Neuroscientists at the Sainsbury Wellcome Centre and Gatsby Computational Neuroscience Unit at UCL have discovered that exploratory actions play a crucial role in animal learning. Their findings have implications for the development of artificial intelligence (AI) agents that can learn more efficiently with less experience. The researchers conducted experiments to test the hypothesis that purposeful actions, such as darting towards objects, are important for animals to learn how to navigate their environment. They found that preventing animals from performing exploratory runs hindered their ability to learn, indicating that these instinctual actions are instrumental in building a cognitive map of the world. The team also explored different models of reinforcement learning to understand the algorithms the brain might be using. They identified two main classes of reinforcement learning models – model-free and model-based – and observed that mice exhibit both types of behavior under different conditions. This insight into learning algorithms can inform the development of AI agents that can learn more efficiently. The researchers also aim to investigate the connection between exploratory actions and the representation of subgoals in the brain. By studying which areas of the brain are involved in representing subgoals and how exploratory actions lead to the formation of these representations, they hope to further enhance our understanding of animal learning. The research was funded by Wellcome and the Gatsby Charitable Foundation.