Profili costituzionali della amministrazione algoritmica

Data has made machines smart. This revolution also affects and involves the exercise of public powers. Artificial intelligence, especially in its most recent versions, not only enables the elaboration of cognitive models of whatis happening; on the basis of machine learning systems, it also allows for assessments and predictions. If the public administration decides to make use of these technologies, the constitutional and legal implications would be enormous. Is it possible to conceive a public administration that is «exclusively» algorithmic? The issue of «robotic» administrative decisions is one that is already relevant to everyday life, given that in the last three years, numerous judicial decisions on the subject have been issued. The European legislator does not appear to exclude this possibility (GDPR, Article 22(2)), subjecting it simply to the condition of consent; on the other hand, the perspective ofItalian administrative courts on «exclusively» automated administrative deci-sions appears to be much more cautious and oriented by constitutional principles. The Italian Constitution and the legal protection of rights require non-impersonal (i.e. voluntary), knowable, and — above all — reasoned decisions. Decisions of which it is always possible to know the reasons, so that citizens can validly defend themselves.