News

Prof. Dr. Jakob Schoeffer von der Universität Groningen wird am wird als Teil der "Bayreuth Lectures" 20.01.2025 von 18:15–19:45 Uhr in H15 (NW I) einen Vortrag zum Thema "Challenges and Opportunities in Human-AI Collaboration" halten.

Abstract: Human-AI collaboration seeks to combine the unique strengths of both humans and artificial intelligence, while mitigating their individual limitations. Typically, this involves AI systems providing decision recommendations, with humans retaining the ability to override them. A central goal is to ensure appropriate reliance, meaning that humans should only override AI recommendations when they are incorrect. However, existing research indicates that achieving this is difficult. In this talk, I will share findings from a large study on the effectiveness of popular explainable AI techniques designed to foster appropriate reliance. I will show that, in some cases, explainable AI can unintentionally influence human decision-making in ways that hinder appropriate reliance. Drawing from these results, I will discuss the current challenges in human-AI collaboration and explore potential strategies to address them.