Beschreibung
We, as humans, routinely talk about ourselves, but what is this “self”, how does it arise, and what influences it? What are the underlying mechanisms that help humans perceive themselves and act in the world? Recent work studies the so-called minimal self and shows how the sense of body-ownership, agency and control contribute to it. In our research, as part of the DFG SPP “the active self”, we investigate the minimal self and its emergence through sensorimotor experiences of embodied agents by utilizing information theory and reinforcement learning. Focusing on artificial agents enables us to extensively experiment and analyze their internal representations to further understand the minimal self.
Keywords
minimal self
reinforcement learning
information theory
embodied intelligence
machine learning
