Lun
05
Ott
Seminari e Convegni
Who Presses the Button? Human Oversight of AI Between Claim and Practice, tenuto da Niklas Kühl
Lunedì 5 ottobre 2026, ore 16,30, si svolgerà il seminario online "Who Presses the Button? Human Oversight of AI Between Claim and Practice", tenuto da Niklas Kühl, dell'University of Bayreuth and Fraunhofer FIT. L’evento è svolto nell'ambito del ciclo di seminari “Responsible Artificial Intelligence” (RAI).
Abstract
In 1995, Homer Simpson automated his safety job at the nuclear plant with a drinking bird. What was a joke then is now law: Article 14 of the EU AI Act requires effective human oversight of high-risk systems. But the question is not whether a human sits in the loop; it is whether that person can actually disagree. Drawing on studies with 136 lay and expert decision-makers and with 101 radiologists, this keynote shows that explanations only help when their format invites verification, and that oversight is a design property of the system, not a checkbox in a process.
Biografia
Niklas Kühl is a Full Professor of information systems and human-centric AI at the University of Bayreuth, Director of the AI Responsibility Centre (ARC) and Group Lead with Fraunhofer FIT. He received a Ph.D. in information systems with a focus on applied machine learning and his Habilitation in applied computer science at the KIT. In his studies, he is working on conceptualizing, designing, and implementing AI products, focusing on fair and effective collaboration within human–AI teams. He has worked on machine learning (ML) and AI in different domains since 2014 and has published his work in leading outlets in information systems (e.g., EJIS, CAIS, BISE), computer science (e.g., ICML, AAAI, CVPR), and human-computer interaction (e.g., CHI, CSCW, FAccT).
Per partecipare online è possibile collegarsi tramite il seguente link.
Per ulteriori informazioni e per ricevere gli avvisi dei talk, è possibile iscriversi alla mailing list inviando un'email vuota con oggetto "Subscribe RAI" all'indirizzo dedicato.
Abstract
In 1995, Homer Simpson automated his safety job at the nuclear plant with a drinking bird. What was a joke then is now law: Article 14 of the EU AI Act requires effective human oversight of high-risk systems. But the question is not whether a human sits in the loop; it is whether that person can actually disagree. Drawing on studies with 136 lay and expert decision-makers and with 101 radiologists, this keynote shows that explanations only help when their format invites verification, and that oversight is a design property of the system, not a checkbox in a process.
Biografia
Niklas Kühl is a Full Professor of information systems and human-centric AI at the University of Bayreuth, Director of the AI Responsibility Centre (ARC) and Group Lead with Fraunhofer FIT. He received a Ph.D. in information systems with a focus on applied machine learning and his Habilitation in applied computer science at the KIT. In his studies, he is working on conceptualizing, designing, and implementing AI products, focusing on fair and effective collaboration within human–AI teams. He has worked on machine learning (ML) and AI in different domains since 2014 and has published his work in leading outlets in information systems (e.g., EJIS, CAIS, BISE), computer science (e.g., ICML, AAAI, CVPR), and human-computer interaction (e.g., CHI, CSCW, FAccT).
Per partecipare online è possibile collegarsi tramite il seguente link.
Per ulteriori informazioni e per ricevere gli avvisi dei talk, è possibile iscriversi alla mailing list inviando un'email vuota con oggetto "Subscribe RAI" all'indirizzo dedicato.