| Taken at face value, the world is complicated and confusing. When operating in such complexity, we are greatly advantaged by our ability to infer the underlying structure of the world -- that is, the relationships between our observations and the underlying latent causes that generate them. At any given time, inferring the latent causes that are currently active -- i.e., the current situation -- allows us to execute the most appropriate actions and cognitive processes. In theories of episodic memory, this definition of the current situation is related to the cognitive constructs of "schemas" and "context". In reinforcement learning and decision-making, representations of the current situation are called the "state". In this work, I begin to uncover the computations and neural mechanisms that underlie our inference of the causal structure of the world, including inferences of the current situation, and also how the inferred situation affects decision-making and memory. Throughout this work, the overlapping brain areas of ventromedial prefrontal cortex (vmPFC) and orbitofrontal cortex (OFC) play a prominent role in the neural circuits that perform this inference.;In the first experiment, I show that overall levels of activity in the OFC are related to learning about one type of causal structure -- transitions between states of the world. In the second experiment, I present evidence that the OFC represents a belief distribution (a posterior probability distribution) over the underlying situation. In the third experiment, I present evidence that, in accordance with current theories of episodic memory and temporal context, memories seem to be organized according to information in the brain about the semantics of recent experience, which may serve as a heuristic proxy for the current situation. |
