PhD
Chair of Artificial Intelligence and Machine Learning at LMU
Institute of Informatics
Akademiestr. 7
80799 München
Biosketch
Asma holds a B.S in Mathematics degree from University of Massachusetts Boston. After finishing her degree she worked in IT health care industry for about 3 years; her experience was mainly pivoted around data analytics to enhance the overall well-being of a defined group of individuals. Asma pursued M.S in Machine Learning from Mohammed Bin Zayed University of Artificial Intelligence to further the passion in the field; her master’s thesis was revolving around learning from noisy labels and uncertainty. She is currently a PhD student in LMU under the supervision of Eyke Hüllermeier to gain a deeper understanding in the domain.
relAI Research
Beyond Explanations: Quantifying Uncertainty in XAI
Artificial intelligence now pervades high-stakes decision-making across science, industry, and society. As models grow in complexity, however, their internal reasoning becomes increasingly opaque. Explainable AI (XAI) aims to address this by offering interpretable insights into model behavior, supporting accountability and human-AI trust. Yet explanations themselves are frequently unstable and sensitive to minor perturbations, stemming from variability in data, model parameters, or the explanation technique itself. Without a principled understanding of where this uncertainty originates, or reliable methods to communicate it, explanations risk becoming misleading and eroding the very trust they are meant to build.
We aim to focus on Uncertainty in Explainable AI (UXAI), formalizing how uncertainty propagates through the explanation pipeline, from data and model to final output. We want to explore this particularly in LLMs and other modalities of data.
Publications
https://epubs.siam.org/doi/abs/10.1137/1.9781611978520.9#afterNav-sdj