GREAT BAY--Amnesty International Netherlands has called on the Dutch municipality of Heerlen to abandon an artificial intelligence project designed to help predict crime, nuisance and public safety risks, warning that predictive policing can reinforce discrimination and lead to human rights violations.
The project, known as PULSE-TWIN, is developing what authorities describe as a Social Digital Twin of Heerlen. The virtual model combines information about the physical city with social and safety data, allowing artificial intelligence to identify patterns and make predictions that could support decisions about neighborhood safety and quality of life.
PULSE-TWIN is being developed by the Municipality of Heerlen together with partners including the Dutch national police, Maastricht University and other organizations. The project is supported through the European Urban Initiative and is intended to test whether technology and community data can be combined to improve urban safety.
The initiative focuses heavily on Heerlen-Noord, where more than 40 percent of residents reportedly feel unsafe, compared with a national average of about 14 percent. Authorities say the project is intended to help address persistent concerns involving safety, quality of life and social conditions in the area.
Under the project, AI is expected to analyze information from official sources, professionals and residents to identify patterns and predict possible developments. The system is intended to support decision-making by local authorities and professionals working on neighborhood safety and liveability.
That predictive element is at the center of Amnesty International’s objection.
Alexander Laufer, a technology and human rights researcher with Amnesty International Netherlands, has said the organization does not question Heerlen’s stated intention to improve public safety. Amnesty argues, however, that governments have a special responsibility to protect human rights and that the widespread use of AI in other areas does not justify using it to predict criminal activity.
Amnesty maintains that predictive policing systems can create a self-reinforcing cycle. If authorities identify a particular neighborhood as higher risk, that area may receive greater surveillance and enforcement attention. Increased scrutiny can then produce more recorded incidents and more data from the same area, reinforcing the original assessment while activity elsewhere receives less attention.
The organization has also warned that automated risk profiling can reproduce existing biases in policing data and disproportionately affect lower-income communities and minority groups. Amnesty says predictive policing can raise serious concerns involving privacy, discrimination and ethnic profiling.
Heerlen, however, maintains that PULSE-TWIN is being developed with safeguards. Privacy-by-design principles are being incorporated into the project, Maastricht University is involved in its development, an ethics committee is monitoring the initiative and the municipality’s data protection officer is participating. A Data Protection Impact Assessment is also planned.
The municipality has stressed that the project is a pilot and that applications found to be ineffective or irresponsible will not automatically continue. Officials have said the initiative is intended to test whether AI can responsibly support efforts to make neighborhoods safer and more liveable.
That assurance has not satisfied Amnesty. Laufer has argued that a pilot still involves real communities and that residents can be affected by decisions made on the basis of predictions, even if the system does not rely on individual personal data.
The project’s scope also extends beyond conventional policing. PULSE-TWIN is intended to combine social and spatial information to help professionals make decisions about neighborhood liveability, public safety and community interventions. Other European cities are also involved as transfer partners examining whether elements of the project could be adapted elsewhere.
The dispute highlights a wider challenge facing governments as artificial intelligence becomes more deeply integrated into public administration. At issue is whether systems designed to identify patterns and forecast risks can be used without turning historical inequalities, incomplete data or existing enforcement patterns into automated predictions about communities.
For Heerlen, PULSE-TWIN is being presented as an experiment in using technology to support safer neighborhoods. For Amnesty International, the concern goes further, raising the question of whether predictive technology should be used in public safety and law enforcement at all when inaccurate or biased predictions can directly affect the people living in the communities being analyzed.
