This project focused on the use of algorithms in relation to the police and public security. I specifically focused on the Predictive Policing system that is used in The Netherlands known as CAS (Criminaliteits Anticipatie Systeem). This system, that is built by the Police, uses data from recorded criminal activity and locational information to predict when and where a crime is likely to occur. The predictions manifest themselves as 125m2 hotspots. Once one of these spaces is considered a hotspot, the police will pay closer attention to it.
Whilst in theory the system should work, the outcome is far from perfect. The data used by the police contains all the biases that the police are prone to which includes discrimination against minorities and working class communities. The predictive system only amplifies these biases by solidifying them in a form that seems to represent a totality. This vision of the future is used and trusted by the police yet can lead to increased policing in already over policed areas and feedback loops when crime is found in hotspot areas yet not in others.
The video uses a method I created for visualising what the predictive policing machine might see if it were to have vision. Using data that was scraped from the Police's website, I determined which neighbourhood in The Hague which they believed to have the highest rate of criminal offences. Then, using all available google street view images in this area I trained a Generative Adversarial Network (machine learning algorithm) to recognise elements of the landscape. I then asked it to recreate the architecture of The Hague using the images of this particular neighbourhood.
The video was shown at the Violent Patterns Symposium.
An essay, with the same title, acts as an accompaniment to the video which will soon be found here.