How do people choose what book to read? At times, people choose to exploit well-known options that are likely to lead to high enjoyment. However, readers must also effectively explore novel books in order to reduce uncertainty in a way that allows for efficient learning about less-known alternatives that might also lead to high enjoyment. It is unknown precisely how and why readers make these sequential book selection decisions. By placing book options in a semantic embedding space, we show that people decide which book to explore using a structured generalization mechanism based on semantic similarities between known and unknown books and a directed exploration mechanism that incentivizes seeking books in high uncertainty. In addition, we demonstrate that people’s directed and random book exploration patterns are modulated by individual differences in curiosity, which fosters reading enjoyment and promotes exploring unfamiliar books. In summary, our study demonstrates that these exploration mechanisms guide readers' decisions about what books to read in a way that leads to high levels of enjoyment.

Cognitive control is an important framework for understanding the neuropsychological processes that underlie and enable the successful completion of everyday tasks. Only recently has research in this area investigated motivational contributions to control allocation. An important gap in our understanding is the way in which intrinsic rewards associated with a task motivate the sustained allocation of cognitive control. In three behavioral and one functional magnetic resonance imaging studies, we use a naturalistic and open-sourced simulator to show that changes in the balance between task difficulty and an individual’s ability to perform the task result in different levels of intrinsic reward, which motivates dynamic shifts between networked brain states. Specifically, high levels of intrinsic reward associated with a balance between task difficulty and individual ability are associated with increased connectivity between cognitive control and reward networks. By comparison, a mismatch between task difficulty and individual ability is associated with lower levels of intrinsic reward and corresponds to increased activity within the default mode network. Insular activation suggests that motivational salience, as defined by the level of intrinsic reward, drives shifts between networked brain states associated with task engagement or disengagement. These results implicate reward processing as a critical component of cognitive control.