07.05.2026 IfI Colloquium: Human-Centric Machine Learning: Transparency, Explainability, and Guidance by Human Feedback
Speaker:
Prof. Dr. Matthias Zeppelzauer, Computer Science and Security, University St. Pölten, Austria
Date: Thursday, 7 May 2026, 17:15
Location: room BIN 2.A.01 at the Department of Informatics (IfI), Binzmühlestrasse 14, 8050 Zürich
Details about the format of the talk shall be checked always just ahead of a certain presentation date: (information here)
Abstract
Many machine learning models in use today exhibit black-box behavior: their internal mechanisms and decision-making processes are not readily interpretable by humans. Moreover, these models are typically user-agnostic—despite being trained on vast amounts of data, they rarely account for the specific needs, goals, or constraints of human users. In this talk, I will present methodological advances from my group on human-centric machine learning, with a particular emphasis on (i) constructing transparent (white-box) classification models, (ii) developing explanation techniques for complex deep learning architectures, and (iii) incorporating user feedback and explanation-based supervision into training to steer model behavior, reduce bias, and improve alignment with user requirements. The talk will illustrate these ideas through a range of use cases, including natural language understanding, image classification, and medical time-series analysis.
Bio
Matthias Zeppelzauer is a professor and head of the Media Computing Research Group at the University of Applied Sciences St. Pölten in Austria. He received his PhD in Computer Science from TU Wien in 2011 with highest distinction. In 2020, he completed his habilitation at TU Wien in Computer Science on Retrieval of Multimodal Media Data. His research focuses on computer vision, machine learning, visual analytics, and multimedia information retrieval. Focus topics of his ongoing research include explainable and trustworthy machine learning, multimodal machine learning, collaborative machine learning as well as social media retrieval. He thereby pursues an interdisciplinary approach to research and investigates machine learning problems in fields such as the social sciences, medicine, biology, and in the humanities. Matthias was involved intensively in the acquisition and execution of numerous basic and applied research projects at national and international levels and has contributed to raising more than 9.8 million Euro of third-party funding. He is a lecturer for undergraduate and graduate programs and a mentor for younger researchers. He was awarded by the Austrian Computer Society for outstanding achievements in the area of pattern recognition as well as with the Austrian Open Source Award.