2.3: Case Study- Predictive Policing in the US

Predictive policing is the use of data analysis, machine learning, and artificial intelligence (AI) to predict where crimes are most likely to occur and who is most likely to commit them.

It is used by law enforcement agencies to allocate resources and personnel, identify potential criminal suspects, and prevent crime before it happens. However, there are concerns about the potential for bias and discrimination in predictive policing algorithms, as well as questions about the legality and ethics of using AI to predict criminal behaviour.

Critics also argue that predictive policing can reinforce existing biases and inequalities in the criminal justice system, leading to unjust and discriminatory outcomes.

This is because the datasets often include biases which are a product of systemic racism, including police mugshot databases with an inordinate amount of black people and people of colour.

In August 2016, a coalition of 17 organisations, including the American Civil Liberties Union (ACLU), issued a statement expressing concern about predictive policing tools used by law enforcement in the United States. The statement highlighted the technology’s racial biases, lack of transparency, and other flaws that lead to injustice, particularly for people of color.

The statement called for transparency about predictive policing systems, evaluation of their short- and long-term effects, monitoring of their racial impact, and the use of data-driven approaches to address police misconduct. The statement also emphasised the importance of community needs and the potential of social services interventions to address problems for at-risk individuals and communities before crimes occur.

Facial recognition technology poses special risks of disparate impact for historically marginalised communities, such as black individuals who are more likely to be stopped by police officers and are overrepresented in law enforcement databases. Recent studies demonstrate that these technical inaccuracies are systemic and biased against people with darker skin.

Companies have announced actions to improve the accuracy of their facial recognition algorithms and diversity of their training datasets, but the scope and effectiveness of such efforts vary across vendors.

There remains an ethical question of if or when it is appropriate to use facial recognition to address legitimate security concerns, regardless of its accuracy. Guardrails are needed to ensure more equitable use of enhanced surveillance technologies, including facial recognition.