ESI SyNC 2023: September 14 - 15

Linking hypotheses: where neuroscience, computation, and cognition meet

Many nervous systems develop and form functional circuits through a long period of development involving a myriad of mechanisms. Some of these are determined by genes and molecules, while others depend on neural activity patterns. I will present how these diverse mechanisms work together to set up neural circuits shortly after an animal is born, enabling it to gradually acquire its cognitive and behavioral capabilities. I will focus on the visual system, and demonstrate how neural circuits became established and capable of performing different computations. I will focus on some specific mechanisms such as inhibitory synaptic plasticity and the emergence of excitatory and inhibitory balance, and its role in the detection of novel stimuli.

How does the brain represent and store facts such as “Jenny painted her bedroom window yesterday” or “Children like sweets”? Across semantic theories, representation of the corresponding meanings is achieved by assuming a lexicon of predicates (like, paint) that take arguments (children, sweets, Jenny, …) as inputs to yield descriptions of events or states. Psychology and neuroscience provide a different type of answer, where the guiding principle is an association or its neural counterpart, a Hebbian synapse. Such an associative view, while bolstered by much neuroscientific evidence, lacks the means to encode different predicates and fails to provide a satisfactory answer as to how different relations between the same arguments can be maintained. In this talk we resolve the tension by emphasizing another type of neural machinery, most prominently demonstrated in (but not restricted to) the literature on rodent spatial navigation. We argue that cell types such as object cells, boundary cells, head-direction cells, etc. are predicates that enable expressing various abstract meanings, and that a (para-)hippocampal navigational system can be viewed as a model for a more complex system present in human cognition.

Memory research is in the midst of an “engram renaissance” (Josselyn, Kohler, & Frankland, 2017), as new tools make possible a more detailed and precise understanding of the neural mechanisms of memory than ever before. These findings have re-invigorated interest in multi-level, multi-field exploration of the capacity to remember. This revival also provides an opportunity to develop a richer conception of the engram — one that is required, I argue, to continue moving memory research forward. In this talk, I explore how a range of theoretical assumptions — about the nature of cognition, constraints on computation, and features of neural systems — influence how researchers think about the engram. I will attempt to show how making these assumptions explicit can yield not only more productive debates about this fundamental concept, but also lead to more fruitful lines of inquiry.

The reinforcement learning (RL) theory constitutes a framework for an artificial agent to learn actions that maximize rewards in the environment. It has been successfully applied to Neuroscience to account for animal neural and behavioral processes in simple laboratory tasks, such as Pavlovian and instrumental conditioning, and single-step economic decision-making tasks. It moreover became very popular due to its account for dopamine reward prediction error signals. However, more complex multi-step tasks, such as navigation and social interaction tasks, illustrate their computational limitations. In parallel, researches in engineering, in robotics in particular, have emphasized the complementarity between different learning strategies when facing complex tasks, and explored solutions to combine these different strategies. One central distinction is between model-based and model-free reinforcement learning strategies: In the former case, an agent learns a statistical model of the effects of its actions in the environment, and then use this model to plan sequences of actions towards desired goals. In contrast, model-free strategies are relevant when the environment statistics are too noisy to learn a good internal model. In this case, RL agents can rather learn local action values and adapt reactively in each state of the environment. In this presentation, I will show a series of work where we used a coordination of model-based and model-free reinforcement learning to account for a diversity of behavioral and neural observations in humans, non-human primates and rodents in different paradigms: Navigation, instrumental and Pavlovian conditioning. I will moreover present recent robotics results where the same algorithm with the same parameters produces optimal performance in simple navigation and social interaction tasks, with a drastically reduced computational cost compared to classical methods. Finally, I will show how the patterns of mental simulation within such internal models can mimic experimentally observed reactivations of the rodent hippocampus in spatial cognition tasks, and raise new predictions for future experiments.

Bats are extreme aviators and amazing navigators. Many bat species nightly commute dozens of kilometres in search of food, and some bat species annually migrate over thousands of kilometres. Studying bats in their natural environment has always been extremely challenging because of their small size (mostly <50 gr) and agile nature. We have developed novel miniature sensors allowing us to GPS-tag small bats, thus opening a new window to document their behaviour in the wild. We have used this technology to track bat pups over months from birth to adulthood. Following the bats’ full movement history allowed us to show that they use novel short-cuts which are typical for cognitive-map based navigation. Using miniature microphones placed on the bats, we have also inferred and studied their foraging success and social behaviour. This novel technology thus allows us to document and model foraging decision making in real-life large scale over long time periods.

With the internet, society has been exposed to puzzling and exquisite animal behavior, while neuroscience has vastly concentrated on a few inbred animal models studied in trained unnatural settings. Changing this trend is imperative. The onset of wireless recording technologies has allowed the development of more naturalistic inquiries into the brain and behavior of animals. By developing a novel behavioral paradigm in neuroscience, playing 'Hide and Seek' with rats, we were able to study mammalian play and its neural bases. We played 'Hide and Seek' with rats in a 30 square meter room, and found that they acquired the game easily and played by the rules. Rats played strategically and, without being conditioned, developed remarkable game specific vocalizations patterns. The freedom afforded to our subject animals allowed us to probe new forms of linking neural and behavioral phenomena. We reversed the arrow of inquiry by doing unsupervised clustering of neural population data allowing us to blindly refer back to the behavior of the rat during the game and uncover neural-behavioral associations unsuspected by the experimenters. My work in large scale wireless rat play laid the foundation for expanding neuroscience in two future directions, 1) the use of non-traditional rodent models (Deer mice) to understand the evolution of how play contributes to cognition and 2) taking novel neurotechnologies out of the lab to study the brains and behavior of animals in the wild.