Predictive policing is a technology-driven approach that uses algorithms and big data to predict and prevent crime. Although it sounds like an efficient way to enhance security, it also raises fundamental questions about privacy, discrimination and feasibility. In this blog, we delve into how predictive policing works, how it is applied in the Netherlands, and where the ethical and legal boundaries lie.

What is predictive policing?

Predictive policing uses historical data, such as crime figures, location information and even personal characteristics, to predict where en when crime is likely to occur. There are two main forms:

• Location-based forecastssuch as the Dutch CAS system (Crime Anticipation System), which identifies risk areas based on previous incidents.
• Personalised predictionswhere individuals are analysed based on risk factors such as previous arrests, behavioural patterns or even medical records.

How does it work in the Netherlands?

Since 2014, the Dutch police have been experimenting with predictive policing, including the CAS system and the controversial Sensing Project in Roermond. In this latest project, vehicles were scored based on characteristics such as country of origin and destination, which led to mass surveillance and ethnic profiling

The ethical and legal boundaries

Although predictive policing has potential, there are serious objections:

1. Privacy and data use

The processing of sensitive personal data, such as ethnicity or medical history, is at odds with privacy legislation. The GDPR (General Data Protection Regulation) offers protection, but is not always sufficient for group-targeted analyses

2. Discrimination and bias

Algorithms are not neutral. They are fed by data that are often already biased. This can lead to discriminatory risk profiles, especially for vulnerable groups who are overrepresented in police data

3. Transparency and accountability

Many AI systems are self-learning and opaque. It is often not clear why a system makes a certain prediction. This makes it difficult to audit or challenge decisions

4. Feasibility

Even if a system identifies dozens of individuals at risk, it is practically impossible to respond to that effectively. Police capacity and operational feasibility are often overlooked

What does the law say?

The Netherlands does not yet have specific legislation for predictive policing. However, organisations such as Amnesty International are calling for:

• One binding human rights test when using algorithms.
• One ban on the use of ethnicity and nationality in risk profiles.
• One independent algorithm regulator with binding advice.

Conclusion: Where does one draw the line?

Predictive policing can contribute to a safer society, but only if it transparent, accountable and people-centred is deployed. The line is drawn where technology begins to undermine the rights and dignity of citizens. It is up to policymakers, researchers and citizens to strictly guard that line.