During flexible behaviour, dopamine is thought to carry reward prediction errors (RPEs), which update values and hence modify future behaviour. However, in real-world situations where the statistical relationships in the environment can be learned, continuously adapting values is not always the most efficient way of adapting to change. In such partial observable structured environments, as is found in many real-world situations, it is not well understood what kind of information dopamine conveys or its causal role in shaping adaptive behaviour. Here, I’ll describe a set of studies where we measured and manipulated dopamine while mice performed a sequential decision task. I’ll show that, while dopamine tracks a wide range of information, shaped by the inferred state of the task, it does not carry the key information that animals are using to rapidly update their choices.