News

Prof. Dr. Thomas Krödel von der Universität Hamburg wird als Teil des "Philosophy Research Forum" am 27.01.2026 von 16:15–17:45 Uhr in S 110 (INF) einen Vortrag zum Thema "Causal Complications for Explainable AI" halten.

Abstract: How can we explain the behaviour of AI agents such as large language models? A natural approach is to explain their behaviour causally by identifying internal representations of the system and their causal relations to the system’s output. It turns out, however, that the nature of these representations makes it difficult to determine their causal roles. What seems to be our best theory of causation, namely the interventionist theory, delivers incorrect verdicts about these roles, and there is no quick fix.