By Maša Galič, Abhijit Das and Marc Schuilenburg

There is great enthusiasm for the use of Artificial Intelligence (AI) in the criminal justice domain in the Netherlands. This enthusiasm is connected to a strong belief – at least on the side of the government – that experimenting with new technologies can enhance security as well as improve government efficiency. New digital systems are considered as leading to rational, scientific and value-neutral ways to generate knowledge and expertise within the criminal justice domain. AI in this domain therefore holds a central position not only in policy documents,3 but can also be seen in numerous examples in practice. The Dutch police stand at the forefront of predictive policing practices, at least in Europe, being the first to deploy an AI-based system for predictive policing nation-wide, and continue to set up an increasing number of predictive policing projects. Facial recognition technology is increasingly used in public space, both by the police and municipalities, often in public-private partnerships constituted within smart city initiatives. And AI-based systems, such as Hansken, are used for the purpose of finding evidence among huge amounts of data gathered in contemporary criminal investigations.6 It should be noted, however, that in the Dutch public sector the term AI is oftentimes used in a broad manner, including algorithmic systems of various complexity. The term AI is used not only for data-driven algorithms (where algorithms are trained on the basis of input data) or rule-based algorithms (where the steps, methodologies and outcomes can be traced to pre-programmed instructions implemented by a human), but also for older and much simpler types of statistical analysis (e.g., actuarial risk assessment tools, which are based on the correlation between certain factors and past statistics concerning recidivism). Because of this broad use of the term AI and a lack of publicly available information on the functioning of many technological systems used in practice, it is sometimes difficult to know whether the system used in the criminal justice domain is strictly speaking AI-based or not. In any case, older methods for statistical analysis should be seen as a precursor of contemporary advanced AI techniques. The development of risk assessment technology, such as predictive policing and tools used for the assessment of the risk of recidivism, is namely taking place on a continuum, where several generations can be discerned.

e-Revue Internationale de Droit Pénal .2023. 57p.