We are pleased to announce an upcoming talk by Prof. Dr. Martin V. Butz from University of Tübingen, who will be presenting on April 2, 2025, at 15:00.
Abstract:
Research from various cognitive science-related disciplines suggests that our minds actively contextualize our thoughts and behavior. Active contextualization enables us to focus on task- and goal-relevant aspects, while ignoring irrelevant components. As a result, it enables deeper planning, reasoning, and the selection of appropriate interaction routines, while minimizing cognitive effort. I will show active inference-based and deep learning neural network-based modeling approaches that implement key aspects of these principles. I argue that active contextualization is a key to develop a versatile language of thought – the core conceptual middle-ground between language and sensorimotor data – which constitutes the missing piece for a maximally effective, goal-directed inference of solutions – in the form of chains of thoughts – to challenging reasoning tasks.
