people believe that predictive policing is just state monitoring disguising itself as internal security because of its extreme opacity These worries are justified because it is anticipated that personal information about Indian citizens, such as bank account details, will be accessible to the country's primary intelligence database, the National Intelligence Grid (NATGRID). In light of the hacking of the Maharashtra Criminal Investigation Department website last year, also worried about the security of the databases holding personal information.¹⁸

(B) Transparency:

In the age of big data policing, the second controversy is about police transparency and is strongly tied to privacy concerns. This problem arises throughout the whole predictive policing deployment process, from data collection to the post-action phase. In many areas, the predictive algorithms have been kept a proprietary secret for years. There are major ramifications for both police legitimacy and community relations with this so-called "algorithmic secrecy." As stakeholders, citizens argue that they have paid taxes and given their personally identifiable information (PII) only to be denied access to information about how that data is being utilized.

Citizens' frustration has grown to the point where some of them, along with advocacy groups and organizations, have even filed lawsuits against police departments that have used predictive policing technologies without disclosing the specifics of the algorithms (e.g., Brennan Center for Justice v. New York Police Department; Smith, Joseph, Kalven, Chicago Sun-Times v. Chicago Police Department). However, police departments contend that disclosing information related to predictive policing will: (1) reveal trade secrets of vendors, breach non-disclosure agreements they have signed, and ultimately jeopardize their relationship with those vendors and others; and (2) undermine predictive policing initiatives, giving potential offenders the ability to “somehow game the system” and “anticipate and thwart police response strategies, exposing officers and the public to risk of harm,” among other things. Transparency is a problem that affects predictive policing not just during the data collecting phase but also during the forecast, action, and post-action phases. Transparency guarantees accountability and fosters community trust in the police during the prediction and action phases. When racial profiling is brought up during the post-action phase, it is critical to locate "smoking gun" direct evidence to support or refute the existence of discriminatory intent and/or result in the decision-making process¹⁹. Thus, transparency becomes essential for jurors and judges as well as complainants and accused police

(C) False Positives and False Negatives:

There is always a chance of false positives and false negatives during the prediction stage, even if we assume perfect, unbiased data input and an impartial predictive algorithm. A false positive in predictive policing is an error whereby: (1) a person predicted to be a victim or perpetrator of crime turns out not to be one; or (2) a location predicted to experience crime in the future turns out not to experience crime in the anticipated time period.²⁰ Conversely, a false negative is a mistake that occurs when either (1) a person who was not expected to be a victim or a perpetrator of crime turns out to be one, or (2) a location where no crime is expected in the future turns out to experience crime within the expected time period.

False-positive and false-negative forecasts will squander police resources that may be better used on actual criminals or crime scenes in all of the aforementioned instances. Optimizing resource allocation is one of the main goals of predictive policing, as it increases the police's ability to thwart criminal possibilities. The fundamental purpose of using predictive policing is challenged by false positives and false negatives, which could also have a detrimental effect on officers' trust in the system.

(D) Admissibility of Evidence:

The debate over the admissibility of evidence in court invariably arises from the question of probable cause and reasonable suspicion, as was demonstrated in Terry v. Ohio (1968). This problem arises during predictive policing's post-action phase and has a significant and broad influence on how criminal cases turn out as well as the law. The weapon that the police officer had taken from the suspect (the petitioner in this case) was "properly admitted into evidence against him, since the search which led to its seizure was reasonable under the Fourth Amendment," the Supreme Court said in Terry v. Ohio. Applying this approach to predictive policing raises the question of whether evidence discovered after a search and seizure carried out under the orders of predictive police software should be accepted as admissible.

(E) Accountability and Liability:

Determining accountability when AI systems lead to wrongful arrests or other legal violations poses significant challenges. If law enforcement agencies rely on AI tools that produce biased or inaccurate results, questions arise about who is responsible for these outcomes—whether it's the developers of the software, the police departments using it, or the government entities overseeing its implementation²¹. Establishing clear legal frameworks that delineate responsibilities is crucial for addressing these accountability issues.²²

(F) Scope and limitation:

The purpose of this study is to look at how community relations and law enforcement procedures may be affected by AI-driven predictive policing. It will look at the effects of AI integration on officer discretion, due process rights, and the possibility of bias in police decisions. The paper will evaluate the effects on fairness, and individual rights by examining the ethical issues and accountability methods required for responsible AI use. In the end, the study aims to shed light on how to reconcile the need to protect fundamental rights in the use of predictive policing with technical improvements.

This research exclusively examined the theological perspective. Because this paper only consults secondary sources, the perspective on artificial intelligence may be constrained. Using other study approaches, future researchers can expand the scope.

(G) Research Method:

This paper utilises Doctrinal research methodology, a tool used in legal research, to explore and support a thorough understanding of the topic. The author uses secondary data sources to guide the research objectives and inquiries. The analysis of the study is descriptive in nature. It seeks to identify numerous aspects and qualities. Data for the study were gathered from many sources, including government publication, journal articles, websites and reports.

VII. Findings

Using AI technologies, like facial recognition software, frequently necessitates gathering a lot of data, including biometric information. People's rights to privacy and the protection of their personal information may be violated by this activity. There are serious worries about surveillance overreach and the possible exploitation of personal data in countries like India, where the government has installed facial recognition technology without strict safeguards. Addressing these accountability concerns requires the establishment of precise legal frameworks that specify obligations.

VIII. Recommendation

It is possible to establish the following three primary policy recommendations for the application of predictive policing: (1) police reaction strategy, (2) clear communication between various police units and hierarchy levels, and (3) dependable data collecting. First and foremost, it's critical to offer a trustworthy data gathering. After all, the final quality of the risk projections will be greatly influenced by the quality of the data. For predictive policing to be implemented successfully, a police department's data collection procedure and data quality must be assessed. Analyzing how problems with data quality can affect prediction performance might also be a good idea. Conduct routine evaluations of AI systems to make sure they adhere to moral principles and represent the ideals of the community. Put strategies in place to recognize and lessen algorithmic biases. Assess algorithms for fairness on a regular basis and make any required adjustments. To reduce bias, make sure the data used to train AI models is representative of the population and varied. Only gather the information required for particular crime prevention initiatives. Steer clear of data collecting that violates people's privacy or excessive surveillance. Increasing openness and equity will boost public confidence and guarantee that AI systems support public safety campaigns.

IX. Conclusion

By using AI into predictive policing, law enforcement will be able to protect communities more effectively and pro-actively, revolutionizing crime prevention tactics. Law enforcement organizations can use data insights to anticipate crime trends, allocate resources optimally, and make data-driven choices by utilizing AI algorithms. To guarantee that the use of AI in predictive policing is consistent with the values of justice, fairness, and privacy, however, ethical issues, accountability, and openness are essential. The ethical and appropriate application of AI in crime prevention will become more and more crucial as technology develops, contributing to the development of safer and more secure communities.

This paper has argued that predictive policing changes crime control and the citizen-officer relationship, even though modern technology cannot be definitively labelled as incompatible with criminal justice. Programs like risk evaluations have an impact on the exercise of true officer discretion even though they theoretically operate impartially. The absence of context on which an officer would base their discretion affects the establishment of reasonable suspicion, even in cases where risk assessments using perfect information are used to forecast crime. When the data is subjected to algorithmic processes, people who have interacted with police in any capacity, regardless of the outcome of the interaction, may be deemed more likely to commit a crime in the future where past crimes are typically inferred from data of past stops or arrests. When police use the derived information—a risk assessment—to justify stops or arrests, they are probably going to breach the presumption of innocence. This kind of discretionary use is not only modified, but also possibly predicated on inaccurate data. It is obvious that changing the physical and informational environment in which discretion functions also modifies the system that safeguards fundamental rights. Therefore, serious consideration must also be given to the invisible but cumulative consequences on group and individual rights when discussing the fairness and efficacy of predictive policing in preventing crime.

*****

Footnotes

1. Author is a student at SASTRA Deemed to Be University, India.

2. Author is a student at SASTRA Deemed to Be University, India.

3. Crimecast: A crime prediction and strategy direction service, Nafiz Mahmud, 2016. ↩

4. Big data analytics and mining for crime data analysis, visualization and prediction, Mingche Feng, Jiangbin Zheng, 2018. ↩

5. Spatio- Temporal crime hotspot detection and prediction: A Systematic Literature Review, Umair Muneer Butt, Sukumar Letchmunan, 2020. ↩

6. AI in crime prevention and prediction, Mrs Aafrin Gouri, 2024. ↩

7. Artificial Intelligence and crime prediction: A systematic literature review, 2022. ↩

8. The Rise Of Technology In Crime Prevention: Opportunities, Challenges And Practitioners’ Perspectives, Dario Ortega Anderez, 2021. ↩

9. Artificial Intelligence, Predictive Policing, and Risk Assessment for law enforcement,Richard A. Berk, 2021. ↩

10. A comparative study to evaluate filtering methods for crime data feature selection, Abdul Jalil, 2017. ↩

11. Predictive Policing Explained, Lau, T., 2020. ↩

12. Predictive policing: the role of crime forecasting in law enforcement operations. Perry WL, 2013. ↩

13. To predict and serve,Kristian Lum, William Isaac, 2016. ↩

14. Reasonably Suspicious Algorithms: Predictive policing at the United States border, Lindsey barrett. ↩

15. ways to reduce crime in the world’s most violent cities. Larsson, N., 2015. ↩

16. A survey of research into artificial neural networks for crime prediction, Forkan Albo, 2021. ↩

17. Windows into the soul: Surveillance and society in an age of high technology.Marx, G. T., 2016. ↩

18. Maharashtra CID website hacked, defaced.Mengle, G., 2020. ↩

19. Preventing crime before it happens: How data is helping Delhi Police. Hindustan Times, Singh, K., 2020. ↩

20. The right to be presumed innocent. Crim Law. Stewart H, 2014. ↩

21. Data in New Delhi’s Predictive Policing System. Marda, V. and Narayan, S., 2020. In: FAT* 20′: Conference on Fairness, Accountability, and Transparency. ↩

22. AI in Police Work, Dr. Kamal Kishore Singh, IPS. ↩