Abstract
Dieci principi per un uso responsabile degli strumenti di rilevamento AI
La diffusione dei modelli di intelligenza artificiale generativa ha spinto molte istituzioni accademiche e di ricerca ad adottare strumenti automatici di rilevamento dei contenuti generati artificialmente. La letteratura indipendente documenta tuttavia limiti significativi: accuratezze reali inferiori alle dichiarazioni commerciali, bias sistematici contro popolazioni marginalizzate e opacità dei meccanismi decisionali. A ciò si aggiungono implicazioni rilevanti sul piano della sovranità digitale e della protezione dei dati. Il contributo propone un framework operativo in dieci principi, raggruppati in tre aree — cautela tecnologica, tutela dei diritti e trasformazione delle pratiche valutative — rivolto alle istituzioni accademiche e di ricerca italiane, che bilanci la tutela dell'integrità della produzione intellettuale con la protezione dei diritti individuali.
Presentazioni
- DOCUMENTO
Video
Bio
GianMarco Schiesaro works at the Central Directorate for Information Technology of the Italian National Institute of Statistics (Istat), where he focuses on developing IT skills and on innovation projects related to artificial intelligence. He has gained significant experience in the field of International Development Cooperation, with a focus on the design and management of institutional training programmes, particularly in the context of developing countries and the adoption of innovative educational technologies. He has been a member of the Board and of the Scientific Committee of the Master’s programme in Peace Studies at Roma Tre University, of the Master in Cooperation and Development of the CDN Consortium in Pavia, and of the MICAD Master’s programme at Bethlehem University (UNESCO Chair, Palestine), with which he has carried out research activities in the field of ICT for Development.
Abstract
Ten principles for the responsible use of AI detection tools
The spread of generative artificial intelligence models has led many academic and research institutions to adopt automated tools for detecting AI-generated content. However, independent literature highlights significant limitations, including real-world accuracy levels that fall short of commercial claims, systematic biases against marginalised groups, and a lack of transparency in decision-making mechanisms. These issues are further compounded by important concerns regarding digital sovereignty and data protection. This contribution proposes an operational framework based on ten principles, organised into three main areas: technological caution, rights protection, and the transformation of assessment practices. The framework is intended for Italian academic and research institutions and aims to balance the safeguarding of intellectual output integrity with the protection of individual rights.
Attribuzione 4.0 Internazionale