Explaining Explainability – the podcast on Explainable Artificial Intelligence.

At TRR 318, scaffolding is recognized as a crucial aspect of explaining. It involves supporting learners, whether they are children acquiring their first words or robots interacting with caregivers. In the fifth episode of the podcast "Explaining Explainability," Prof. Britta Wrede engages with experts Dr. Angela Grimminger from Paderborn University and Prof. Dr.-Ing. Anna-Lisa Vollmer from Bielefeld University. They explore how scaffolding functions in various contexts, such as explanatory situations, and its role in their TRR 318 projects.

As TRR 318 understands it, explanations are not delivered in a one-way process - they are co-constructed, by the explainer and the explainee. And it seems that Large Language Models (LLMs) are pretty good at doing just that - but do they really co-construct? And what is co-construction? Prof Britta Wrede discusses these questions with two experts on LLMs, Prof Axel Ngonga Ngomo from Paderborn University and Prof Henning Wachsmuth from Leibniz Universität Hannover.

How do you know when someone has understood something? And how can explainers adapt their approach to promote better understanding? In this episode, Prof. Britta Wrede discusses these questions with Prof. Hendrik Buschmeier, a computational linguist at Bielefeld University, and Prof. Heike Buhl, a psychologist at Paderborn University. (Episode in German; English transcript available.)

What is the difference between explainability and explaining? How does the act of explanation, done by humans, work – and how can it be applied to AI systems? Why is it even of such great importance for XAI? Prof. Britta Wrede is discussing these and other questions again with Prof. Katharina Rohlfing, psycholinguist at Paderborn University and Prof. Philipp Cimiano, computer scientist at Bielefeld University.