Immigration, the democratic use of predictive algorithms
According to the United Nations Population Division, to guarantee the "support ratio", i.e. the number of people in employment per retiree, Germany would need about 3.4 million immigrants a year until 2050. The study used an algorithm to which another question was also posed: how many immigrants does Germany need to keep the number of employed persons constant, regardless of their ratio with respect pensioners? In this case, the answer was 460,000 people.
Reading these data without giving in to the temptation of superficial synthesis requires a preliminary clarification on the elements of the calculation. For example, it must be specified whether the AI is processing the overall migration balance, i.e. the net value coming from the difference between the number of immigrants and emigrants, or instead the total number of entries and, therefore, only immigration flows. In other words, the data of the predictive algorithm can be interpreted scientifically only if the terms of the calculation are understood. Otherwise, the algorithmic operation can be offered to the public loaded with risks linked both in terms of its functioning and its outcome.
The development of a predictive algorithm that works on big data regarding migratory flows is often motivated by a series of concerns related to immigration policy. The prediction of flows has been represented as a solution capable of inducing at least three virtuous effects: allowing public authorities to prepare adequate measures to manage inflows and reception; prevent social tensions usually associated with an increase in immigration in a certain area of the territory; understand the direction of the demographic growth of a given territory and, therefore, understand the consequences of a change in the composition of the population on welfare systems and the labor market.
However, the availability of information of this type could also stimulate the adoption of policies strongly oriented towards the reduction of arrivals and the strategic use of border controls. The effect, in this case, would be to encourage preventive immigration controls, with dramatic consequences for protection of fundamental rights, first of all the right to asylum.
The problem of a "populist risk", i.e. of a knee-jerk political reaction activated by algorithmic predictions, questions constitutional law at several levels. For this reason, it is necessary to explore the opportunities for a democratic use of the predictive tool. The risk of populist use of the predictive algorithm can be averted, in fact, only if citizens maintain their political freedoms and understand, with a reasonable intellectual effort, the topics covered by civic and political debate and the reasons behing political action.
Understanding civic issues that are processed through algorithms requires an adequate level of knowledge about AI. From this point of view, an entirely algorithmic decision-making process is compatible with democracy insofar as it is assisted by guarantees of minimum levels of technological education, as it emerges, moreover, from EU secondary law. Algorithmic decision-making therefore needs to be grounded in a widespread and shared understanding of what AI-based decision-making is. Only in this way can the use of algorithms increase the sense of responsibility of public decision-makers and push them to adopt solutions that share and analyze data with a view to fueling democratic processes. In this framework, research on immigration law can play a crucial role in contributing to a process of orienting algorithmic technology towards the rights and principles of the European constitutional tradition.