Hosted by the School of Psychological Science
Abstract: How do humans learn and organize information to form abstract knowledge? The hippocampal formation seems to be crucial for many cognitive processes including learning, memory, navigation, and concept formation, and is therefore a strong candidate for a domain-general system for organizing knowledge. But how does it work?
In the first part of my talk, I will show that a domain-general, clustering learning algorithm explains key findings in both spatial and conceptual domains. When the clustering model is applied to spatial contexts, hippocampal place cell and entorhinal grid cell-like representations arise. The same mechanism applied to conceptual tasks lead to representations aligned with conceptual knowledge. In the second part of my talk, I will present a multilevel account of this model to address how the hippocampal formation supports concept learning that bridges from behaviour, to cognitive constructs, down to the level of neurons. Inspired by algorithms that capture flocking in birds, we introduce a novel learning rule with a recurrence mechanism and propose that the hippocampus exhibits “neural flocking", where coordinated activity arises from local rules. The learning rule naturally leads to similarly-tuned neural assemblies, demonstrating how concept representations, concept cells, symbols, and their underlying neural assemblies (e.g., place-cell assemblies) in the hippocampus can form.