LLMs and the Patterns of Human Language Use – Christoph Durt
Large Language Models (LLMs) use enormous amounts of data derived from human language use to generate text that is experienced as meaningful by humans. While they replicate stochastic patterns, however, they thereby also prompt patterns of thought and experience. The presentation investigates the relationship between these patterns.
Christoph Durt
Chris Durt is a philosophical researcher and lecturer at the Technical University of Munich. His main research focus is digital technology and the human mind. Rather than seeing digital technology as a mind by itself, he investigates its intricate interrelation with the human mind and other relevant factors such as language and culture. Chris finds phenomenology especially helpful for this investigation, but also builds on insights from Ancient Greek and Early Modern Philosophy, Nietzsche, Wittgenstein, and others. He engages in long-standing interdisciplinary exchanges with psychology, psychiatry, computational linguistics, and AI science. For more information, please see his personal website – www.durt.de
LLMs, Cognitive Biases, and the Path to Cognitive Humility – Marcin Rządeczka & Maciej Wodziński
We are trying to examine the limitations of LLM-based mental health chatbots, highlighting their susceptibility to cognitive biases and the illusion of cognitive humility. We introduce the Simulated Intellectual Humility Scale (SIHS-LLM) to assess epistemic self-awareness in AI and propose strategies for enhancing metacognitive reasoning to improve the epistemic reliability of LLMs.
Marcin Rządeczka
Marcin Rządeczka, Ph.D. is an assistant professor at Maria Curie-Skłodowska University in Lublin, specializes in cognitive science, computational psychiatry, evolutionary psychopathology, and the philosophy of AI for mental health. His research explores cognitive biases, simulated Theory of Mind in AI, and the epistemic challenges of neurodiverse-friendly artificial intelligence. He has contributed to interdisciplinary studies on the role of AI in mental health, and the evolutionary perspective on cognitive biases in autism and schizophrenia spectra. As the head of the Multimodality Research Laboratory (MultiLab), he integrates philosophical, computational, and psychological approaches to understanding human cognition and AI alignment in therapeutic contexts.
Maciej Wodziński
Maciej Wodziński, Ph. D. is a researcher at the Institute of Philosophy, Maria Curie-Skłodowska University in Lublin, Poland. His research focuses on the philosophy of mental health, with a particular interest in epistemic injustice in psychiatric practice and critical autism studies. His interests also extend to the role of AI in philosophy, especially in using AI to analyze first-person experiences of neurodivergent individuals. His work aims to bridge philosophy, mental health, and AI to develop more just and nuanced approaches to neurodiversity.
Quantifying Borderline Experience – Marcin Moskalewicz
Limits of Formalization. Which Aspects of Human Experience Can Enter AI Models? Marianne Broeker
Computational psychiatry is a quickly evolving discipline that aims to understand psychopathology in terms of algorithmic processes. While cognitive phenomena, especially beliefs or ways of “reasoning”, can more easily be formalized, meaning re-described in mathematical terms and then entered computational models, there is speculation as to whether phenomenology might be formalizable too. Here, we explore the possibility of formalizing and modeling a phenomenological account of schizophrenia and other mental illnesses, using concepts from phenomenology, such as of “minimal self” and “intentionality”. We then try to apply the concepts to a computational logic, to test whether an AI can “possess” them. Overall, we are asking via what conditions phenomenology can enter a computational logic.
Marianne Broeker
Marianne D. Broeker is a 3rd year PhD student at the Department of Experimental Psychology at the University of Oxford. Her research looks at changes in conscious in psychosis and takes place at the intersection of clinical psychology and philosophy. However, she is also very interested and in the philosophy of psychiatry and social sciences related to psychiatry, as well as political activism. She has written about and critiqued computational approaches to understanding psychiatric categories and other complex psychological phenomena. Next to her academic work, she works as a existential-analytic psychotherapist in a practice in London.
An gentle introduction to computational (neuro) phenomenology – Maxwell J. D. Ramstead
This presentation gently introduces the field of computational (neuro) phenomenology which is premised on novel advances in computational (generative) modelling. I first present the basic concepts and formal elements of generative modelling. I then examine how one can leverage these advances in the generative modelling of lived experience to further the contemporary project of neurophenomenology.
Maxwell J. D. Ramstead
Maxwell J. D. Ramstead is Chief Science Officer at Noumenal Labs and an Honorary Research Fellow, Queen Square Institute of Neurology, University College London, where he works closely with Karl Friston. Ramstead’s research focuses on on the free-energy principle, Bayesian mechanics, multiscale active inference, and computational phenomenology. Since 2020, Ramstead has moved moved into the field of artificial intelligence, both in academic research and industry.