Trust in AI Models
Mohsen Abbaspour Onari studied how explainable AI influences user trust in decision support systems, combining technical metrics with human-centered evaluations.
Mohsen Abbaspour Onari, PhD researcher in the Information Systems (IS) group, defended his dissertation on November 4 at the Department of Industrial Engineering & Innovation Sciences. His research explores how explainable AI affects human trust in decision support systems, especially in sensitive domains like healthcare.
Societal Impact
Artificial Intelligence is increasingly embedded in daily life, from chatbots to medical diagnostics. While these systems offer efficiency and innovation, they also raise concerns about bias, misinformation, and transparency. Entrepreneurs and regulators face the challenge of designing AI systems that not only perform well but also earn public trust. Abbaspour Onari鈥檚 work responds to this need by examining how explanations influence user trust behavior in AI decisions.
Measuring Trust
The research combines objective metrics with subjective user studies to evaluate AI explanations. In one study, medical experts assessed the interpretability of an AI model to diagnose distal myopathy, revealing how technical clarity and human perception interact. Another study focused on COVID-19 cases, modeling how users' trust perception in an AI model. The third study clearly separated perceived trust, which reflects what users say, from demonstrated trust, which shows whether they delegate the decision to AI or not.
Design Implications
Abbaspour Onari鈥檚 findings show that trust in AI is not built by explanation alone. It depends on how well those explanations align with user expectations, domain knowledge, and cognitive biases. For developers and policymakers, this means that building trustworthy AI requires interdisciplinary insight, combining technical design with psychological understanding.
Broader Relevance
As AI continues to shape decisions in healthcare, finance, and public services, understanding trust becomes essential. This research offers tools and insights to guide the development of decision support systems that are not only transparent but also socially responsible. It encourages a shift from algorithmic performance to human-centered design.
Mohsen Abbaspour Onari defended his thesis on November 4. Title of the thesis: . Supervisors: Yingqian Zhang, Marco Nobile, and Isel Grau Garcia.