AI at the moment of critical decisions

Artificial intelligence (AI) is rapidly transforming healthcare. From supporting diagnoses to assisting with medical decision-making, AI is increasingly becoming part of everyday medical practice. Yet one crucial question remains largely unanswered: How do end users, including patients, experience AI when facing some of the most difficult decisions of their lives?

This question lies at the heart of the research conducted by Zeineb Sassi, a doctoral researcher at the Institute of Epidemiology and Preventive Medicine, Medical Sociology at the University of Regensburg, Germany. With a background in psychology and medical informatics, she combines health services research, patient-centered care, and emerging digital technologies. Her particular interest lies in understanding how patients experience decision-making processes and, increasingly, interactions with AI in complex healthcare settings.

This year, Zeineb Sassi was invited to attend the 75th Lindau Nobel Laureate Meeting, one of the world's most prestigious scientific gatherings, bringing together Nobel Laureates and outstanding young researchers from across the globe. She is among approximately 600 young scientists selected through a highly competitive international process and one of only twelve researchers chosen to present her work and research interests as part of the "Next Generation in Life Science" presentation series.

In her doctoral project, she investigated how AI influences not only medical decision-making but also the emotional environment in which these decisions take place. Her research focused on post-transplant care, where patients are often confronted with life-changing decisions under conditions of uncertainty, anxiety, and psychological distress. Although modern medicine emphasizes shared decision-making between physicians and patients, many individuals still report feeling insufficiently informed and emotionally supported during critical moments.

Ms. Sassi, in your study, you interviewed patients who had experienced AI-based care in kidney transplantation. How did you conduct the study?

As part of my doctoral research, I interviewed kidney transplant patients during their post-transplant follow-up care. The aim was to better understand their perspectives and experiences as end users interacting with AI-based decision support systems (AI-DSS) and to gather their feedback regarding their interactions with this tool.

How does AI support this group of patients?

AI supports both patients and physicians by calculating individualized risk predictions based on large amounts of clinical data and presenting these predictions in an understandable format. For example, risks of organ rejection, infection, or other complications following kidney transplantation can be estimated more accurately. Visualizing these predictions through graphical risk curves can help identify changes in a patient's condition at an early stage, facilitate discussions about treatment options, and support more informed shared decision-making during follow-up care. Importantly, the final decisions remain in the hands of physicians and patients.

How many patients did you interview?

I interviewed 36 kidney transplant patients. All participants were receiving follow-up care in which an AI-based decision support system was being used. The system is a machine learning–based predictive model that estimates the probability of graft loss within the next 90 days and presents this prediction through a graphical risk curve.

What exactly were you interested in investigating?

I was interested in a question that has received surprisingly little attention in AI research so far: What happens to physician-patient interactions when AI becomes part of medical decision-making? At the center of the study was therefore human-AI interaction in one of the most sensitive and high-risk areas of medicine: the follow-up care of kidney transplant patients.

Did your doctoral project deliberately incorporate AI? Is this already common practice in such situations? Did patients still have the usual conversations with their physicians?

The doctoral project accompanied the implementation of an AI-based decision support system in transplant medicine that was developed by the German Research Center for Artificial Intelligence (DFKI) in Berlin. The system analyzes clinical data and calculates an individual's risk of graft loss. During follow-up appointments, these risk predictions were presented graphically to both physicians and patients. Such systems are increasingly being developed and tested, particularly in high-risk medical fields where decision-making is associated with substantial uncertainty and pressure. Importantly, the AI did not replace physicians. Instead, I was able to observe how interactions changed when AI became an additional source of information present in the consultation. In a sense, a new form of collaboration emerged among patients, physicians, and the AI system.

What was the most important finding of the doctoral study?

The most important finding was that AI influences not only clinical decisions but also social and relational processes, altering the roles of both patients and physicians with the introduction of AI. Many patients trusted AI predictions because they believed that AI systems could analyze vast amounts of information that would be difficult for humans to process. At the same time, new uncertainties emerged. What happens when the AI's assessment differs from the physician's judgment? And who ultimately bears responsibility for a decision?

Particularly interesting was the finding that acceptance of AI depended less on its technical performance and more on how it was integrated into care. When physicians discussed AI-generated results with patients using a clear and user-friendly interface, patients reported greater understanding and involvement. In contrast, when results were mentioned only briefly or not explained, some patients felt excluded from the decision-making process.

What surprised you most?

I was surprised that many patients sometimes perceived AI as more attentive or more knowledgeable than humans- not because they considered it more empathetic, but because they believed it could process substantially more information simultaneously.

At the same time, it became very clear that patients placed great importance on the human aspects of medicine, particularly physicians' ability to consider individual life circumstances and personal experiences. The study suggests that human-AI interactions should be designed in ways that strengthen rather than replace human relationships in healthcare.

What role could your findings play in the future use of AI in patient care?

The findings demonstrate that the future of medical AI depends not only on algorithms but also on how these systems interact with human users. As AI becomes increasingly integrated into high-risk medical settings, systems must be not only technically reliable but also implemented in an ethically responsible manner. I approach this from a human-AI interaction perspective: the critical question is not only what AI calculates, but how its outputs are communicated, interpreted, and incorporated into clinical practice.

For healthcare practice, this means that physicians need support in explaining AI-generated information in ways that patients can understand and discuss. For developers and interface designers, it means that transparency, accountability, and patient involvement must be considered from the outset.

The central challenge of digital medicine is therefore not simply to use AI as a tool, but to design it as an integral component of human-AI interaction. As AI increasingly influences clinical reasoning and decision-making processes- particularly in high-risk areas of medicine- it becomes essential to clearly define how AI should be designed, when AI should be used, and how human responsibility, judgment, and patient participation can be preserved.