Periodic Reporting for period 4 - GEOCOG (Cognitive Geometry: Deciphering neural concept spaces and engineering knowledge to empower smart brains in a smart society)
Reporting period: 2021-11-01 to 2023-04-30
The fundamental question in cognitive neuroscience—what are the key coding principles of the brain enabling human thinking—still remains largely unanswered. Insights come from one of the most fascinating discoveries in neuroscience, the Nobel Prize-awarded identification of spatially responsive cells in the rodent brain, in a region called the hippocampal formation. So-called hippocampal place cells, and grid cells in the nearby located entorhinal cortex, signal—in concert with other spatially tuned cells—position, direction, distance and speed. Thereby they provide an internal spatial map, the brain’s SatNav, the most intriguing coding scheme in the brain outside the sensory system. However, these brain mechanisms and in particular the role in higher level cognition are poorly understood in humans. Our framework is concerned with the key idea that this navigation system in the brain—potentially as a result of evolution—provides a fundamental neural metric for human cognition. Specifically, we propose that the brain represents experience in so-called ‘cognitive spaces’. For illustration, consider the simple example of describing cars, which you might do along two dimensions, their engine power and their weight. Depending on the two features, racing cars, for instance, would occupy a region characterized by high power and low weight, whereas campers by low power and high weight. We test the overarching hypothesis that—akin to representing places and paths in a spatial map—similar coding principles are involved in the formation of such cognitive spaces. In the current ERC project, we aim to understand these brain processes in more detail, especially the mechanisms which underlie the formation of new conceptual knowledge.
In the first phase of the project, we implemented our key experimental framework. In brief, we developed different new cognitive tasks, built new analytical pipelines for our behavioural and neuroimaging data and developed important modelling tools, including encoding models and deep neural network models for the analysis of behavioural as well as neural data. We carried out several experiments to examine the formation of conceptual spaces and to track knowledge acquisition in human participants. To identify the neural building blocks of concept spaces, we developed a series of new psychological paradigms including two-dimensional conceptual spaces, multimodal concept learning tasks and a psychophysical approach to examine temporal coding of cognitive spaces. Furthermore, we also investigated concept space formation while participants physically navigated on an omnidirectional treadmill ‘through’ abstract concepts. Key preliminary findings include the identification of knowledge from physical to conceptual spaces and the successful application of deep neural network models to describe and decode fine-grained behavior from neuronal and imaging data.
We aim to describe key coding principles of the human brain underlying the formation of abstract knowledge. Our goal is to identify some of the central neural principles in the hippocampal-entorhinal system which help us to build new concepts. We also aim to develop AI-inspired models to explain complex human behaviour.