Ethics of Smart City Predictive Policing

Spurred by the ethics of utilitarianism, smart cities have risen all around the world in an attempt to transform the ideal vision of urban life into reality. Even though the implementation of technology-driven solutions has bolstered the security sector to become significantly stronger and more efficient, ethical concerns have appropriately surfaced. The perfect example is predictive policing, a form of policing that uses statistical predictions and quantitative techniques to predict high-frequency areas for crime. Centered around AI algorithms and historical data, predictive policing seeks to balance the benefits of increased safety with the drawbacks of social inequality. As a result, it is often at risk of falling into a cycle that perpetuates policing prejudices and exacerbates implicit biases. This struggle to find a common ground between humanistic values and data-driven urbanism presents the question of whether this “sweet spot” really exists. However, through careful consideration of all aspects, an intersection between security and ethics can be found. Specifically, satisfying ethical standards in smart city predictive policing requires prioritizing transparency over the efficiency of streamlining operations that could potentially perpetuate racial biases in the criminal justice system.

Smart cities are modern cities that emphasize technological solutions and services to municipal problems. As the development of these areas pushes society towards a more sustainable and adaptive future, traditional forms of policing are being replaced by AI computer algorithms and advanced data analysis at an increasing rate; predictive policing serves as the catalyst for this progression towards increased safety through digital intelligence. Precisely, this approach entails improving tactical strategies by using methods such as advanced hot spot identification models or risk terrain analysis. These advancements allow law enforcement to more effectively deploy resources in criminal areas, conserving both money and time. Prior to predictive policing, historical data on criminal activity was collected by pooling data from law enforcement agencies. Since it is not uncommon for police officers to hold biases and preconceived notions regarding certain demographics, this affects how, where, and who law enforcement targets while being unaware of their tendencies. Years and years of data that have been compiled only strengthen the presence of these implicit biases, and these biases will be reinforced and amplified when fed into predictive policing algorithms..

Due to the inherent prejudices in city data, crime is often represented inaccurately, leading to a misleading reputation and flawed understanding of certain neighborhoods that could result in the unjust over-policing of targeted demographics. Law enforcement uses large amounts of criminal data to gauge the general safety of a given area, rendering human input essential to the process of predictive policing. Due to humans being at the forefront of policing for centuries, implicit biases are guaranteed to be embedded within. This unreliable data commonly results from “periods when the police are engaged in discriminatory practices against certain communities, thereby unnecessarily or incorrectly classifying areas as ‘high risk’” [1]. Clearly, specific criminal populations are not protected from being overrepresented because of police’s tendency to target the same areas, with the lack of a standardized data collection method being responsible for this critical flaw. Not only do racial and socioeconomic prejudices within law enforcement manifest themselves in criminal data, but the societal expectation for what crime typically consists of also results in a selection bias that skews the data. There is a general focus on “street crime,” which is primarily made up of offenses like theft and drug-trafficking. This type of coverage is severely unrepresentative of the array of crimes in society, and hosts the dangerous potential of misinforming the public on overall city safety. For instance, specific areas are thought to be more unsafe than in reality, while other areas escape this spotlight. The inability of criminal data to capture all inherent nuances in crime gives rise to biases. Avoiding targeting non-stereotypical criminal types and neighborhoods only results in ethical issues when this data is fed into a predictive policing system. As a result, glossing over these limitations may enable predictive policing to become a tool more harmful than beneficial to our democracy.

The predictive policing seesaw between enhanced safety and reinforced stereotypes is one that struggles to find an ethical equilibrium point. With decades of criminal data compiled in metropolitan areas, it can be easily assumed that they are unrepresentative of the true distribution of crime. The circulation of these biases within an AI algorithmic system has the potential to magnify previous discriminatory practices to a much larger scale, creating problems that could jeopardize the fundamental liberties that form the basis of our democracy. However, transparency and accountability measures can be taken through increased community engagement and reforms in criminal data collection, marking an increase in democratic participation that may pose significant benefits in enacting predictive policing in our society. As we become a more technologically-based and information-dense world, new approaches to advance efficiency are bound to emerge. Thus, instead of opposing and criticizing the shortcomings of predictive policing, communities should address them head on and readily adapt to mitigate its disadvantages.

References

1. Ghimire, Mukesh. “Ethics, Artificial Intelligence and Predictive Policing.” The Security Distillery, The Security Distillery, 23 July 2021, https://thesecuritydistillery.org/all-articles/ethics-artificial-intelligence-and-predictive-policing.