The Italian Supreme Court has provided important clarifications on algorithmic transparency, regarding automated decision-making processes also carried out through profiling and AI systems.
In a well-known case concerning a platform for the automated determination of the “reputational” rating of users, the Court specified the necessary steps to ensure transparency in the processing of personal data through automated decision-making processes, which are also necessary for the collection of valid privacy consents.
Companies are not required to provide users with information about the algorithm expressed in mathematical language, which, moreover, is unrelated to the need to protect confidential information. In order to obtain valid consent (in particular, in compliance with the GDPR requirements of “freedom” and “specificity”), users need to be able to know the algorithm as a reliable process for obtaining a given result or solving a given problem within a finite time, based on an unambiguous description; and anyway, detailed.
In other words, it is important to ensure that the process leading to the expected result is understood, and that the execution scheme in which the algorithm is expressed is as transparent and comprehensible as possible, with reference to the functional characteristics of the algorithm.
The decision of the Supreme Court, at least in part, anticipated the one subsequently received by the EU Court of Justice, in a case of central importance for the issue of the explainability of automated processes, in which it was affirmed, with reference to the obligations of the “controller” by Art. 22 GDPR, that the need to provide ‘meaningful information on the logic used’ in the automated process requires explanations, through relevant and concise, transparent, understandable and easily accessible information, on the procedure and principles applied in practice to use, by automated means, personal data relating to a particular person for the purpose of obtaining a specific result.
Articoli Correlati