Talk title: Model-Based and Model-Free Contributions to Directed and Random Exploration Strategies

Abstract

In the realms of psychology and neuroscience, two divergent yet crucial research trajectories delve into how humans navigate the delicate balance between exploration and exploitation, as well as the dichotomy of goal-directed versus habitual control. On one front, addressing the exploration-exploitation dilemma reveals two robust families of strategies: directed exploration, biased toward gaining valuable insights and resolving uncertainties, and random exploration, characterized by heightened behavioural variability or ‘decision noise.’ On another front, the study of dual-control processes uncovers the reliance on two distinct mechanisms: the swift and reflective Model Free control, driven by past successes and the ‘law of effect,’ and the more deliberate Model-Based control, leveraging a cognitive map to plan actions and predict future rewards. However, a critical gap persists in our understanding, as the relationship between exploration strategies and control processes remains unexplored. This lacuna in knowledge hampers the development of targeted interventions to enhance exploration. In this presentation, I unveil a novel task designed to dissect aspects of directed and random exploration within both model-based and model-free control. Additionally, I present experimental findings showcasing a distinctive ‘division of labour,’ wherein directed exploration aligns with model-free control and random exploration aligns with model-based control. These findings pave the way for a richer understanding of the control processes guiding distinct exploration strategies.