Lecture Series “Robust AI”: Claudia Müller-Birn

The next lecture in our lecture series will be given by Prof. Dr. Claudia Müller-Birn (FU Berlin) on the topic of “Designing for Human Value: A Human-Centered Perspective on AI System Design in Healthcare”. This talk ist hosted by the Human-Centered AI Lab of Prof. Dr. Hanna Drimalla.

When & where:

Thursday, January 23, 2025, 2 pm at Bielefeld University (Room CITEC 2.015) or online. Join us via Zoom https://uni-bielefeld.zoom-x.de/j/65573907891?pwd=EpbTNEgRMzVzJds7EDSvXrhbe9W9E0.1

Abstract:

The increasing presence of Artificial Intelligence (AI) in healthcare holds great promise for improving physician decision-making and patient outcomes. However, current AI systems are often evaluated using performance metrics (e.g., accuracy), which may not align with human values in specific contexts of use. In this talk, I present a human-centered perspective on AI system design, emphasizing the importance of understanding human values and needs in the design process. To achieve this, I focus on three key areas: (1) designing for human-AI collaboration, (2) value-centered dataset creation, and (3) participatory design of AI systems.

In the first area, we conducted a series of studies to investigate how the design of AI systems, particularly the provision of uncertainty representations, impacts human decision-making in treatment decisions. Our findings suggest that uncertainty representations alone do not increase human-AI team performance, but when combined with guidance, they can enhance outcomes. In the second area, we explored the decision processes involved in medical data donation and investigated the role of reflection prompts on decision-making. Our results show that a value-oriented reflection prompt supports value congruence, especially for individuals with high privacy concerns. Finally, we examined the potential of participatory approaches to design AI systems by prioritizing human values and needs. I provide several examples that illustrate how patients, clinicians, and researchers can co-design AI systems.

By combining qualitative and quantitative research methods, we aim to provide a more comprehensive understanding of the complex healthcare context and demonstrate that a human-centered approach to AI system design leads to meaningful insights into patient values and needs.

Speaker:

Claudia Müller-Birn is the head of the Human-Centered Computing research group at the Institute for Computer Science at Freie Universität Berlin. Prior to joining Freie Universität Berlin, she was a postdoctoral fellow at Carnegie Mellon University School of Computer Science, supported by a Feodor Lynen Fellowship from the Alexander von Humboldt Foundation. In her interdisciplinary research projects, she focuses on human-AI collaboration, value-led dataset creation, and the participatory design of data-driven systems.