Trust in Algorithms: How Does It Evolve?
Patricia Kahr earned her PhD with research on how trust and reliance in AI systems dynamically evolve across time and context, vital for responsible algorithm use.
Patricia Kahr, PhD researcher in the Human Technology Interaction (HTI) group, defended her dissertation on December 9 on trust and reliance in human interactions with AI systems. Her work provides insight into how these processes evolve over time and across contexts, a question that is increasingly important as algorithms play a role in critical decisions.
Can I trust an algorithm?
Algorithmic systems assist with everyday decisions, from planning delivery routes to assessing legal cases and supporting medical diagnoses. Yet trusting these systems is not always straightforward. People often wonder: when can I trust an algorithm, and when should I remain cautious? This uncertainty can lead to poor decisions or ignoring valuable advice.
Complex task
Organizations and society face a complex task: ensuring that people use algorithms responsibly. Kahr鈥檚 research shows that trust and reliance are not static traits but dynamic processes. They change through repeated interactions, shifting contexts, and system performance. This means that AI policy and design must account for these developments.
Key Insights
Kahr examined how prior experiences with algorithms influence future decisions and how errors - from occasional mistakes to fluctuating accuracy - affect the recovery of trust. Through experiments and field studies in domains such as logistics and legal decision-making, her work demonstrates that trust is not just about technology but also about human needs, their values and decision-making context.
Patricia Kahr defended her thesis on December 9. Title of the thesis: Supervisors: Chris Snijders, Gerrit Rooks, and Martijn Willemsen.