The application of artificial intelligence (AI) is a development that is considered rather ambivalent. On the one hand, it gives tremendous opportunities to achieve more efficient and effective results by analyzing large amounts of different data ("big data") using AI. On the other hand, there is a fear that the use of self-learning algorithms and AI will result in further encroachment on the already shrinking field of personal privacy.
Government and business are eagerly resorting to the use of AI for profiling citizens and customers. This involves (deep/self-learning) algorithms as the basis of the system fed with (big) data from which analyses lead to results that form the basis for ultimate decision-making.
Well-known examples of the use of AI include facial recognition and license plate registration, as well as more far-reaching population surveys based on background and behavioral data of citizens. Systems such as SyRi (System Risk Indication), in which the government links citizens' personal data to detect various forms of fraud, abuse and violations, seem to overshoot their mark.
Fraud investigations in earmarked districts give the impression of already starting on a discriminatory basis. A limited number of municipalities have received ministerial approval for such investigations: Eindhoven, Capelle aan den IJssel, Haarlem and Rotterdam. Citizens are not informed and are a priori suspect, which is how the actions are perceived. In Rotterdam, based on SyRi research, 1,200 addressees were placed in the "Risk Notification Register" because, according to SyRi, they had an increased risk of fraud or other crimes. Meanwhile, Mayor Aboutaleb put a stop to the SyRi experiments in that city himself. He considered the application of SyRi disproportionate with regard to the infringement on the privacy of citizens, especially since the information obtained was linked to secret police information. Eindhoven also stopped the experiments.
Although the government is supposed to handle the personal data of its citizens with care, the requirements set forth in the AVG for the processing of personal data, whereby processing must be lawful, proper and transparent, do not always seem to be well observed. Indeed, the data subject (that is, the person whose personal data is being processed) has the right to know the logic behind the processing. In addition, that person has the right not to be subjected to profiling that leads to a decision with legal consequences.
The problem is that actual implementation of the requirements laid down in the AVG is difficult. The required transparency will almost never (be able to) be disclosed when operating algorithms in practice. Apart from the fact that that operation is not understandable to virtually anyone, the enforceability of those rules by the government will be virtually impossible anyway. In addition, the risks of disclosing those algorithms will be far too great. Trade secrets, copyright violations and invasion of privacy by less well-intentioned people will stand in the way. The government itself also has no inclination to make the algorithms used transparent. It was only after a WOB request by the NOS that various government departments wanted to provide insight into the way in which artificial intelligence is applied by the government, for example. (1)
The application of artificial intelligence by government can certainly be applied positively and take some of the workload off the government's hands. But care should be taken not to rely on it completely and, in addition, creeping discrimination through selective data analysis should be avoided as part of that application. Carefulness, legality and proportionality in both the design and application of AI systems is a "conditio sine qua non", especially for the government.
(1) https://app.nos.nl/datavisualisatie/2018/algoritmen/index.html
This article can also be found in the Digital Transformation and Big Data dossier
