What Makes a Good Conversation? Emotions, Confidence and Explainability

Large language models (LLMs) have transformed the field of artificial intelligence, achieving or even surpassing human performance in numerous natural language processing tasks. In her lecture, Prof. Milica Gašić will explore what makes a conversation effective by contrasting LLMs with task-oriented dialogue systems — systems that interface with external knowledge bases and explicitly model the user’s goals.

A particular focus will be placed on the crucial role of reinforcement learning and the impact of reward functions on performance. Prof. Gašić will illustrate different approaches, including reward functions that enable continual learning, emotion-based reward mechanisms, and confidence-driven reward models inspired by human psychology. She will also demonstrate how topological data analysis can enhance the explainability of LLMs and hypothesize how combining LLMs with task-oriented dialogue systems may lead to conversational agents with a broad range of desirable properties.

About the Speaker
Prof. Milica Gašić is Professor of the Dialog Systems and Machine Learning Group at Heinrich Heine University Düsseldorf. Her research focuses on fundamental questions of human-computer dialogue modelling at the intersection of natural language processing and machine learning. Before her current position, she was a Lecturer at the University of Cambridge, where she led the Dialogue Systems Group. Prof. Gašić earned her PhD under the supervision of Prof. Steve Young, receiving the prestigious EPSRC PhD Plus Award for her work on statistical dialogue modelling.

Event Details
Date: September 18, 2025
Time: 10:15 AM – 11:45 AM
Location: Joseph-von-Fraunhofer-Str. 25, Dortmund, Floor 3, Room 303 (Conference Room Lamarr / RC Trust Dortmund)

AI Colloquium
The AI Colloquium is a series of lectures dedicated to cutting-edge research in the field of machine learning and artificial intelligence, coorganized by the Lamarr Institute for Machine Learning and Artificial Intelligence (Lamarr Institute), the Research Center Trustworthy Data Science and Security (RC Trust), and the Center for Data Science & Simulation at TU Dortmund University (DoDas).