Designing AI for Hybrid Decision-Making or...Why Assessing Appropriate Reliance Matters

To optimize human-AI collaboration, particularly in high-stakes fields like medicine, it is crucial to assess "appropriate reliance"—a user's ability to trust AI when it is right and distrust it when it is wrong. Traditional AI performance metrics like accuracy are insufficient, as real-world results are often inconsistent due to cognitive biases such as automation bias (over-relying on incorrect AI) and self-anchoring bias (distrusting correct AI). To address this, the talk introduces a framework and an online tool, Metimeter, which analyzes user "reliance patterns" to quantify appropriate reliance. This assessment helps generate a "protocol diagram," a nomogram that identifies the most effective human-AI interaction protocol (e.g., Traditional, Displacement, Inhibition) for a specific team and context. The core finding is that no single protocol is universally superior; the optimal choice is context-dependent. By periodically evaluating a team's appropriate reliance, AI systems can be configured to achieve the best and safest outcomes in situated decision-making settings.

Speaker: Federico Cabitza (University of Milano-Bicocca)

Biography
Federico Cabitza is an Associate Professor at the University of Milano-Bicocca, where he teaches Human-Computer Interaction and Decision Support Systems in several master’s programs. He leads the MUDI Laboratory and serves as director of the local node of the national “Informatics and Society” laboratory within the CINI consortium. Since 2016, he has collaborated with various hospitals, including the IRCCS Galeazzi Sant’Ambrogio Hospital in Milan, with which he holds a formal affiliation. He is also a co-founder and scientific director of the university spin-off Red Open srl, dedicated to AI impact assessment. He has authored more than 250 scientific publications and has been ranked for several years among the world’s top 2% of scientists, according to the Stanford University database. He is also co-author, with Luciano Floridi, of Intelligenza Artificiale. L’uso delle nuove macchine (Bompiani, 2021).

Online seminar Designing AI for Hybrid Decision-Making or... Why Assessing Appropriate Reliance Matters, held by Federico Cabitza.
The seminar will be held on September 22 at 4:30 PM, online at this link.

For more information contact professor Daniele Quercia.