BIO-AI: The mechanistic and theoretical foundations of exploratory learning in rats

About This Grant

How do you remember where you have been or how to get from one place to another? Although animals are very good at these forms of spatial memory, the mechanisms remain poorly understood. Understanding these mechanisms is important, because the skills they control are crucial for animal survival. Such skills also have practical applications for designing human devices. For example, a robot vacuum should remember where it has and has not cleaned, and it should find its way back to the charger. Yet, remarkable as they may seem, such smart tools are much less capable than animals at spatial memory-dependent tasks. This project seeks to better understand how animals perform such tasks and translate that understanding into advances in artificial intelligence. Importantly, this research focuses on a previously overlooked strategy that animals, rats in this case, often use when navigating: they take a break from doing the navigation task at hand and change posture, seemingly to absorb information from a wider or different angle. Preliminary studies showed that animals that engage in this additional posture shift perform better in navigation tasks. In this project, experiments will be used to determine what information animals absorb during the shift in strategy and how neural activity in the hippocampus--a brain region critical for spatial navigation--changes in association with the behavioral strategy. In combination with the experimental studies, computational models and machine learning will be used to understand the biology while advancing artificial intelligence systems for use in a wide range of devices to make daily lives of humans better. The project will also provide opportunities for students at all levels to gain training at the intersection of neurophysiology and machine learning/artificial intelligence. Spatial memory is well known to depend on intact hippocampal function. Yet, the hippocampus's role in spatial memory encoding remains unclear. The investigators recently found that blocking hippocampal activity during rearing (when rats stand on their hind legs) impairs spatial memory, identifying rearing as a key behavioral epoch for memory encoding. The discovery that rearing is key for hippocampal-dependent spatial memory encoding is significant, as it expands the very short list of behavioral markers linked to critical hippocampal functions. This project builds on this finding by determining: 1) what information rearing provides to update the cognitive map, 2) whether other behaviors elicit similar hippocampal dynamics and if blocking activity during these behaviors affects memory, and 3) whether rearing indicates curiosity-driven information foraging. Addressing these unknowns has implications for multiple fields of study. First, by examining how new information acquired during rearing is integrated into existing cognitive maps, the project uncovers principles of continual learning through active sampling. Second, by investigating what other behaviors elicit hippocampal dynamics like rearing, the project may identify additional exploratory behaviors that support spatial memory and connect them to underlying neurophysiological mechanisms. Third, by testing whether rearing reflects curiosity-driven information foraging, the project evaluates normative models explaining when and why rats choose to rear. The investigators address these questions using a combination of high-density electrophysiology, closed-loop optogenetics, behavioral analysis, machine learning and computational modeling. This project is supported by the Directorate of Biological Sciences Division of Integrative Organismal Systems Neural Systems Cluster Modulation Program and by the Division of Emerging Frontiers. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.